Serving Substance Abuse Professionals Since 1993

Last Update: 08.09.11

C O R K O N L I N E

powerpoint presentations
CORK database search
resource materials
bibliographies
clinical tools
user services
newsletters
about cork
home

CORK Bibliography: Research, Data Analysis

92 citations. January 2007 to present

Prepared: September 2011

Ahmed SH; Graupner M; Gutkin B. Computational approaches to the neurobiology of drug addiction. Pharmacopsychiatry 42(Supplement 1): S144-S152, 2009. (77 refs.)

To increase our understanding of drug addiction - notably its pharmacological and neurobiological determinants - researchers have begun to formulate computational models of drug self-administration. Currently, one can roughly distinguish between three classes of models which all have in common to attribute to brain dopamine signaling a key role in addiction. The first class of models contains quantitative pharmacological models that describe the influence of pharmacokinetic and pharmacodynamic factors on drug self-administration. These models fail, however, to explain how the drug self-administration behavior is acquired and how it eventually becomes rigid and compulsive with extended drug use. Models belonging to the second class circumvent some of these limitations by modeling how drug use usurps the function of dopamine in reinforcement learning and action selection. However, despite their behavioral plausibility, these latter models lack neurobiological plausibility and ignore the potential role of opponent processes in addiction. The third class of models attempts to surmount these pitfalls by providing a more realistic picture of the midbrain dopamine circuitry and of the complex action of drugs of abuse in the output of this circuitry. Here we provide a brief overview of these different models to illustrate the potential contribution of mathematical modeling to our understanding of the neurobiology of drug addiction.

Albert PS; Follmann DA. Random effects and latent processes approaches for analyzing binary longitudinal data with missingness: A comparison of approaches using opiate clinical trial data. Statistical Methods in Medical Research 16(5): 417-439, 2007. (24 refs.)

The analysis of longitudinal data with non-ignorable missingness remains an active area in biostatistics research. This article discusses various random effects and latent process models which have been proposed for analyzing longitudinal binary data subject to both non-ignorable intermittent missing data and dropout. These models account for non-ignorable missingness by introducing random effects or a latent process which is shared between the response model and the model for the missing-data mechanism. We discuss various random effects and latent processes approaches and compare these approaches with analyses from an opiate clinical trial data set, which had high proportion of intermittent missingness and dropout. We also compare these random effect and latent process approaches with other methods for accounting for non-ignorable missingness using this data set.

Arndt S. Stereotyping and the treatment of missing data for drug and alcohol clinical trials. (editorial). Substance Abuse Treatment, Prevention and Policy 4: e-article 2, 2009. (7 refs.)

Stigma and stereotyping of marginalized groups often is insidious and shows up in unlikely places, for instance in how clinical trials consider dropouts in treatment research. A surprising number of studies presume that people who do not complete the study protocol relapse and code their data as if they had been observed. There is no good statistical rationale for this treatment of missing data and numerous and more defensible alternative methods are available. We need to be mindful about our attitudes and preconceptions about the people we are intending to help. There is no good reason to continue to support science built on this scientifically indefensible stereotyping, however unintentional.

Babor TF. Regression to the mean: elephant in the living room or the delusions of a Swedish student 'out bicycling'? (editorial). Addiction 103(1): 4-5, 2008. (6 refs.)

Barnes SA; Larsen MD; Schroeder D; Hanson A; Decker PA. Missing data assumptions and methods in a smoking cessation study. Addiction 105(3): 431-437, 2010. (31 refs.)

Aim: A sizable percentage of subjects do not respond to follow-up attempts in smoking cessation studies. The usual procedure in the smoking cessation literature is to assume that non-respondents have resumed smoking. This study used data from a study with a high follow-up rate to assess the degree of bias that may be caused by different methods of imputing missing data. Design and methods: Based on a large data set with very little missing follow-up information at 12 months, a simulation study was undertaken to compare and contrast missing data imputation methods (assuming smoking, propensity score matching and optimal matching) under various assumptions as to how the missing data arose (randomly generated missing values, increased non-response from smokers and a hybrid of the two). Findings: Missing data imputation methods all resulted in some degree of bias which increased with the amount of missing data. Conclusion: None of the missing data imputation methods currently available can compensate for bias when there are substantial amounts of missing data.

Bender R; Kuss O; Hildebrandt M; Gehrmann U. Estimating adjusted NNT measures in logistic regression analysis. Statistics in Medicine 26(30): 5586-5595, 2007. (18 refs.)

The number needed to treat (NNT) is a popular measure to describe the absolute effect of a new treatment compared with a standard treatment or placebo in clinical trials with binary outcome. For use of NNT measures in epidemiology to compare exposed and unexposed subjects, the terms 'number needed to be exposed' (NNE) and 'exposure impact number' (EIN) have been proposed. Additionally, in the framework of logistic regression a method was derived to perform point and interval estimation of NNT measures with adjustment for confounding by using the adjusted odds ratio (OR approach). In this paper, a new method is proposed which is based upon the average risk difference over the observed confounder values (ARD approach). A decision has to be made, whether the effect of allocating an exposure to unexposed persons or the effect of removing an exposure from exposed persons should be described. We use the term NNE for the first and the term EIN for the second situation. NNE is the average number of unexposed persons needed to be exposed to observe one extra case; EIN is the average number of exposed persons among one case can be attributed to the exposure. By means of simulations it is shown that the ARD approach is better than the OR approach in terms of bias and coverage probability, especially if the confounder distribution is wide. The proposed method is illustrated by application to data of a cohort study investigating the effect of smoking on coronary heart disease.

Beynon CM; Bellis MA; Church E; Neely S. When is a drug-related death not a drug-related death? Implications for current drug-related death policies in the UK and Europe. Substance Abuse Treatment, Prevention, and Policy 2: article 25, 2007. (11 refs.)

Background: Drug-related death (DRD) figures, published by the national performance management framework, are used to monitor the performance of Drug (and Alcohol) Action Teams (D[A]ATs) in England and Wales with respect to reducing DRDs among drug abusers. To date, no investigation has been made into the types of death included in these figures, the demographic and drug profile of those who died, nor the likelihood of individuals included in DRD figures interacting with services designed to assist drug abusers. The aim of this work was to examine the characteristics of deaths classified as drug-related and to explore their applicability to performance-monitor drug-related services. Liverpool was chosen because it was reported by the national DRD monitoring system to have the highest number of DRDs in 2004. Methods: Information was retrieved from the Liverpool coroner's records and established monitoring systems on individuals reported by the national performance monitoring system as a DRD between 1st January 2004 and 30th June 2005 (n = 70). Analyses assessed differences between those categorised by the national performance monitoring system as 'drug abusers/dependents' and 'non-drug abusers/dependents' using ?2, Fisher's exact test and Mann-Whitney U. Results: Non-drug abusers were significantly older (median age 53.59 vs. 38.23), had no recent contact with drug-related agencies (cv. 31.6% of abusers who had treatment contact) and had different post mortem drug profiles than drug abusers. A significantly greater proportion of non-drug abusers died from drug toxicity - predominantly through anti-depressants, anti-psychotics and analgesics. Conclusion: Our findings suggest that the national DRD performance monitoring system includes deaths of people who are not drug abusers - individuals who are not the current focus of drug prevention, treatment or harm minimisation services. This raises concerns regarding the applicability of these figures to performance monitor D(A)ATs. Furthermore, using the more compact definitions used to monitor trends in DRDs across England, Wales and Europe fails to include a proportion of deaths attributable to drug misuse - such as those attributable blood-borne viruses. Current definitions used to monitor DRDs locally, nationally and across Europe fail to capture the true burden of drug-related mortality.

Bird SM. Database linkage: Outside reflections on health care in prisons. (commentary). Addiction 104(7): 1241-1242, 2009. (12 refs.)

This commentary addresses the article by JN Marzo, M.Rotily, F Meroueh, M Varastet, C Hunault C., et al. " RECAMS Study Group. Maintenance therapy and 3-year outcome of opioid-dependent prisoners: a prospective study in France (2003-06). Addiction 2009; 104: 1233-40. The commentary addresses issures related to data analysis and the sample size required for statistical power and ability to generalize.

Burlew AK; Feaster D; Brecht ML; Hubbard R. Measurement and data analysis in research addressing health disparities in substance abuse. Journal of Substance Abuse Treatment 36(1): 25-43, 2009. (86 refs.)

This article describes concrete strategies for conducting substance abuse research with ethnic minorities. Two issues associated with valid analysis, measurement and data analysis, are included. Both empirical (e.g., confirmatory factor analysis, item response theory, and regression) and nonempirical (e.g., focus groups, expert panels, pilot studies, and translation equivalence) approaches to improve measures are described. A discussion of the use of norms and Cutoff scores derived from a different ethnic group along with the effects of the ethnicity of the interviewer or coder on measurement is included. The section on data analysis describes why the use of race-comparison designs may lead to misleading conclusions. Alternatives to race-comparison analysis including within-group and between-group analyses are described. The shortcomings of combining ethnic groups for analyses are discussed. The article ends with a list of recommendations for research with ethnic minorities.

Carle AC. Cross-cultural invalidity of alcohol dependence measurement across Hispanics and Caucasians in 2001 and 2002. Addictive Behaviors 34(1): 43-50, 2009. (57 refs.)

Aims: Do assessments of alcohol dependence demonstrate similarly validity across Hispanics and non-Hispanic Caucasians? This investigation examined this question. Method: it employed confirmatory factor analyses for ordered-categorical measures to search for measurement bias on the AUDADIS, a standardized measure of alcohol dependence across Hispanic (n = 4819) and non-Hispanic Caucasians (n = 16, 109) in a nationally representative survey of alcohol use in the United States conducted in 2001 and 2002. Measurement: Analyses considered whether 27 items operationalizing the DSM-IV alcohol dependence construct provided equivalent measurement. Findings and conclusions: Nine items revealed statistically significant bias, suggesting strong caution regarding the cross-ethnic validity of alcohol dependence. Sensitivity analyses established that item level differences erroneously impact alcohol dependence estimates among the 2001-2002 US Hispanic population. Biased measurement underestimates differences between Hispanics and non-Hispanic Caucasians, underestimates Hispanics' true use levels, and falsely minimizes current increases in drinking behavior evidenced among Hispanics. Findings urge improved public health efforts among the Hispanic community and underscore the necessity for cultural sensitivity when generalizing measures and constructs developed in the majority to Hispanic individuals.

Chen HY; Gao SS. Estimation of average treatment effect with incompletely observed longitudinal data: Application to a smoking cessation study. Statistics in Medicine 28(19): 2451-2472, 2009. (38 refs.)

We study the problem of estimation and inference on the average treatment effect in a smoking, cessation trial where an outcome and some auxiliary information were measured longitudinally. and both were subject to missing Values. Dynamic generalized linear mixed effects models linking the outcome, the auxiliary information, and the covariates are proposed. The maximum likelihood approach is applied to the estimation and inference of the model parameters. The average treatment effect is estimated by the G-computation approach and the sensitivity of the treatment effect estimate to the nonignorable missing data mechanisms is investigated through the local sensitivity analysis approach. The proposed approach call handle missing data that form arbitrary missing patterns over little. We applied the proposed method to the analysis of the smoking cessation trial.

Ciesla JR; Spear SF. Nonresponse bias in adolescent substance abuse treatment outcomes research: Implications for evaluating care. Journal of Child & Adolescent Substance Abuse 16(3): 125-140, 2007. (17 refs.)

Nonresponse bias was investigated in an outcomes study of adolescent substance abuse treatment. Treatment-related characteristics of respondents (n = 53) were compared with those of non respondents (n = 61). Statistical analysis showed that few differences were seen between respondents and nonrespondents. The results indicate that outcomes data can he collected in this treatment population that are relatively unaffected by nonresponse bias. Missing responses are shown to be very likely missing at random.

Clarke H; Byford M. Addictive drug management policies in a long-run economic model. Australian Economic Papers 48(2): 151-165, 2009. (20 refs.)

A model of illicit, addictive drug use is proposed when users have foresight. Impacts of drug use penalties, penalties on drug use-related crime, support for drug user rehabilitation as well as the effects of health-related, harm-minimisation policies are analysed. In the short run, government policies impact only on the drug use intensities of existing addicted and casual users. Longer term policy-induced user-cost changes impact on new user and addict numbers through their effect on recruitment into addiction and quit dynamics. Effects of policies on user numbers, usage intensities and impacts on long-run social costs are analysed over this long-term horizon. The model provides a setting for analysing the long-run effects of illicit drug management policies on the social costs of illicit drug use and allows assessment of drug use abstinence and harm minimisation policy tradeoffs.

Comulada WS; Weiss RE; Cumberland W; Rotheram-Borus MJ. Reductions in drug use among young people living with HIV. American Journal of Drug and Alcohol Abuse 33(3): 493-501, 2007. (17 refs.)

ZIP models were used to detect reductions in drug abuse among young people living with HIV (YPLH) over 15 months when most young people abstain from use. YPLH (n = 171) aged 16 to 29 years were randomly assigned to an 18 session intervention or a delayed-intervention condition. The ZIP models showed significant reductions in abuse of multiple substances over time in the non-delayed intervention. Previous analyses did not find significant reductions. Intervention efficacy often cannot be detected if there are highly skewed distributions of outcomes, such as drug abuse. ZIP modeling offers an opportunity to more reliably detect behavioral changes.

Connell AM; Dishion TJ; Yasui M; Kavanagh K. An adaptive approach to family intervention: Linking engagement in family-centered intervention to reductions in adolescent problem behavior. Journal of Consulting and Clinical Psychology 75(4): 568-579, 2007. (40 refs.)

This study used Complier Average Causal Effect analysis (CACE; see G. Imbens & D. Rubin, 1997) to examine the impact of an adaptive approach to family intervention in the public schools on rates of substance use and antisocial behavior among students ages 11-17. Students were randomly assigned to a family-centered intervention (N = 998) in 6th grade and offered a multilevel intervention that included (a) a universal classroom-based intervention, (b) the Family Check-Up (selected; T. J. Dishion & K. Kavanagh, 2003), and (c) family management treatment (indicated). All services were voluntary, and approximately 25% of the families engaged in the selected and indicated levels. Participation in the Family Check-Up was predicted by 6th-grade teacher ratings of risk, youth reports of family conflict, and the absence of biological fathers from the youths' primary home. Relative to randomized matched controls, adolescents whose parents engaged in the Family Check-Up exhibited less growth in alcohol, tobacco, and marijuana use and problem behavior during ages I I through 17, along with decreased risk for substance use diagnoses and police records of arrests by age 18.

Connor JP; Symons M; Feeney GEX; Young RM; Wiles J. The application of machine learning techniques as an adjunct to clinical decision making in alcohol dependence treatment. Substance Use & Misuse 42(14): 2193-2206, 2007. (33 refs.)

With few exceptions, research in the addictive sciences has relied on linear statistics and methodologies. Addiction involves a complex array of nonlinear behaviors. This study applies two machine learning techniques, Bayesian and decision tree classifiers, in the assessment of outcome of an alcohol dependence treatment program. These nonlinear approaches are compared to a standard linear analysis. Seventy-three alcohol-dependent subjects undertaking a 12-week cognitive-behavioral therapy (CBT) program and 66 subjects undertaking an identical program but also prescribed the relapse prevention agent Acamprosate were employed in this study. Demographic, alcohol use, dependence severity, craving, health-related quality of life, and psychological measures at baseline were used to predict abstinence at 12 weeks. Decision trees had a 77% predictive accuracy across both data sets, Bayesian networks 73%, and discriminant analysis 42%. Combined with clinical experience, machine learning approaches offer promise in understanding the complex relationships that underlie treatment outcome for abstinence-based alcohol treatment programs.

Cook PJ. Comment on "Explaining change and stasis in alcohol consumption" (editorial). Addiction Research & Theory 17(6): 586-587, 2009. (4 refs.)

The author comments on the article "Explaining Change and Stasis in Alcohol Consumption," which highlights the findings of a research that alcohol consumption in Nordic countries in 2003 and 2004 was unresponsive to reductions in price. He argues that although the research provides a consistent set of findings, he thinks that there was more drinking and alcohol abuse immediately following the reduction in price.

Crits-Christoph P; Johnson J; Gallop R; Gibbons MBC; Ring-Kurtz S; Hamilton JL et al. A generalizability theory analysis of group process ratings in the treatment of cocaine dependence. Psychotherapy Research 21(3): 252-266, 2011. (45 refs.)

Videotaped group drug counseling sessions were rated for alliance, self-disclosure, positive and negative feedback, group cohesion, and degree of participation of each group member. Interrater reliability was good to excellent for most measures. However, generalizability coefficients based on statistical models that included terms for patient, counselor, session, group, and rater revealed that some measures had inadequate dependability at the patient level if only two raters and two sessions were used to create patient-level scores. In contrast, good generalizability coefficients based on two raters and two sessions were obtained for alliance, non-positive learning statements received from counselor, participation variables, and self-disclosures about the past. The implications of the findings for the design of process-outcome studies are discussed.

Cunningham JA; Selby PL. How you assess quit attempts for smoking makes a big difference to your results. (editorial). Addiction 103(11): 1761-1762, 2008. (4 refs.)

Del Boca FK; Darkes J. Enhancing the validity and utility of randomized clinical trials in addictions treatment research: III. Data processing and statistical analysis. Addiction 102(9): 1356-1364, 2007. (63 refs.)

Purpose: This is the third paper in a series that reviews strategies for optimizing the validity and utility of randomized clinical trials (RCTs) in addictions treatment research. Whereas the two previous papers focused on design and implementation, here we address issues pertaining to data processing and statistical analysis. Recommendations for enhancing data quality and utility are offered in sections on data coding and entry; and data format, structure and management. We discuss the need for preliminary data analyses that examine statistical power; patterns of attrition; between-group equivalence; and treatment integrity and discriminability. We discuss tests of treatment efficacy, as well as ancillary analyses aimed at explicating treatment processes. Safeguards are necessary to protect data quality, and advance planning is needed to ensure that data formats are compatible with statistical objectives. In addition to treatment efficacy, statistical analyses should evaluate study internal and external validity, and investigate the change mechanisms that underlie treatment effects.

Del Boca FK; Darkes J. Enhancing the validity and utility of randomized clinical trials in addictions treatment research: II. Participant samples and assessment. Addiction 102(8): 1194-1203, 2007. (89 refs.)

Purpose: This paper is the second in a series that describes strategies for optimizing the validity and utility of randomized clinical trials (RCTs) in addictions treatment research. Whereas the first paper focused on treatment implementation and research design, here we address issues pertaining to participant samples and assessment methods. Scope With respect to participant samples, sections focus on the definition of study populations; informed consent; sample size and statistical power; recruitment and enrollment; sample retention; and participant tracking systems. Assessment topics include eligibility screening and baseline assessment; treatment-related variables; outcome measures; the frequency of follow-up evaluation; and assessment process. A final section highlights the importance of pilot testing. Conclusion Sample recruitment and retention strategies are needed that safeguard both internal and external validity. Daily estimation assessment procedures are recommended because of their versatility for creating a range of outcome measures. Assessment batteries should include measures that permit the investigation of treatment processes and mechanisms of action.

Elliott MN; McCaffrey DF; Lockwood JR. How important is exact balance in treatment and control sample sizes to evaluations? Journal of Substance Abuse Treatment 33(1): 107-110, 2007. (8 refs.)

Because it is well known that the power to measure differences between two groups is typically best with an even distribution of any given fixed sample size, great emphasis is often placed on exactly equal treatment and control allocations in evaluations of substance abuse interventions. Independent randomization of individuals (e.g., a "coin flip") when study participants are enrolled in an ongoing fashion by multiple recruiters and assigned to treatment conditions does not guarantee exact balance, often prompting the use of schemes that are complex and burdensome to implement. Our results suggest that departures from simple randomization are only warranted for single-site trials involving fewer than 77 total subjects or for multisite trials with substantially fewer than 77 subjects per site. With such a rule, simple randomization will produce samples that are at least 95% as efficient as a fully balanced sample of equal size at least 95% of the time.

Falcaro M; Povey AC; Fielder A; Nahit E; Pickles A. Estimating intervention effects in a complex multi-level smoking prevention study. International Journal of Environmental Research and Public Health 6(2): 463-477, 2009. (36 refs.)

This paper illustrates how to estimate cumulative and non-cumulative treatment effects in a complex school-based smoking intervention study. The instrumental variable method is used to tackle non-compliance and measurement error for a range of treatment exposure measures (binary, ordinal and continuous) in the presence of clustering and dropout. The results are compared to more routine analyses. The empirical findings from this study provide little encouragement for believing that poorly resourced school-based interventions can bring about substantial long-lasting reductions in smoking behaviour but that novel components such as a computer game might have some short-term effect.

Finney JW. Regression to the mean in substance use disorder treatment research. Addiction 103(1): 42-52, 2008. (38 refs.)

Aims: Regression to the mean (RTM) refers to the tendency for a group of cases that differ from the population mean to move (regress) towards the mean, on average, when re-assessed, if scores at the two points are less than perfectly correlated. This paper considers factors that affect the magnitude of RTM and how RTM may impact findings from primary studies and reviews of substance use disorder (SUD) treatment. Design and methods The paper is guided largely by A Primer on Regression Artifacts by Campbell and Kenny. It reviews potential RTM effects in three areas of SUD treatment research. One is the extent to which within-group improvement in comparative treatment trials, including 'placebo effects', is a function of RTM. The second is the vulnerability of treatment evaluations employing non-equivalent control group designs to RTM and biased estimates of treatment effects when matching, or statistical equating is used to adjust for pre-existing group differences. The final issue is the impact of RTM in syntheses of research findings on SUD treatments. In particular, the tendency for later studies of a particular intervention to have smaller treatment effect sizes relative to earlier studies is considered as an RTM phenomenon. Findings: RTM is a pervasive, but often unrecognized phenomenon that can bias findings in SUD treatment studies and in systematic reviews of that research. Conclusion: SUD treatment researchers should be aware of RTM, take any available steps to reduce it, and try to diagnose whether it is still affecting research findings.

Foxcroft DR; Kypri K; Simonite V. Bayes' Theorem to estimate population prevalence from Alcohol Use Disorders Identification Test (AUDIT) scores. Addiction 104(7): 1132-1137, 2009. (13 refs.)

The aim in this methodological paper is to demonstrate, using Bayes' Theorem, an approach to estimating the difference in prevalence of a disorder in two groups whose test scores are obtained, illustrated with data from a college student trial where 12-month outcomes are reported for the Alcohol Use Disorders Identification Test (AUDIT). Using known population prevalence as a background probability and diagnostic accuracy information for the AUDIT scale, we calculated the post-test probability of alcohol abuse or dependence for study participants. The difference in post-test probability between the study intervention and control groups indicates the effectiveness of the intervention to reduce alcohol use disorder rates. In the illustrative analysis, at 12-month follow-up there was a mean AUDIT score difference of 2.2 points between the intervention and control groups: an effect size of unclear policy relevance. Using Bayes' Theorem, the post-test probability mean difference between the two groups was 9% (95% confidence interval 3-14%). Interpreted as a prevalence reduction, this is evaluated more easily by policy makers and clinicians. Important information on the probable differences in real world prevalence and impact of prevention and treatment programmes can be produced by applying Bayes' Theorem to studies where diagnostic outcome measures are used. However, the usefulness of this approach relies upon good information on the accuracy of such diagnostic measures for target conditions.

Frigon AP; Krank MD. Self-coded indirect memory associations in a brief school-based intervention for substance use suspensions. Psychology of Addictive Behaviors 23(4): 736-742, 2009. (35 refs.)

This study assessed the concurrent validity of self-generated and self-coded substance use associations for marijuana and alcohol use. Grades seven to twelve students were assessed as part of a brief intervention program in lieu of suspension for substance use infractions in school. During the cognitive assessment, students generated memory associations to probes for high-risk situations and desirable outcomes. Later, the participant rated their responses according to categories including both non-risk and substance use. Three different coding methods were compared: (1) conservative codes using clearly unambiguous responses, (2) liberal scores adding ambiguous, but likely responses, and (3) self-coded. Self-coded scores were higher, had stronger correlations with substance use, and were better predictors of substance use and problems than either conservative or liberal coded scores. These findings suggest that self-coding may be used to improve concurrent validity, decrease ambiguities in coding, and reduce the cost of measuring memory associations. The present method promises a cost effective and valid measure of indirect substance use cognitions that can be readily adapted for interventions.

Gmel G; Daeppen JB. Recall bias for seven-day recall measurement of alcohol consumption among emergency department patients: Implications for case-crossover designs. Journal of Studies on Alcohol and Drugs 68(2): 303-310, 2007. (45 refs.)

Objective: The purpose of this study was to estimate biases in recalling alcohol consumption over short periods. Method: Patients (n = 918) attending the surgical ward of the emergency department (ED) of the Lausanne University Hospital in Switzerland participated in a brief intervention study. Inclusion criteria were an average alcohol consumption exceeding 14 drinks per week for men or 7 drinks per week for women, or the consumption at least once monthly of 5 or more drinks for men or 4 or more drinks for women. Alcohol consumption was measured by means of a retrospective 7-day diary. Results: Recalled alcohol consumption decreased with the length of the recall period. Consumption was 0.9 drinks lower for a recall of 7 days compared with a recall of I day. Biases were apparent for every day of the week, but the bias was highest for consumption to be recalled for Fridays and Saturdays. Recall bias was significant only for sporadic drinkers (those drinking less than 4 days a week) but not for regular drinkers (those drinking 5 or more days a week). Conclusions: Recall bias is a threat for survey measurements of alcohol consumption in general and particularly for research designs in which the bias is differentially distributed across cases and controls. This bias is true for case-crossover designs in which the recalled consumption of an individual for a period farther away from the interview (e.g., past week) serves as the control for the acute intake of the same individual (e.g., in the 6-hour period preceding ED attendance). Because risk estimates of case-crossover designs focus particularly on sporadic drinkers, the finding of recall biases being higher among sporadic drinkers increases the chance of spurious findings in such designs.

Gmel G; Wicki M; Rehm J; Heeb JL. Estimating regression to the mean and true effects of an intervention in a four-wave panel study. Addiction 103(1): 32-41, 2008. (30 refs.)

Objectives: First, to analyse whether a taxation-related decrease in spirit prices had a similar effect on spirit consumption for low-, medium- and high-level drinkers. Secondly, as the relationship between baseline values and post-intervention changes is confounded with regression to the mean (RTM) effects, to apply different approaches for estimating the RTM effect and true change. Sample: Consumption of spirits and total alcohol consumption were analysed in a four- wave panel study (one pre-intervention and three post-intervention measurements) of 889 alcohol consumers sampled from the general population of Switzerland. Methods: Two correlational methods, one method quantitatively estimating the RTM effect and one growth curve approach based on hierarchical linear models (HLM), were used to estimate RTM effects among low-, medium- and high-level drinkers. Results Adjusted for RTM effects, high-level drinkers increased consumption more than lighter drinkers in the short term, but this was not a persisting effect. Changes in taxation affected mainly light and moderate drinkers in the long term. All methods concurred that RTM effects were present to a considerable degree, and methods quantifying the RTM effect or adjusting for it yielded similar estimates. Conclusion: Intervention studies have to consider RTM effects both in the study design and in the evaluation methods. Observed changes can be adjusted for RTM effects and true change can be estimated. The recommended method, particularly if the aim is to estimate change not only for the sample as a whole, but for groups of drinkers with different baseline consumption levels, is growth curve modelling. If reliability of measurement instruments cannot be increased, the incorporation of more than one pre-intervention measurement point may be a valuable adjustment of the study design.

Green CE; Moeller FG; Schmitz JM; Lucke JF; Lane SD; Swann AC et al. Evaluation of heterogeneity in pharmacotherapy trials for drug dependence: A bayesian approach. American Journal of Drug and Alcohol Abuse 35(2): 95-102, 2009. (48 refs.)

Aims: Difficulty identifying effective pharmacotherapies for cocaine dependence has led to suggestions that subgroup differences may account for some of the heterogeneity in treatment response. Well-attested methodological difficulties associated with these analyses recommend the use of Bayesian statistical reasoning for evaluation of salient interaction effects. Methods: A secondary data analysis of a previously published, double-blind, randomized controlled trial examines the interaction of decision-making, as measured by the Iowa Gambling Task, and citalopram in increasing longest sustained abstinence from cocaine use. Results: Bayesian analysis indicated that there was a 99% chance that improved decision-making enhances response to citalopram. Given the strong positive nature of this finding, a formal, quantitative Bayesian approach to evaluate the result from the perspective of a skeptic was applied. Conclusions: Bayesian statistical reasoning provides a formal means of weighing evidence for the presence of an interaction in scenarios where conventional, Frequentist analyses may be less informative. [Supplementary materials are available for this article. Go to the publisher's online edition of The American Journal of Drug and Alcohol Abuse for the following free supplemental resource: Appendix 1].

Grittner U; Gmel G; Ripatti S; Bloomfield K; Wicki M. Missing value imputation in longitudinal measures of alcohol consumption. International Journal of Methods in Psychiatric Research 20(1): 50-61, 2011. (44 refs.)

Attrition in longitudinal studies can lead to biased results. The study is motivated by the unexpected observation that alcohol consumption decreased despite increased availability, which may be due to sample attrition of heavy drinkers. Several imputation methods have been proposed, but rarely compared in longitudinal studies of alcohol consumption. The imputation of consumption level measurements is computationally particularly challenging due to alcohol consumption being a semi-continuous variable (dichotomous drinking status and continuous volume among drinkers), and the non-normality of data in the continuous part. Data come from a longitudinal study in Denmark with four waves (2003-2006) and 1771 individuals at baseline. Five techniques for missing data are compared: Last value carried forward (LVCF) was used as a single, and Hotdeck, Heckman modelling, multivariate imputation by chained equations (MICE), and a Bayesian approach as multiple imputation methods. Predictive mean matching was used to account for non-normality, where instead of imputing regression estimates, "real" observed values from similar cases are imputed. Methods were also compared by means of a simulated dataset. The simulation showed that the Bayesian approach yielded the most unbiased estimates for imputation. The finding of no increase in consumption levels despite a higher availability remained unaltered.

Gueorguieva R; Wu R; Pittman B; Cramer J; Rosenheck RA; O'Malley SS et al. New insights into the efficacy of naltrexone based on trajectory-based reanalyses of two negative clinical trials. Biological Psychiatry 61(11): 1290-1295, 2007. (59 refs.)

Background: The heterogeneity of clinical findings in studies evaluating the efficacy of naltrexone in the treatment of alcohol dependence has led to growing efforts to explore novel approaches to data analysis. The objective of this study was to identify distinct trajectories of daily drinking over time in two negative clinical trials and to determine whether naltrexone affected the probability to follow a particular trajectory. Methods: The Veterans Affairs (VA) Cooperative Study #425 and the Women's Naltrexone Study failed to demonstrate efficacy on primary outcome variables. Separately for each study, we analyzed daily indicators of any drinking and heavy drinking using a semiparametric group-based approach. Results: We estimated three distinct trajectories of daily drinking (both any and heavy drinking) which we described as "abstainer," "sporadic drinker," and "consistent drinker." Naltrexone doubled the odds of following the abstainer trajectory instead of the consistent drinker trajectory but did not significantly change the odds of following the abstainer trajectory as contrasted with the sporadic drinker trajectory. Conclusions: Naltrexone may have a clinically meaningful effect for alcohol-dependent patients with a high chance of consistent drinking, even in studies where it failed to show efficacy in planned analyses.

Gullo MJ; Ward E; Dawe S; Powell J; Jackson CJ. Support for a two-factor model of impulsivity and hazardous substance use in British and Australian young adults. Journal of Research in Personality 45(1): 10-18, 2011. (73 refs.)

Multiple lines of evidence suggest impulsivity comprises two distinct components relevant to substance misuse. Reward drive reflects sensitivity to rewarding stimuli and subsequent approach motivation. Rash impulsiveness reflects the ability to inhibit such approach behavior in light of negative consequences. However, several studies suggest the latter trait to be a more robust predictor. This begs the question as to whether a less parsimonious two-factor model is necessary. This study employed structural equation modeling to compare the fit of one- and two-factor impulsivity models to alcohol and drug use data provided by British (n = 183) and Australian (n = 271) young adults. Results consistently supported the two-factor model and its cross-cultural consistency, with rash impulsiveness being the more robust predictor.

Hedden SL; Woolson RF; Carter RE; Palesch Y; Upadhyaya HP; Malcolm RJ. The impact of loss to follow-up on hypothesis tests of the treatment effect for several statistical methods in substance abuse clinical trials. Journal of Substance Abuse Treatment 37(1): 54-63, 2009. (41 refs.)

"Loss to follow-up" can be substantial in substance abuse clinical trials. When extensive losses to follow-up occur, one must cautiously analyze and interpret the findings of a research study. Aims of this project were to introduce the types of missing data mechanisms and describe several methods for analyzing data with loss to follow-up. Furthermore, a simulation study compared Type I error and power of several methods when missing data amount and mechanism varies. Methods compared were the following: Last observation carried forward (LOCF), multiple imputation (MI), modified stratified summary statistics (SSS), and mixed effects models. Results demonstrated nominal Type I error for all methods; power was high for all methods except LOCF. Mixed effect model, modified SSS, and MI are generally recommended for use; however, many methods require that the data are missing at random or missing completely at random (i.e., "ignorable"). If the missing data are presumed to be nonignorable, a sensitivity analysis is recommended.

Hedden SL; Woolson RF; Malcolm RJ. A comparison of missing data methods for hypothesis tests of the treatment effect in substance abuse clinical trials: a Monte-Carlo simulation study. Substance Abuse Treatment, Prevention and Policy 3: e-article 13, 2008. (39 refs.)

Background: Missing data due to attrition are rampant in substance abuse clinical trials. However, missing data are often ignored in the presentation of substance abuse clinical trials. This paper demonstrates missing data methods which may be used for hypothesis testing. Methods: Methods involving stratifying and weighting individuals based on missing data pattern are shown to produce tests that are robust to missing data mechanisms in terms of Type I error and power. In this article, we describe several methods of combining data that may be used for testing hypotheses of the treatment effect. Furthermore, illustrations of each test's Type I error and power under different missing data percentages and mechanisms are quantified using a Monte-Carlo simulation study. Results: Type I error rates were similar for each method, while powers depended on missing data assumptions. Specifically, power was greatest for the weighted, compared to un-weighted methods, especially for greater missing data percentages. Conclusion: Results of this study as well as extant literature demonstrate the need for standards of design and analysis specific to substance abuse clinical trials. Given the known substantial attrition rates and concern for the missing data mechanism in substance abuse clinical trials, investigators need to incorporate missing data methods a priori. That is, missing data methods should be specified at the outset of the study and not after the data have been collected.

Hedeker D; Mermelstein RJ; Demirtas H. Analysis of binary outcomes with missing data: Missing = smoking, last observation carried forward, and a little multiple imputation. Addiction 102(10): 1564-1573, 2007. (25 refs.)

Aims: Analysis of binary outcomes with missing data is a challenging problem in substance abuse studies. We consider this problem in a simple two-group design where interest centers on comparing the groups in terms of the binary outcome at a single timepoint. Design: We describe how the deterministic assumptions of missing = smoking and last observation carried forward (LOCF) can be relaxed by allowing missingness to be related imperfectly to the binary outcome, either stratified on past values of the outcome or not. We also describe use of multiple imputation to take into account the uncertainty inherent in the imputed data. Setting: Data were analyzed from a published smoking cessation study evaluating the effectiveness of adding group-based treatment adjuncts to an intervention comprised of a television program and self-help materials. Participants Participants were 489 smokers who registered for the television-based program and who indicated an interest in attending group-based meetings. Measurements The measurement of the smoking outcome was conducted via telephone interviews at post-intervention and at 24 months. Findings and conclusions: The significance of the group effect did vary as a function of the assumed relationship between missingness and smoking. The 'conservative' missing = smoking assumption suggested a beneficial group effect on smoking cessation, which was confirmed via a sensitivity analysis only if an extreme odds ratio of 5 between missingness and smoking was assumed. This type of sensitivity analysis is crucial in determining the role that missing data play in arriving at a study's conclusions.

Hellemann G; Conner BT; Anglin MD; Longshore D. Seeing the trees despite the forest: Applying recursive partitioning to the evaluation of drug treatment retention. Journal of Substance Abuse Treatment 36(1): 59-64, 2009. (14 refs.)

Aims: The aim of this study is to demonstrate the utility of recursive partitioning (RP) for analyzing process and outcome data in drug treatment research. The basic methodology of RP is introduced and applied to the prediction of treatment retention. Methods: A total of 315 individuals randomly assigned to one of two treatment conditions; 289 (91.7%) completed a comprehensive baseline assessment battery. Treatment retention was assessed at a 52-week follow-up interview. Findings: The RP approach was successful in generating a parsimonious decision tree that predicted drug treatment retention from the 195 input variables. Severity of drug use (as indicated by length of time speedballing), criminal behavior (as indicated by history of property crimes), level of insight, social network, and age at intake were predictive of treatment retention. The model is estimated to explain 32% of the variability in the population. Conclusions: RP supports the notion that there are early indicators of treatment retention and that specific approaches that are tailored to individuals' needs will be potentially more successful in treatment engagement and retention than the typical "one size fits all" approach. The results also demonstrate the utility of RP for the detection of complex relationships between diverse and interdependent predictors.

Homer JF; Drummond MF; French MT. Economic evaluation of adolescent addiction programs: Methodologic challenges and recommendations. (review). Journal of Adolescent Health 43(6): 529-539, 2008. (60 refs.)

This article identifies and describes several methodologic challenges encountered in economic evaluations of substance abuse interventions for adolescents. Topics include study design, the choice of perspective, the estimation of costs and outcomes, and the generalizability of results. Recommendations are offered for confronting these challenges using examples from research on adolescent substance abuse and dependency/addiction.

Hughes JR; Callas PW. Data to assess the generalizability of samples from studies of adult smokers. Nicotine & Tobacco Research 12(1): 73-76, 2010. (10 refs.)

Introduction: One major determinant of external validity is the representativeness of the sample. This article provides data to help authors and readers assess the generalizability of samples from smoking studies. Methods: We analyzed the 2007 U. S. National Health Interview Survey. Results: We provide means, SEMs, and 95% CIs for demographic and smoking behavior characteristics of never-smokers, ever-smokers, all current smokers, current daily smokers, current nondaily smokers, long-term ex-smokers, and smokers who made a quit attempt in the last year. Discussion: Our results can help studies assess generalizability, set targets for recruitment, or reweigh data to reflect U. S. averages.

Humphreys K; Weingardt KR; Harris AHS. Influence of subject eligibility criteria on compliance with National Institutes of Health guidelines for inclusion of women, minorities, and children in treatment research. Alcoholism: Clinical and Experimental Research 31(6): 988-995, 2007. (17 refs.)

Background: Many alcohol treatment outcome studies exclude some patients with particular problems, such as psychiatric disorders, noncompliance, and homelessness. Such criteria may increase the likelihood of a study being successfully conducted, but may also have the unintended consequence of reducing a study's ability to comply with National Institutes of Health guidelines for inclusion of racial minorities, women, and children in treatment research. Methods and Results: This paper examined this issue empirically using 5 prior studies of treatment systems enrolling over 100,000 alcohol patients. Widely used eligibility criteria in the alcohol treatment field typically exclude between one-fifth to one-third of patients from enrolling in research. Under several eligibility criteria, most notably those for drug use and social/residential instability, women and African-American patients are substantially more likely to be excluded than are men and non-African-American patients, respectively. Conclusions: In designing treatment studies with many eligibility criteria, researchers may therefore inadvertently be thwarting their own good faith efforts to ensure that a range of vulnerable populations are able to participate in research. We analyze the implications of this dilemma for the generalizability of treatment results and for research design, and provide data that may help researchers working in different treatment systems estimate the impact of various eligibility criteria.

Kelly AB; Haynes MA; Marlatt GA. The impact of adolescent tobacco-related associative memory on smoking trajectory: An application of negative binomial regression to highly skewed longitudinal data. Addictive Behaviors 33(5): 640-650, 2008. (24 refs.)

Tobacco use is prevalent in adolescents and understanding factors that contribute to smoking uptake remains a critical public health priority. While there is now good support for the role of implicit (preconscious) cognitive processing in accounting for changes in drug use, these models have not been applied to tobacco use. Longitudinal analysis of smoking data presents unique problems, because these data are usually highly positively skewed (with excess zeros) and render conventional statistical tools (e.g., OLS regression) largely inappropriate. This study advanced the application of implicit memory theory to adolescent smoking by adopting statistical methods that do not rely on assumptions of normality, and produce robust estimates from data with correlated observations. The study examined the longitudinal association of implicit tobacco-related memory associations (TMAs) and smoking in 114 Australian high school students. Participants completed TMA tasks and behavioural checklists designed to obscure the tobacco-related focus of the study. Results showed that the TMA-smoking association remained significant when accounting for within-subject variability, and TMAs at Time I were modestly associated with smoking at Time 2 after accounting for within subject variability. Students with stronger preconscious smoking-related associations appear to be at greater risk of smoking. Strategies that target implicit TMAs may be an effective early intervention or prevention tool. The statistical method will be of use in future research on adolescent smoking, and for research on other behavioural distributions that are highly positively skewed.

Krebs CP; Lindquist CH; Koetse W; Lattimore PK. Assessing the long-term impact of drug court participation on recidivism with generalized estimating equations. Drug and Alcohol Dependence 91(1): 57-68, 2007. (31 refs.)

Drug courts are one of the most common strategies for dealing with the large proportion of criminal offenders who are drug-involved, yet methodological limitations limit the conclusions that can be drawn from many existing evaluations of their effectiveness. The current study 41 examined the long-term impact of drug court participation compared to regular probation on the recidivism of 475 drug-involved offenders under supervision in Hillsborough County, Florida. Using a combination of self-reported data (collected through in-person interviews at baseline, i.e., the beginning of supervision) and administrative records, the study employed a repeated measures framework (examining five 6-month time periods from baseline to 30 months post-baseline) and generalized estimating equations to compare the likelihood of being arrested between drug court participants and a matched sample of comparison offenders. The results indicate that participation in drug court was associated with a significant decrease in the likelihood of being arrested in the 12-18 months post-baseline time period. Although the drug court effect was somewhat delayed (it was not significant prior to 12 months) and short-lived (it was not significant after 18 months), the fact that significant program effects were observed during a time period that coincides with the conclusion of drug court participation for graduates and a time period well beyond initial program exposure, suggests that drug court participants are more likely than comparable offenders not exposed to drug court to remain arrest free when no longer under community supervision.

Labrie JW; Pedersen ER; Tawalbeh S. Classifying risky-drinking college students: Another look at the two-week drinker-type categorization. Journal of Studies on Alcohol and Drugs 68(1): 86-90, 2007. (26 refs.)

Objective: The present study examined the effectiveness of the 2-week period currently used in the categorization of heavy episodic drinking among college students. Two-week drinker-type labels included the following: nonbinge drinker, binge drinker, and frequent binge drinker. Method: Three samples of college student drinkers (104 volunteers, 283 adjudicated students, and 238 freshmen male students) completed the 3-month Timeline Followback assessment of drinking. Drinking behavior during the last 2 weeks of the month before the study was compared with drinking behavior during the first 2 weeks of the same month to compare behavior and resulting labels during both 2-week periods. Results: Inconsistencies existed in drinker-type labels during the first 2 weeks of the month and the last 2 weeks of the month for all three samples. Between 40% and 50% of participants in the three samples were classified as a different drinker type across the month. Nonbinge drinkers experienced a wide range of alcohol-related problems, and much variation existed among the frequent-binge-drinker label. Conclusions: The results suggest that the current definition needs to be modified to accurately identify risky-drinking college students. Expanding the assessment window past 2 weeks of behavior, as well as developing different classification schemes, might categorize risky drinkers more accurately.

Lakey B; Ondersma SJ. A new approach for detecting client-treatment matching in psychological therapy. Journal of Social and Clinical Psychology 27(1): 56-69, 2008. (30 refs.)

Identifying the treatments that are most effective for specific clients (i.e., client-treatment matching) is a major goal of research in psychological therapy. Unfortunately, there is little evidence that clients differ in the treatments to which they respond. This could result from the use of between-subjects designs that might be insensitive to client-treatment matching. In other areas of psychology, Generalizability (G) and Social Relations Model (SRM) designs routinely obtain large, conceptually identical matching effects. The current study investigated client-treatment matching using a G/SRM design. Postpartum women with a history of drug use completed three computer-delivered treatment segments of a brief motivational intervention, and rated themselves on state motivation for change following each segment. Strong client-treatment matching effects were found when using G/SRM analyses, but not when using between-subjects analyses. G/SRM methods might be more sensitive to client-treatment matching effects than are commonly used between-subjects designs.

Lee J-H; Herzog TA; Meade CD; Webb MS; Brandon TH. The use of GEE for analyzing longitudinal binomial data: A primer using data from a tobacco intervention. Addictive Behaviors 32(1): 187-193, 2007. (4 refs.)

Longitudinal study designs in addictive behaviors research are common as researchers have focused increasingly on how various explanatory variables affect responses over time. In particular, such designs are used in intervention studies that have multiple follow-up points. These designs typically involve repeated measurement of participants' responses, and thus correlation within each participant is expected. Correct inferences can only be obtained by taking into account this within-participant correlation between repeated measurements, which can complicate the analysis of longitudinal data. In recent years, generalized estimating equations (GEE) has become a standard method for analyzing non-normal longitudinal data, yet it often is not utilized by addiction researchers. The goal of this article is to provide an overview of the GEE approach for analyzing correlated binary data for behavioral researchers, using data from an intervention study on the prevention of relapse to tobacco smoking.

Loughran H; McCann ME. A case for developing community drug indicators. Social Indicators Research 102(2): 229-244, 2011. (28 refs.)

The EU Action Plan on Drugs (2005-2008) calls for member states of the European Union to provide information on five key epidemiological indicators. These are: general population surveys, prevalence and patterns of problem drug use, drug related infectious diseases, drug related deaths and mortality of drug users, and demand for drug treatment. The goal is to improve the comparability of data across the Member States, which is a central task of the EMCDDA (European Monitoring Centre for Drugs and Drug Addiction). Ireland has made progress on a national level in meeting this obligation. Currently the core information systems used to monitor the drugs problem in Ireland and to inform policy making are in the health and law enforcement areas including treatment, mortality and crime data. The dominance of such objective indicators and treatment outcome measures has contributed to obscuring the view of communities experiencing drugs problems on a day to day basis. The data are summations of the individual experience of drug problems and contribute little to understanding the broader question of how drug problem effect communities. This article draws on a community drugs study to review the contribution of traditional indicators of drug problems and consider some of the limitations of this data. It then presents an analysis of community data to identify possible community indicators of drug problems.

Lum C. The geography of drug activity and violence: Analyzing spatial relationships of non-homogenous crime event types. Substance Use & Misuse 43(2): 179-201, 2008. (71 refs.)

The pervasiveness of interest regarding the theme of a relationship between street-level drug activity and violence has been reflected throughout criminal justice research, policy, and practice as well as in public opinion. Most research has focused on the connection between the two at the individual level. This study extends previous research by examining the place-based relationship between drugs and violence. To do so, this project employs three spatial statistical approaches-measures of spatial intensity/density, measures of spatial dependence for drugs and violence separately, and a modified spatial dependence approach for non-homogenous populations to explore the relationship between drug activity and violence. The findings indicate that while drugs and violence often exhibit overlapping spatial patterns, important variations exist in the spatial relationship between the two.

Luo S; Crainiceanu CM; Louis TA; Chatterjee N. Analysis of smoking cessation patterns using a stochastic mixed-effects model with a latent cured state. Journal of the American Statistical Association 103(483): 1002-1013, 2008. (47 refs.)

We develop a mixed model to capture the complex stochastic nature of tobacco abuse and dependence. This model describes transition processes among addiction and nonaddiction stages. An important innovation of our model is allowing an unobserved cure state, or permanent quitting, in contrast to transient quitting. This distinction is necessary to model data from situations where censoring prevents unambiguous determination that a person has been "cured," Moreover, the processes that described transient and permanent quitting are likely to be different and have different policy-making implications. For example, when analyzing factors that influence smoking and can be targeted by interventions, it is more important to target those factors that are associated with permanent quitting rather than transient quitting. We apply our methodology to the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) study, a large (29,133 participants) longitudinal cohort study. While ATBC was designed as a cancer prevention study. It contains unique information about the smoking status of each participant during every 4-month period of the study. These data are used to model smoking cessation patterns using a discrete-time stochastic mixed-effects model with threee states: smoking, transient cessation, and permanent cessation (absorbent state). Random participant-specific transition probabilities among these states are used to account for participant-to-participant heterogeneity. Another important innovation in our article is to design computationally practical methods for dealing with the size of the dataset and complexity of the models. This is achieved using the marginal likelihood obtained by integrating over the Beta distribution of random effects

Mazanov J; Byrne DG. Modelling change in adolescent smoking behaviour: Stability of predictors across analytic models. British Journal of Health Psychology 13(Part 3): 361-379, 2008. (53 refs.)

Objectives. The current paper examined the variability of predictors of changes in adolescent smoking across linear and nonlinear analytic models. Design. Three analytic models typically used to model adolescent smoking behaviour were tested: one linear model of change (standard linear), one static linear model (pre-post linear) and one nonlinear model of change (cusp catastrophe). Variability in model composition was assessed by examining the pattern of variables achieving statistical significance and proportion of variance explained. Methods. Model testing was conducted on data from Australian adolescents successfully tracked through a 12-month longitudinal study of smoking (N = 779). The survey measured demographics, self-reported smoking, smoking among friends and family, self-esteem, neuroticism, coping, stress and risk taking. Results. The results indicated that while predictors of change were invariant across analytic models explanatory power varied markedly. Models of change in smoking that included simple, interacted or polynomial forms of initial conditions (past behaviour) explained more than four times the variance of models without. Conclusions. These results justified confidence in the predictors of change in adolescent smoking across analytic models. A secondary implication was that more research into past behaviour's role in the context of dynamical models of adolescent smoking and other health behaviour is needed.

McCoy TP; Ip EH; Blocker JN; Champion H; Rhodes SD; Wagoner KG et al. Attrition bias in a US internet survey of alcohol use among college freshmen. Journal of Studies on Alcohol and Drugs 70(4): 606-614, 2009. (38 refs.)

Objective: Attrition bias is an important issue in survey research on alcohol, tobacco, and other drug use. The issue is even more salient for Internet studies, because these studies often have higher rates of attrition than face-to-face or telephone surveys, and there is limited research examining the issue in the field of drug usage, specifically for college underclassmen. This study assessed whether measures of high-risk drinking and alcohol-related consequences were related to attrition groups ("stayers" or "leavers") in a cohort of college freshmen. Method: Data were collected in 2003 and 2004 from 2,144 first-year college students at 10 universities in the southeastern United States. Demographics, indicators of high-risk drinking, and alcohol-related consequences were compared between cohort stayers and leavers in statistical analyses using two methods. Results: Analyses indicated that cohort leavers reported significantly higher levels of high-risk drinking (past-30-day heavy episodic drinking, weekly drunkenness) and past-30-day smoking but not significantly increased alcohol-related consequences. The directionality of bias was modestly consistent across outcomes and comparison methods. Conclusions: The current study's findings suggest that intervention efforts to reduce smoking or high-risk drinking need to consider attrition bias during study follow-up or account for it in analyses.

McIntosh J. Is alcohol consumption good for you? Results from the 2005 Canadian Community Health Survey. Addiction Research & Theory 16(6): 553-563, 2008. (13 refs.)

Data from the Statistics Canada 2005 Canadian Community Health Survey is used to test the hypothesis that classification errors of the type noted by Fillmore et al. (2006) could invalidate the statistical results on the effects of alcohol consumption on self-rated health and the incidence of heart disease and diabetes. The results obtained in this study show that the beneficial effects of moderate alcohol use that so many studies have found, still appear even when the correct classification of alcohol use is employed. However, parameter biases and inferential errors can occur when researchers fail to distinguish between former drinkers and never drinkers within the non-drinking group.

McMillan GP; Bedrick E; C'deBaca J. A Bayesian model for estimating the effects of drug use when drug use may be under-reported. Addiction 104(11): 1820-1826, 2009. (15 refs.)

Aims: We present a statistical model for evaluating the effects of substance use when substance use might be under-reported. The model is a special case of the Bayesian formulation of the 'classical' measurement error model, requiring that the analyst quantify prior beliefs about rates of under-reporting and the true prevalence of substance use in the study population. Design: Prospective study. Setting: A diversion program for youths on probation for drug-related crimes. Participants: A total of 257 youths at risk for re-incarceration. Measurements: The effects of true cocaine use on recidivism risks while accounting for possible under-reporting. Findings: The proposed model showed a 60% lower mean time to re-incarceration among actual cocaine users. This effect size is about 75% larger than that estimated in the analysis that relies only on self-reported cocaine use. Sensitivity analysis comparing different prior beliefs about prevalence of cocaine use and rates of under-reporting universally indicate larger effects than the analysis that assumes that everyone tells the truth about their drug use. Conclusion: The proposed Bayesian model allows one to estimate the effect of actual drug use on study outcome measures.

Midanik LT; Greenfield TK; Bond J. Addiction sciences and its psychometrics: The measurement of alcohol-related problems. Addiction 102(11): 1701-1710, 2007. (77 refs.)

Aims: The focus of this paper is on psychometric issues related to the measurement of alcohol problems. Methods: Taking a broad perspective, this paper first examines several issues around the use of instruments to provide diagnostic categories in surveys, including dimensionality, severity and alcohol consumption. Secondly, a discussion of some of the political issues surrounding measurement of alcohol problems is presented, including some of the conflicts that arise when the psychometric properties of commonly used instruments are questioned. Finally, newer statistical techniques that can be applied to scale development in the alcohol field are examined, including non-linear multivariate analyses and confirmatory/hypothesis-based methods. Results and conclusions Continued scholarly discussion needs to be encouraged around these psychometric issues so that instrument development and maintenance in the addiction sciences becomes an ongoing academic pursuit as we strive to measure alcohol problems in the best way possible.

Miller PG; Johnston J; McElwee PR; Noble R. A pilot study using the internet to study patterns of party drug use: Processes, findings and limitations. Drug and Alcohol Review 26(2): 169-174, 2007. (31 refs.)

Since the 1990s there has been a rise in both the prevalence of party drug use in Australia and the use of party drug- related websites. This study investigates whether it is feasible to recruit and survey party drug users via the internet. It took place in Victoria, Australia. Participants were directed to a website where they completed a brief, structured internet- based survey. A total of 460 responses were received over 31 days, 393 of which fitted all inclusion criteria. The sample consisted predominately of young, male polydrug users and is one of the largest samples of party drug users in Australia reported thus far. It was concluded that it is feasible to recruit and survey current party drug users via the internet and that this method is quicker and cheaper than traditional survey methods, although samples are not necessarily representative of the party drug- using population. Other limitations and advantages are discussed.

Miller WR; Manuel JK. How large must a treatment effect be before it matters to practitioners? An estimation method and demonstration. Drug and Alcohol Review 27(5): 524-528, 2008. (18 refs.)

Introduction and Aims. Treatment research is sometimes criticised as lacking in clinical relevance, and one potential source of this friction is a disconnection between statistical significance and what clinicians regard to be a meaningful difference in outcomes. This report demonstrates a novel methodology for estimating what substance abuse practitioners regard to be clinically important differences. Design and Methods. To illustrate the estimation method, we surveyed 50 substance abuse treatment providers participating in the National Institute on Drug Abuse (NIDA) Clinical Trials Network. Practitioners identified thresholds for clinically meaningful differences on nine common outcome variables, indicated the size of effect that would justify their learning a new treatment method and estimated current outcomes from their services. Results. Clinicians judged a difference between two treatments to be meaningful if outcomes were improved by about 10-12 points on the percentage of patients totally abstaining, arrested for driving while intoxicated, employed or having abnormal liver enzymes. A 5 percentage-point reduction in patient mortality was regarded as clinically significant. On continuous outcome measures (such as percentage of days abstinent or drinks per drinking day), practitioners judged an outcome to be significant when it doubled or halved the base rate. When a new treatment meets such criteria, practitioners were interested in learning it. Discussion and Conclusions. Effects that are statistically significant in clinical trials may be unimpressive to practitioners. Clinicians' judgements of meaningful differences can inform the powering of clinical trials.

Monga N; Rehm J; Fischer B; Brissette S; Bruneau J; El-Guebaly N et al. Using latent class analysis (LCA) to analyze patterns of drug use in a population of illegal opioid users. Drug and Alcohol Dependence 88(1): 1-8, 2007. (62 refs.)

Background: The objective of this paper is to empirically determine a categorization of illegal opioid users in Canada in order to describe and analyze drug use patterns within this population. Methods: Drug use patterns of 679 eligible illegal opioid users outside treatment from the OPICAN study, a pan-Canadian cohort (recruited March to December, 2002) involving the cities of Toronto, Montreal, Vancouver, Edmonton and Quebec City, were empirically examined using latent class analysis. These latent classes were then further analyzed for associations using chi-square and t-test statistics. Findings: The opioid and other drug user sample surveyed were categorized into three latent classes. Class I (N=256) was characterized by the use of Tylenol 3 and benzodiazepines along with high levels of depression and self-reported pain. Class 2 (N=68) was described by the non-injection use of both heroin and crack while having a high level of homelessness. Class 3 (N=344) was shown to consist of injection drug users of heroin and cocaine exhibiting the highest levels of HIV and Hepatitis C infections amongst the classes. Conclusions: Using latent class analysis we found distinct patterns of drug use amongst illegal opioid users differing in terms of type of drugs co-used, social context, and co-morbid pathologies. These data may be useful as the empirical basis for the planning of specific prevention and treatment interventions.

Morgan-Lopez AA; Fals-Stewart W. Analytic methods for modeling longitudinal data from rolling therapy groups with membership turnover. Journal of Consulting and Clinical Psychology 75(4): 580-593, 2007. (44 refs.)

Interventions for a variety of emotional and behavioral problems are commonly delivered in the context of treatment groups, with many using rolling admission to sustain membership (i.e., admission, dropout, and discharge from group are perpetual and ongoing). The authors present an overview of the analytic challenges inherent in rolling group data and outline commonly used (but flawed) analytic and design approaches to addressing (or sidestepping) these issues. Moreover, the authors propose use of latent class pattern mixture models (LCPMMs) as a statistically and conceptually defensible approach for modeling treatment data from rolling groups. The LCPMM approach is illustrated with rolling group data from a group-based alcoholism pilot treatment trial (N = 128). Different inferences were made with regard to treatment efficacy under LCPMM vs. the commonly used standard group-clustered latent growth model (LGM); coupled with other preliminary findings in this area, inferences from LGMs may be overly liberal when applied to data from rolling groups. Continued work on data analytic difficulties in groups with membership turnover is critical for furthering the ecological validity of research on behavioral treatments.

Nahra TA; Mendez D; Alexander JA. Employing super-efficiency analysis as an alternative to DEA: An application in outpatient substance abuse treatment. European Journal of Operational Research 196(3): 1097-1106, 2009. (78 refs.)

A common technique for conducting efficiency analyses consists of a two-stage procedure that combines data envelopment analysis (DEA) with Tobit regression. As the DEA scores are censored at one, this method has the drawback of masking important information at the upper tail of the distribution of scores. In this paper, we present a DEA-based methodology for a two-stage efficiency analysis where the upper bound constraint of one for the efficiency scores is relaxed. This method, super-efficiency DEA, is contrasted with the two-stage approach that employs traditional, bounded DEA scores. We use data from the National Drug Abuse Treatment Survey to examine how the relative efficiency of the treatment units is affected by the organizational structures, operating characteristics and treatment modalities of a nationally representative sample of outpatient substance abuse treatment units. Our results show that the super-efficiency DEA approach offers advantages over the traditional methodology. It is easy to implement, and, for the same sample size provides more information.

Neal DJ; Simons JS. Inference in regression models of heavily skewed alcohol use data: A comparison of ordinary least squares, generalized linear models, and bootstrap resampling. Psychology of Addictive Behaviors 21(4): 441-452, 2007. (27 refs.)

Analysis of alcohol use data and other low base rate risk behaviors using ordinary least squares regression models can be problematic. This article presents 2 alternative statistical approaches, generalized linear models and bootstrapping, that may be more appropriate for such data. First, the basic theory behind the approaches is presented. Then, using a data set of alcohol use behaviors and consequences, results based on these approaches are contrasted with the results from ordinary least squares regression. The less traditional approaches consistently demonstrated better fit with model assumptions, as demonstrated by graphical analysis of residuals, and identified more significant variables potentially resulting in theoretically different interpretations of the models of alcohol use. In conclusion, these models show significant promise for furthering the understanding of alcohol-related behaviors.

Nordt C; Stohler R. Estimating heroin epidemics with data of patients in methadone maintenance treatment, collected during a single treatment day. Addiction 103(4): 591-597, 2008. (16 refs.)

Aims: Effects of differing drug policies are difficult to evaluate, because time trends in the spread of heroin use, the most problematic illicit drug world-wide, are unknown in almost all countries. We aimed to develop a simple method to estimate these dynamics with data that can be gathered from patients in substitution treatment within a single day. Design: We tested the assumption that being in substitution treatment on any day depends solely upon individual time since onset of regular heroin use (following a 'general inclusion function'). We used data from the case register for substitution treatments in the canton of Zurich (1992-2004), comprising 9518 patients, to model a 'general inclusion function'. Applying this function, we calculated 30 incidence curves for heroin dependence, each with data of one of 30 randomly chosen treatment days between 1992 and 2004. Findings Incidence modelling led to 30 similar curves, and therefore our hypothesis was corroborated. Additionally, our approach also revealed a restricted access to substitution treatment in the early 1990s and a decline in demand due to the introduction of heroin-assisted treatment from 1994 onwards. Conclusions In the canton of Zurich, the probability of being in substitution treatment can be described by a 'general inclusion function', and therefore dynamics of heroin epidemics can be estimated based on data of a single treatment day. Adaptation of our function to areas with a more restricted access to substitution treatment may permit these estimations also in other regions or countries. Thus, our approach facilitates the urgently needed assessment of the effects of different drug policies.

Nunes EV; Levin FR. Treatment of co-occurring depression and substance dependence: Using meta-analysis to guide clinical recommendations. Psychiatric Annals 38(11): 730-738, 2008. (50 refs.)

The authors consider the use of meta-analysis as a tool in guiding clinical care. They review the process of meta-analysis and highlight key elements, such as effect size, and then apply these prinicples to examination of studies on the effects of antidepressants, not only in treating Mood Disorder, but also in treating substance abuse.

Nygaard P; Bright K; Saltz R; McGaffigan R. Archival data: Collection and use in community alcohol projects. Substance Use & Misuse 42(12-13): 1945-1953, 2007. (6 refs.)

Archival data are considered useful for identifying problem areas, assessing levels of problems, and evaluation of interventions. However, few publications describe the process of collecting them and related potential obstacles. For the Safer California Universities study, archival data is expected to play a major role in identifying problem settings and the extent of alcohol use-related problems on the campuses. The project has experienced a number of obstacles in collecting these data. This article discusses strategies for collecting data, obstacles related to collecting them, solutions to these obstacles, and communication with partners on the campuses. The study's limitations are noted.

Paddock SM. Bayesian variable selection for longitudinal substance abuse treatment data subject to informative censoring. Journal of the Royal Statistical Society. Series C, Applied Statistics 56(Part 3): 293-311, 2007. (48 refs.)

Measuring the process of care in substance abuse treatment requires analysing repeated client assessments at critical time points during treatment tenure. Assessments are frequently censored because of early departure from treatment. Most analyses accounting for informative censoring define the censoring time to be that of the last observed assessment. However, if missing assessments for those who remain in treatment are attributable to logistical reasons rather than to the underlying treatment process being measured, then the length of stay in treatment might better characterize censoring than would time of measurement. Bayesian variable selection is incorporated in the conditional linear model to assess whether time of measurement or length of stay better characterizes informative censoring. Marginal posterior distributions of the trajectory of treatment process scores are obtained that incorporate model uncertainty. The methodology is motivated by data from an on-going study of the quality of care in in-patient substance abuse treatment.

Piper ME; Loh WY; Smith SS; Japuntich SJ; Baker TB. Using decision tree analysis to identify risk factors for relapse to smoking. Substance Use & Misuse 46(4): 492-510, 2011. (75 refs.)

This research used classification tree analysis and logistic regression models to identify risk factors related to short- and long-term abstinence. Baseline and cessation outcome data from two smoking cessation trials, conducted from 2001 to 2002 in two Midwestern urban areas, were analyzed. There were 928 participants (53.1%% women, 81.8%% White) with complete data. Both analyses suggest that relapse risk is produced by interactions of risk factors and that early and late cessation outcomes reflect different vulnerability factors. The results illustrate the dynamic nature of relapse risk and suggest the importance of efficient modeling of interactions in relapse prediction.

Poulsen PB; Dollerup J; Moller AM. Is a percentage a percentage? Systematic review of the effectiveness of Scandinavian behavioural modification smoking cessation programmes. (review). Clinical Respiratory Journal 4(1): 3-12, 2010. (46 refs.)

Introduction: Tobacco smoke is the leading preventable cause of death in the world. A total of 50% of all smokers will die from a smoking-related disease with a major impact upon quality of life and health-care costs. Tobacco-controlling policies, including smoking cessation, have increasingly been implemented across European countries. Reported effectiveness data on smoking cessation interventions are important for decision making. Objective: This study aimed to conduct a literature review on how the effectiveness (quit rates) of behavioural modification smoking cessation programmes (BMSCPs) - counselling, quitlines and quit-and-win contests - were analysed in Denmark, Sweden and Norway. Methods: A systematic review was carried out by using the search engines Medline (U.S. National Library of Medicine, Bethesda, MD, USA), Cinahl (CINAHL Information Systems, EBSCO Industries, Ipswich, MA, USA), Embase (Elsevier, New York, NY, USA) and the grey literature. Following the Russell Standards, studies were selected according to design, analysis of data [intention-to-treat (ITT)/per protocol (PP)], documentation of abstinence and length of follow-up. Cochrane reviews of pharmacological studies were used as the benchmark. Results: Although ITT analysis is the standard scientific approach advocated, most studies of BMSCPs reviewed were analysed by using the PP approach and were based on self-reported point prevalence estimates. This resulted in the reported 1-year quit rates between 16%-45% (PP) and 9%-23% (ITT). In contrast, pharmacological studies are conservative, as they are randomised, use ITT analysis and have continuous quit rates with biochemical verification of abstinence. Conclusion: This literature review reveals that quit rates of smoking cessation interventions are not always comparable. Scandinavian BMSCPs reported optimistic quit rates, confirmed by Cochrane literature review criteria. Care should be exercised when comparing smoking cessation interventions.

Prendergast M; Huang D; Hsef YI. Patterns of crime and drug use trajectories in relation to treatment initiation and 5-year outcomes - An application of growth mixture modeling across three data sets. Evaluation Review 32(1): 59-82, 2008. (43 refs.)

Drug abusers vary considerably in their drug use and criminal behavior over time, and these trajectories are likely to influence drug treatment participation and treatment outcomes. Drawing on longitudinal natural history data from three samples of adult male drug users, we identify four groups with distinctive drug use and crime trajectories during the 5 years prior to their first treatment episode. The groups' characteristics of initial treatment are compared. The trajectory groups are then included in Poisson growth curve models to predict drug use, incarceration, and employment during the 5 years following first treatment. Findings indicate that posttreatment drug use decreased and posttreatment employment increased. There was little change in posttreatment incarceration. Posttreatment trajectories for drug use, incarceration, and employment were significantly different across the four trajectory groups.

Raiff BR; Faix C; Turturici M; Dallery J. Breath carbon monoxide output is affected by speed of emptying the lungs: Implications for laboratory and smoking cessation research. Nicotine & Tobacco Research 12(8): 834-838, 2010. (16 refs.)

Introduction: Researchers have used breath carbon monoxide (CO) cutoff values ranging from 4 to 10 ppm to define abstinence in cigarette-smoking cessation research and reductions in CO as a measure of acute abstinence in laboratory research. The current study used a reversal design to investigate effects of exhalation speed on CO output in four groups (non-, light, moderate, and heavy smokers; n = 20 per group). Methods: In one condition, participants were instructed to empty their lungs as quickly as possible (fast), whereas in a different condition, participants were instructed to empty their lungs at a slow pace (slow). Conditions were counterbalanced and repeated twice for each participant. Results: For all groups, speed of exhalation was significantly lower during the slow condition than during the fast condition, and CO output was significantly higher during the slow condition than during the fast condition. Sensitivity and specificity analyses revealed that the optimal CO cutoff for smoking abstinence was 3 ppm during the fast condition versus 4 ppm during the slow condition. Additionally, when heavy smokers switched from exhaling slow to exhaling fast, they showed an approximately 30% reduction in CO. Discussion: The results suggest that exhalation speed should be monitored when CO is used as a measure of smoking status for laboratory and smoking cessation research. If exhalation speed is not monitored when using CO to verify smoking cessation, then more conservative CO cutoff values should be used to avoid false negative CO readings.

Resnicow K; Zhang NH; Vaughan RD; Reddy SP; James S; Murray DM. When intraclass correlation coefficients go awry: A case study from a school-based smoking prevention study in South Africa. American Journal of Public Health 100(9): 1714-1718, 2010. (21 refs.)

Objectives. We conducted a group randomized trial of 2 South African school-based smoking prevention programs and examined possible sources and implications of why our actual intraclass correlation coefficients (ICCs) were significantly higher than the ICC of 0.02 used to compute initial sample size requirements. Methods. Thirty-six South African high schools were randomly assigned to 1 of 3 experimental groups. On 3 occasions, students completed questionnaires on tobacco and drug use attitudes and behaviors. We used mixed-effects models to partition individual and school-level variance components, with and without covariate adjustment. Results. For 30-day smoking, unadjusted ICCs ranged from 0.12 to 0.17 across the 3 time points. For lifetime smoking, ICCs ranged from 0.18 to 0.22; for other drug use variables, 0.02 to 0.10; and for psychosocial variables, 0.09 to 0.23. Covariate adjustment substantially reduced most ICCs. Conclusions. The unadjusted ICCs we observed for smoking behaviors were considerably higher than those previously reported. This effectively reduced our sample size by a factor of 17. Future studies that anticipate significant cluster-level racial homogeneity may consider using higher-value ICCs in sample-size calculations to ensure adequate statistical power.

Riddell S; Shanahan M; Degenhardt L; Roxburgh A. A review of the use of US-derived aetiological fractions in an Australian setting for antenatal problems related to cocaine use. (review). Australian and New Zealand Journal of Public Health 32(4): 393-394, 2008. (17 refs.)

Aetiological fractions are often used as an indirect measure of morbidity and mortality related to a specific risk factor. Aetiological fractions previously used in Australia for cocaine-related antenatal haemorrhage and low birth weight newborns have relied on risk ratios calculated from US-based studies. As outlined in this paper, there are several differences in the use and prevalence of cocaine and its associated harms between the two nations. As such, it is recommended that any use of these aetiological fractions with Australian data should occur with caution.

Ripatti S; Mvaela P. Conditional models accounting for regression to the mean in observational multi-wave panel studies on alcohol consumption. Addiction 103(1): 24-31, 2008. (14 refs.)

Aims: To develop statistical methodology needed for studying whether effects of an acute-onset intervention differ by consumption group that accounts correctly for the effect of regression to the mean (RTM) in observational panel studies with three or more measurement waves. Design: A general statistical modelling framework, based on conditional models, is presented for analysing alcohol panel data with three or more measurements, that models the dependence between initial drinking level and change in consumption controlling for RTM. The method is illustrated by panel data from Finland, southern Sweden and Denmark, where the effects of large changes in alcohol taxes and travellers' allowances were studied. Findings: The suggested model allows for drawing statistical inference of the parameters of interest and also the identification of non-linear effects of an intervention by initial consumption using standard statistical software modelling tools. There was no evidence in any of the countries of the changes being larger among heavy drinkers, but in southern Sweden there was evidence that light drinkers raised their level of consumption. Conclusions: Conditional models are a versatile modelling framework that offers a flexible tool for modelling and testing changes due to intervention in consumption by initial consumption while controlling simultaneously for RTM.

Ritz B; Ascherio A; Checkoway H; Marder KS; Nelson LM; Rocca WA et al. Pooled analysis of tobacco use and risk of Parkinson disease. Archives of Neurology 64(7): 990-997, 2007. (35 refs.)

Context: Epidemiologic studies have reported that cigarette smoking is inversely associated with Parkinson disease (PD). However, questions remain regarding the effect of age at smoking onset, time since quitting, and race/ethnicity that have not been addressed due to sample size constraints. This comprehensive assessment of the apparent reduced risk of PD associated with smoking may provide important leads for treatment and prevention. Objective: To determine whether race/ethnicity, sex, education, age at diagnosis, and type of tobacco modify the observed effects of smoking on PD. Design, Setting, and Participants: We conducted the first ever pooled analysis of PD combining individual-level data from 8 US case-control and 3 cohort studies (Nurses' Health Study, Health Professionals Follow-Up Study, and Honolulu-Asia Aging Study) conducted between 1960 and 2004. Case-control studies provided data for 2328 PD cases and 4113 controls matched by age, sex, and ethnicity; cohort studies contributed 488 cases and 4880 controls selected from age- and sex-matched risk sets. Main Outcome Measure: Incident PD. Results: We confirmed inverse associations between PD and smoking and found these to be generally stronger in current compared with former smokers; the associations were stronger in cohort than in case-control studies. We observed inverse trends with pack-years smoked at every age at onset except the very elderly (> 75 years of age), and the reduction of risk lessened with years since quitting smoking. The risk reductions we observed for white and Asian patients were not seen in Hispanic and African American patients. We also found an inverse association both for smoking cigars and/or pipes and for chewing tobacco in male subjects. Conclusions: Our data support a dose-dependent reduction of PD risk associated with cigarette smoking and potentially with other types of tobacco use. Importantly, effects seemed not to be influenced by sex or education. Differences observed by race and age at diagnosis warrant further study.

Rojas NL; Sherrit L; Harris S; Knight JR. The role of parental consent in adolescent substance use research. Journal of Adolescent Health 42(2): 192-197, 2008. (33 refs.)

Purpose: The objective of our study was to assess the effects of requiring parental consent upon study participation and self-reported substance-related problems among 14-18-year-olds. Methods: This was a secondary analysis of combined data from two similar studies of adolescent substance use that recruited participants from the same adolescent clinic at Children's Hospital Boston. Study I waived parental consent, whereas Study 2 required parental consent. The combined dataset included demographic characteristics and Car, Relax, Alone, Forget, Friends, Trouble (CRAFFT) study screening test responses. The CRAFFT is an orally administered screen that yields a score from 0-6 and that has been shown to be a valid and reliable measure of risk for substance-related problems. Results: The participation refusal rate in Study 1, where consent was waived, was 19.7% (132 of 670 eligible individuals) and in Study 2 (243 of 411 eligible individuals), where consent was required, it was 59.1% (p <.0001). Participants did not differ significantly with respect to gender and age but did differ by self-identified race/ethnicity between the two studies. Because the CRAFFT score distributions were highly skewed, we used the nonparametric Mann-Whitney U test for differences in mean rank. The mean rank in Study I was significantly higher than in Study 2 (mean rank 362 vs. 325, p =.02). After controlling for age, gender, and race/ethnicity, the adjusted proportional odds ratio for a one-point increase in CRAFFT score was 1.47 (CI 1.03, 2.10) for Study 1 compared with Study 2. Conclusions: The research requirement of parental consent may result in substantial self-selection bias towards a lower risk sample.

Rooke SE; Hine DW; Thorsteinsson EB. Implicit cognition and substance use: A meta-analysis. (review). Addictive Behaviors 33(10): 1314-1328, 2008. (131 refs.)

A meta-analysis of 89 effect sizes based on the responses of 19,930 participants was conducted to estimate the magnitude of the relationship between substance-related implicit cognitions and the use of legal and illegal substances. The analysis produced a weighted average effect size of r=.31. Moderation analyses revealed significant heterogeneity in effect sizes related to facet of implicit cognition, measurement strategy, sample composition, and substance type. The largest effect sizes were found in studies that assessed implicit semantic associations, employed word association measures, and focused on marijuana use. The findings suggest that implicit cognition is a reliable predictor of substance use, although effect sizes vary as a function of several methodological factors.

Schmidt CM; Tauchmann H. Heterogeneity in the intergenerational transmission of alcohol consumption: A quantile regression approach. Journal of Health Economics 30(1): 33-42, 2011. (24 refs.)

This paper addresses the question of whether the effect of parental drinking on children's later consumption of alcohol - which is frequently found to be of positive sign - exhibits a certain pattern of heterogeneity. In particular, if this effect is more prominent in the upper tail than elsewhere in the distribution of children's alcohol consumption, conventional regression analyses that focus on the mean effect may substantially underrate parental drinking as a risk factor for children's later alcohol abuse. In our empirical application, we address this issue by applying censored quantile regression methods to German survey data. The supposed pattern of heterogeneity is indeed found in the data, at least for daily parental drinking. In addition, the intergenerational transmission of alcohol consumption exhibits gender-specific heterogeneity.

Schuckit MA; Smith TL; Trim RS; Tolentino NJ; Hall SA. Comparing structural equation models that use different measures of the level of response to alcohol. Alcoholism: Clinical and Experimental Research 34(5): 861-868, 2010. (50 refs.)

Background: The two measures of a low level of response (LR) to alcohol, an alcohol challenge and the retrospective Self-Report of the Effects of Alcohol questionnaire (SRE), each identify individuals at high risk for heavy drinking and alcohol problems. These measures also perform similarly in identifying subjects with unique functional brain imaging characteristics. However, few data are available regarding whether alcohol challenge-based and SRE-based LRs operate similarly in structural equation models (SEMs) that search for characteristics, which help to mediate how LR impacts alcohol outcomes. Methods: Two hundred and ninety-four men from the San Diego Prospective Study were evaluated for their LR to alcohol using alcohol challenges at similar to age 20. At similar to age 35, the same subjects filled out the SRE regarding the number of drinks needed for effects 15 to 20 years earlier. The two different LR scores for these men were used in SEM analyses evaluating how LR relates to future heavy drinking and to drinking in peers (PEER), alcohol expectancies (EXPECT), and drinking to cope (COPE) as potential mediators of the LR to drinking pattern (ALCOUT) relationships. Results: While the 2 LR measures that were determined 15 years apart related to each other at a modest level (r = 0.17, p < 0.01), the SEM results were similar regardless of the LR source. In both alcohol challenge-based and SRE-based LR models, LR related directly to ALCOUT, with partial mediation from PEER and COPE, but not through EXPECT in these 35-year-old men. Conclusions: Consistent with the > 60% overlap in prediction of outcomes for the 2 LR measures, and with results from functional brain imaging, alcohol challenge- and SRE-based LR values operated similarly in SEM models in these men.

Snyder JL; Bowers TG. The efficacy of acamprosate and naltrexone in the treatment of alcohol dependence: A Relative Benefits analysis of randomized controlled trials. American Journal of Drug and Alcohol Abuse 34(4): 449-461, 2008. (53 refs.)

Random controlled trials on the efficacy of naltrexone and acamprosate in the treatment of alcohol dependence were reviewed, using a Relative Benefit (RB) analysis approach. A total of 42 studies were included, showing acamprosate use demonstrated a modest improvement, with a RB of 1.76 at three month follow-up. Short-term administration of naltrexone significantly reduced the relapse rate, but was not associated with modification in the abstinence rate. There was insufficient data available to ascertain the efficacy of naltrexone and acamprosate over prolonged periods of time, or the effectiveness of the medications relative to each other.

Sterling KL; Diamond PM; Mullen PD; Pallonen U; Ford KH; McAlister AL. Smoking-related self-efficacy, beliefs, and intention: Assessing factorial validity and structural relationships in 9th-12th grade current smokers. Addictive Behaviors 32(9): 1863-1876, 2007. (33 refs.)

Smoking-related self-efficacy and beliefs about the benefits of smoking are consistently related to intention to continue smoking, a common proximal outcome in youth smoking cessation studies. Some measures of these constructs are used frequently in national and state youth tobacco surveys, despite little evidence of validity for high school smokers. Further, the association of the constructs with intention has not been demonstrated in this group. The factorial validity of the measures and the cross-sectional correlations among self-efficacy, beliefs, and intention were examined among 9th-12th grade current smokers (N= 2767, 13.8% reporting smoking > 1 cigarette in the previous 30 days; mean age 16.2; 61.2% white, 6.2% Black, 17.8% Hispanic, 5.0% Asian, 3.5% other; response rate 70%) from a convenience sample of 22 Texas schools. Confirmatory factor analyses supported evidence of factorial validity for the scales in this sample. Structural equation modeling analyses suggested youth smokers have low confidence in their ability to avoid smoking, believe smoking offers emotional or social benefits, and intend to continue smoking. The scales assess smoking-related self-efficacy, beliefs, and intention in this sample. Prospective studies are needed before intervention development implications are suggested.

Stout RL. Advancing the analysis of treatment process. (review). Addiction 102(10): 1539-1545, 2007. (45 refs.)

Aims: To review the role of process research in clinical research, to summarize progress in statistical methods for process analyses, and to describe a dynamic analytical approach that can provide new insights into the processes responsible for the effects of treatments and other variables. Summary: Process research helps us to understand what happens during our interventions, and can yield valuable knowledge regardless of whether an intervention is found to have significant effects. This is a review of recent statistical advances for dealing with missing data, tests for mediation and hierarchical modeling and demonstrate how these advances can help process researchers overcome obstacles that had limited past studies. However, the standard paradigm for process analysis, although conceptually sound, is based upon a static model that does little justice to the dynamics of treatment. Therefore, it is proposed that the paradigm is extended to study the time-course of dynamic processes, using existing statistical methods. Hierarchical linear modeling, structural regression modeling and event history methods are among the most promising tools for more advanced process analyses because of their ability to incorporate time-varying predictors. Conclusions: The function of process analysis is to probe into the mechanisms of action of treatment to locate both weaknesses and strengths, but methods for process research are still rudimentary. By conceptualizing process analysis as a problem of relating multiple time-series, many new analytical opportunities, and challenges, present themselves. Modern statistical methods can help to lead to broad advances in our understanding of the processes that affect treatment success.

Stuart EA; Green KM. Using full matching to estimate causal effects in nonexperimental studies: Examining the relationship between adolescent marijuana use and adult outcomes. Developmental Psychology 44(2): 395-406, 2008. (50 refs.)

Matching methods such as nearest neighbor propensity score matching are increasingly popular techniques for controlling confounding in nonexperimental studies. However, simple k: 1 matching methods, which select k well-matched comparison individuals for each treated individual, are sometimes criticized for being overly restrictive and discarding data (the unmatched comparison individuals). The authors illustrate the use of a more flexible method called full matching. Full matching makes use of all individuals in the data by forming a series of matched sets in which each set h as either I treated individual and multiple comparison individuals or I comparison individual and multiple treated individuals. Full matching has been shown to be particularly effective at reducing bias due to observed confounding variables. The authors illustrate this approach using data from the Woodlawn Study, examining the relationship between adolescent marijuana use and adult outcomes.

Tonigan JS. Statistical considerations in identifying mechanisms of change. Alcoholism: Clinical and Experimental Research 31(10, Supplement S): 55S-56S, 2007. (4 refs.)

The statistical search for mechanisms of change involves multiple inferential tests, ones that generally follow a fixed sequence designed to demonstrate mediation. While there are several popular approaches to conducting such tests, e.g., SEM and MRA, the inflated Type I error rate problem associated with conducting these tests has received little, if any, attention. This paper offers 2 solutions to avoid committing Type I errors associated with mediational tests. Most straightforward, investigators may choose to use a Bonferroni adjustment. In contrast, a design-based approach can be used that tests rival explanations for the observed effects. Examples drawn from addiction research are provided.

Trujols J; Guardia J; Pero M; Freixa M; Sinol N; Tejero A; de los Cobos JP. Multi-episode survival analysis: An application modelling readmission rates of heroin dependents at an inpatient detoxification unit. Addictive Behaviors 32(10): 2391-2397, 2007. (12 refs.)

The purpose of this study is to describe the characteristics of a statistical technique appropriate for analysing multi-episode data (multi-episode survival analysis), and to show its application in modelling the flow of readmissions at an inpatient detoxification unit. Data are from 784 opioid-dependent patients admitted at an inpatient detoxification unit, who totalled 1255 admission episodes. Information stored prospectively at the unit database was reviewed for the following variables at the time of each patient discharge: episode serial number, sex, route of heroin administration, reason for discharge, time of discharge, and transition time (re-entry into the inpatient detoxification unit). Cox's semi-parametric regression model seems the most appropriate for describing the series of episodes. Amongst the parametric models, most noteworthy was the superior fit of the Gompertz-Makeham model, suggesting that the transition rate decreases monotonically with time. The influence of the variables assessed differed based on the serial number of the episode. The results suggest that multi-episode survival analysis is a statistical method that can fully address the long-term perspective on treatment utilization.

Twardella D; Brenner H. Implications of nonresponse patterns in the analysis of smoking cessation trials. Nicotine & Tobacco Research 10(5): 891-896, 2008. (21 refs.)

In the statistical analysis of smoking cessation trials, participants with missing outcome data are commonly assumed to be continued smokers. Using algebraic formulas, a numerical example, and a real-life example, we evaluated the implications of nonresponse patterns on results obtained with a "missing = smoking" (MS) analysis compared with results obtained with an "available case" (AC) analysis, which excludes participants with missing outcome data. The algebraic formulas showed that MS and AC analysis provide consistent estimates of relative quit rates (RQR) when response rates in the treatment and control group are equal, regardless of the validity of the underlying assumption of both approaches. However, as shown in our numerical example, RQR estimated with both approaches can differ substantially in case of differential response rates. In the real-life example the proportion abstinent decreased from 16% to 5% in later response waves but did not reach zero. The estimates of the intervention effect from MS analysis and AC analysis converged when high and comparable response rates were achieved in both the treatment and control groups after multiple reminders. We conclude that smoking cessation studies should aim for high and equal response rates in the compared groups to ensure identification of all successful quitters and to be less susceptible to potential bias related to violation of the assumptions underlying the analytic strategies.

Verbitskaya EV; Krupitsky EM; Burakov A; Tsoy-Podosenina MV; Egorova VY; Bushara N et al. Nonverbal behavior of human addicts: Multimetric analysis. Addictive Behaviors 32(10): 2260-2267, 2007. (16 refs.)

Aims: Ethological approach followed by multimetric statistical analysis was applied to characterize and discriminate alcohol, heroin and dual, alcohol and heroin, dependent subjects. Design: Heroin, alcohol, and dual dependent patients (n=51) after one month of stabilization of remission and control volunteers (n=34) without a history of significant drug or alcohol use were interviewed and videotaped during the interview by approbation. Nonverbal behavioral cues monitored during the interview were analyzed by means of general linear procedure followed by correlation, factor and discriminant function analyses. Findings: By using this approach the attempt to discriminate addicted groups between each other failed. Therefore we found acceptable to combine subjects in one group and to suggest the similarity between alcohol and heroin dependence. It was found that principal markers of behavioral structure in addicted subjects were higher responsivity to communicate distance, less expression of affiliation behavioral pattern, low level of correlations between different behavioral patterns, and unclear factor structure. Behavioral pattern "affiliation" was identified as discriminate behavior between control and addicted subjects. Conclusions: Nonverbal cues of human behavior identified clear differences between healthy control and addictive subjects. Therefore, ethological approach described in this paper could be recommended for future use in clinical practice.

Wang JC; Carlson RG; Falck RS; Leukefeld C; Booth BM. Multi-sample standardization and decomposition analysis: An application to comparisons of methamphetamine use among rural drug users in three American states. Statistics in Medicine 26(19): 3612-3623, 2007. (24 refs.)

This study demonstrates how to use standardization and decomposition analysis (SDA) techniques to compare outcome measures simultaneously among multiple populations. Methamphetamine use among rural stimulant drug users in three geographically distinct areas of the US (Arkansas, Kentucky, and Ohio) is presented as an example of applying SDA. Findings show that the observed crude rate of 'ever used' methamphetamine in the past 30 days and the frequency of methamphetamine use in the past 30 days were much higher in Kentucky than in the other two states. However, after the compositions of socio-demographic confounding factors were standardized across the samples, the two measures of methamphetamine use ranked highest in Arkansas, followed by Kentucky, and then Ohio. Confounding factors contributed in various dimensions to the differences in the observed outcome measures among the distinct samples. The study shows that SDA is a useful technique for multi-population comparisons, providing an opportunity to look at data from a different perspective in medical studies.

Wheeler DC; Waller LA. Comparing spatially varying coefficient models: A case study examining violent crime rates and their relationships to alcohol outlets and illegal drug arrests. Journal of Geographical Systems 11(1): 1-22, 2009. (24 refs.)

In this paper, we compare and contrast a Bayesian spatially varying coefficient process (SVCP) model with a geographically weighted regression (GWR) model for the estimation of the potentially spatially varying regression effects of alcohol outlets and illegal drug activity on violent crime in Houston, Texas. In addition, we focus on the inherent coefficient shrinkage properties of the Bayesian SVCP model as a way to address increased coefficient variance that follows from collinearity in GWR models. We outline the advantages of the Bayesian model in terms of reducing inflated coefficient variance, enhanced model flexibility, and more formal measuring of model uncertainty for prediction. We find spatially varying effects for alcohol outlets and drug violations, but the amount of variation depends on the type of model used. For the Bayesian model, this variation is controllable through the amount of prior influence placed on the variance of the coefficients. For example, the spatial pattern of coefficients is similar for the GWR and Bayesian models when a relatively large prior variance is used in the Bayesian model.

Wicki M; Gustafsson NK; Makela P; Gmel G. Dimensionality of drinking consequences - cross-cultural comparability and stability over time. Addiction Research & Theory 17(1): 2-16, 2009. (40 refs.)

Despite the long tradition for asking about the negative social and health consequences of alcohol consumption in surveys, little is known about the dimensionality of these consequences. Analysing cross-sectional and longitudinal data from the Nordic Taxation Study collected for Sweden, Finland, and Denmark in two waves in 2003 and 2004 by means of an explorative principal component analysis for categorical data (CATPCA), it is tested whether consequences have a single underlying dimension across cultures. It further tests the reliability, replicability, concurrent and predictive validity of the consequence scales. A one-dimensional solution was commonly preferable. Whereas the two-dimensional solution was unable to distinguish clearly between different concepts of consequences, the one-dimensional solution resulted in interpretable, generally very stable scales within countries across different samples and time.

Wirtz PW. Advances in causal chain development and testing in alcohol research: Mediation, suppression, moderation, mediated moderation, and moderated mediation. Alcoholism: Clinical and Experimental Research 31(10, Supplement S): 57S-63S, 2007. (17 refs.)

Background: While causal modeling is generally well known to alcohol researchers, several causal structures (including suppression, mediated moderation, and moderated mediation) are often poorly understood and seldom employed when investigators seek to model the complex mechanisms of behavior change, despite their widespread applicability to the field. Methods: This paper compares and contrasts five basic structures of causal modeling in the context of contemporary alcohol research and demonstrates how mechanisms of behavior change can be conceptualized and tested as parallel and serial sequences of these basic causal structures, forming causal chains. Conclusion: Recent methodological developments, while representing an important advancement for the field, fail to adequately address the complexities of alcohol dependence phenomena. A differentiation between frequently combined forms of these causal structures is proposed that would better address the needs of the field.

Woods ER; Buka SL; Martin CR; Salganik M; Howard MB; Gueguen JA et al. Assessing youth risk behavior in a clinical trial setting: Lessons from the Infant Health and Development Program. Journal of Adolescent Health 46(5): 429-436, 2010. (36 refs.)

Purpose: The purpose of this article was to describe the use of the Youth Risk Behavior Surveillance System (YRBSS) with known 17-18-year-old patients in follow-up of a multisite randomized clinical trial, and to develop a new scoring algorithm indicating the degree of risk-taking behavior for between-group analyses. Methods: Seventy-five questions from the YRBSS were incorporated into the study questionnaire, with the development of safety plans to guide the disposition of participants. The YRBSS questions were grouped into two categories (with three subdomains each) named problem behaviors (conduct problems, sexual behavior, and suicide/hopelessness) and substance use (cigarettes, alcohol, and marijuana use), with scores for each subdomain indicating high, moderate, and low risk. Results: Of the 677 participants, the safety plan was activated 215 times for 199 (29.4%) of participants. Risk behaviors included binge drinking (149), alcohol/substance use and driving (41), depression (22), hopelessness (37), and suicidal ideation (13; all in the past). No emergency room evaluations were required. The subdomain scaling was analyzed by demographic characteristics, and findings were consistent with the literature; for example, higher rates of conduct problems in males, more suicidal ideation in females, greater sexual risk in African Americans, more substance use in males and whites, and more alcohol use in youth with mothers with higher levels of education. Conclusions: YRBSS can be administered in a research setting with appropriate safety precautions. These results should provide a useful guide to the application of the YRBSS to other adolescent populations in the future.

Wright DA; Bobashev G; Folsom R. Understanding the relative influence of neighborhood, family, and youth on adolescent drug use. Substance Use & Misuse 42(14): 2159-2171, 2007. (22 refs.)

In the United States, a variety of programs have been developed to prevent substance use among youth. These programs often target youth directly, and may also have components that address the relational influence of families, schools, and communities. We discuss clustering of youth marijuana use within and between households and neighborhoods. As often discussed in the literature, we consider analyzing "components of variance" in a hierarchical sample design with two or more levels. With a continuous outcome variable, the estimated relative size of variance components at each level can be interpreted as its relative "importance." We estimate variance components when the outcome is dichotomous, and find that for the use of marijuana in the past year, the role of the individual (individual adolescent vs. role of household vs. role of neighborhood) is quite prominent (79% of variation). A similar result is observed for the continuous scale variable of individual positive attitudes toward drug use (83%). For continuous constructs related to either household (parental monitoring) or neighborhood (neighborhood disorganization) the majority of variation still occurs at the individual level (67% and 51%, respectively), although they reveal significant percent variation (about 30%) at the corresponding family or neighborhood levels as well. We discuss the use of variance component methodology and the relevance for prevention programs.

Ye Y; Kaskutas LA. Using propensity scores to adjust for selection bias when assessing the effectiveness of Alcoholics Anonymous in observational studies. Drug and Alcohol Dependence 104(1-2): 56-64, 2009. (60 refs.)

Background: The effectiveness of Alcoholics Anonymous (AA) is difficult to establish. Observational studies consistently find strong dose-response relationships between AA meeting attendance and abstinence, and the only experimental studies favoring AA have been of 12-step facilitation treatment rather than of AA per se. Pending future randomized trials, this paper uses propensity score (PS) method to address the selection bias that potentially confounds the effect of AA in observational studies. Method: The study followed a treatment sample for 1 year to assess post-treatment AA attendance and abstinence (n = 569). Propensity scores were constructed based on known confounders including motivation, problem severity, and prior help-seeking. AA attendance during the 12-month follow-up period was studied as a predictor of alcohol abstinence for 30 days prior to the follow-up interview. PS stratification and PS matching techniques were used to adjust for the self-select bias associated with respondents' propensity to attend AA. Results: The overall advantage in abstinence initially observed narrowed when adjusted. The odds ratio associated with AA attendance reduced from 3.6 to 3.0 after PS stratification and 2.6 after PS matching to AA-attenders. Support for AA effectiveness was strengthened in the quintile with lower propensity scores and when AA-nonattenders were matched as the target group, but was weakened among those in the higher PS quintiles and when matching to AA-attenders. Discussion: These results confirm the robustness of AA effectiveness overall, because the results for higher abstinence associated with AA attendance following propensity score adjustment remained significant, and the reduction in the magnitude of AA's effect was moderate. However, the effect modification by propensity scores in both PS stratification and PS matching approaches seems to suggest that AA may be most helpful, or matter more, for those with a lower propensity to attend AA. Conversely, for those with a high propensity to go to AA (operationalized as higher motivation, greater problem severity, more prior AA and treatment exposure, etc.), attending AA may not make as much of a difference. It will be important that future studies replicate our results, as this is the first paper to use propensity score adjustment in this context.

Yeh PH; Gazdzinski S; Durazzo TC; Sjostrand K; Meyerhoff DJ. Hierarchical linear modeling (HLM) of longitudinal brain structural and cognitive changes in alcohol-dependent individuals during sobriety. Drug and Alcohol Dependence 91(2/3): 195-204, 2007. (49 refs.)

Background: Hierarchical linear modeling (HLM) can reveal complex relationships between longitudinal outcome measures and their covariates under proper consideration of potentially unequal error variances. We demonstrate the application of FILM to the study of magnetic resonance imaging (MRI)-derived brain volume changes and cognitive changes in abstinent alcohol-dependent individuals as a function of smoking status, smoking severity, and drinking quantities. Methods: Twenty non-smoking recovering alcoholics (nsALC) and 30 age-matched smoking recovering alcoholics (sALC) underwent quantitative MRI and cognitive assessments at 1 week, 1 month, and 7 months of sobriety. Eight non-smoking light drinking controls were studied at baseline and 7 months later. Brain and ventricle volumes at each time point were quantified using MRI masks, while the boundary shift integral method measured volume changes between time points. Using HLM, we modeled volumetric and cognitive outcome measures as a function of cigarette and alcohol use variables. Results: Different hierarchical linear models with unique model structures are presented and discussed. The results show that smaller brain volumes at baseline predict faster brain volume gains, which were also related to greater smoking and drinking severities. Over 7 months of abstinence from alcohol, sALC compared to nsALC showed less improvements in visuospatial learning and memory despite larger brain volume gains and ventricular shrinkage. Conclusions: Different and unique hierarchical linear models allow assessments of the complex relationships among outcome measures of longitudinal data sets. These HLM applications suggest that chronic cigarette smoking modulates the temporal dynamics of brain structural and cognitive changes in alcoholics during prolonged sobriety.

Zhang JY; Borland R; Coghill K; Petrovic-Lazarevic S; Young D; Yeh CH et al. Evaluating the effect of health warnings in influencing Australian smokers' psychosocial and quitting behaviours using fuzzy causal network. Expert Systems with Applications 38(6): 6430-6438, 2011. (13 refs.)

This paper explores the application of fuzzy causal networks (FCNs) to evaluating effect of health warnings in influencing Australian smokers' psychosocial and quitting behaviour. The sample data used in this study are selected from the International Tobacco Control Policy Evaluation Survey project. Our research findings have demonstrated that new health warnings implemented in Australia have obvious impacts on smokers' psychosocial and quitting behaviours. FCN is a useful framework to investigate such impacts that overcome the limitation of using traditional statistical techniques, such as linear regression and logistics regression, to analyse non-linear data.