ࡱ > bjbjss s K
U 8 ! D | # & ?% U% U% }% 0S T U P { { { { { { { $ H R { U ZR 0S U U { a a U% }% , p| R] R] R] U d a R U% ( }% -t R] U { R] R] q " r }% N NY T q t | 0 | q Z 4 r r U U R] U U U U U { { R] U U U | U U U U U U U U U U U U U : Online Supplementary material
Continuity of Cannabis use and Violent Offending Over the Life Course
Includes
Supplementary Appendix S1. Study sample
Supplementary Appendix S2. Measures
Supplementary Appendix S3. Analysis
Supplementary Appendix S4. Supplementary Results, including
Supplementary Fig. S1. Flow chart: Follow up assessments
Supplementary Fig. S2. Structural equation: Reciprocal causation model
Supplementary Table S1. Differences in demographics and violence data between completers and drop outs
Supplementary Table S2. Temporal relationship between cannabis use and violence
Supplementary Table S3. Multivariate logistic regression
Supplementary Table S4. Multivariate logistic regression
Supplementary Table S5. Fixed effects cross-lagged logistic regression
Supplementary Table S6. Cross-lagged structural equation models
Supplementary Appendix S5. References
Supplementary Appendix S1. Study sample
The Cambridge Study of Delinquent Development (CSDD), originally designed by Donald J. West and directed since 1982 by David P. Farrington, is a prospective longitudinal study of the development of offending and antisocial behavior in a cohort of 411 boys born mostly in 1953 living in homogenous, working class urban area of South London [a review of major findings may be found in several books ADDIN EN.CITE ADDIN EN.CITE.DATA (HYPERLINK \l "_ENREF_15" \o "West, 1977 #130"West and Farrington 1977, HYPERLINK \l "_ENREF_14" \o "West, 1973 #35"West and Farrington 1973, HYPERLINK \l "_ENREF_13" \o "West, 1982 #132"West 1982, HYPERLINK \l "_ENREF_12" \o "West, 1969 #133"West 1969, HYPERLINK \l "_ENREF_8" \o "Piquero, 2007 #139"Piquero et al. 2007, HYPERLINK \l "_ENREF_2" \o "Farrington, 2013 #140"Farrington et al. 2013) as well as in several in summary papers ADDIN EN.CITE ADDIN EN.CITE.DATA (HYPERLINK \l "_ENREF_1" \o "Farrington, 2006 #135"Farrington et al. 2006, HYPERLINK \l "_ENREF_4" \o "Farrington, 1995 #33"Farrington 1995, HYPERLINK \l "_ENREF_5" \o "Farrington, 1990 #134"Farrington and West 1990)]. The men represented the complete population of boys who were 8 years of age at that time (1961/62) and were attending one of six primary schools in a deprived area in South London. There were multiple waves (T1- T8) of data collection which included participants being interviewed in their school [at ages 8 (T1), 10 (T2), and 14 (T3)], in research offices (at ages of 16 (T4), 18 (T5), and 21 (T6)] or in their homes (at ages 32 (T7) and 48 (T8)] by social science graduates. Parents were interviewed (about once per year) and questionnaires were completed by the boys teachers (about once every two years) between ages 8 and 15 to complement information about troublesome/aggressive behavior in school and difficulties at home. 97% of the sample was white and all were raised in two-parent working class household ADDIN EN.CITE Farrington199533333317Farrington, David PThe development of offending and antisocial behaviour from childhood: Key findings from the Cambridge Study in Delinquent DevelopmentJournal of Child Psychology and PsychiatryJournal of Child Psychology and Psychiatry929-96436619950021-9630(HYPERLINK \l "_ENREF_4" \o "Farrington, 1995 #33"Farrington 1995).
Supplementary Appendix S2. Measures
Violent conviction: Criminal records
Conviction information was obtained for every follow-up year from age 10 to age 56 through searches at the central Criminal Record Office in London, a central repository containing records of all relatively serious offenses committed in Great Britain or Ireland, as well as minor juvenile offenses committed in the London area. Violent conviction (VC) was defined as conviction for robbery, assault, threatening behavior, or possessing an offensive weapon. In the case of 18 males who had emigrated outside Great Britain and Ireland by age 32, applications were made to search their criminal records in the 8 countries where they had settled, and searches were carried out in five countries. We estimated 2 separate dependent variables ADDIN EN.CITE Seillier201013713713717Seillier, AlexandreAdvani, TusharCassano, TomassoHensler, Julie GGiuffrida, AndreaInhibition of fatty-acid amide hydrolase and CB1 receptor antagonism differentially affect behavioural responses in normal and PCP-treated ratsThe International Journal of NeuropsychopharmacologyThe International Journal of Neuropsychopharmacology373-386130320101469-5111(HYPERLINK \l "_ENREF_11" \o "Seillier, 2010 #137"Seillier et al. 2010):
DV1VC [cumulative number of (subsequent) VCs]: The variable was computed by calculating the cumulative mean number per year from age 10 to 56. For cannabis users, only convictions that were committed subsequent to cannabis use were counted, excluding those that happened prior to use.
DV2VC [risk of (subsequent) VC): This dichotomized dependent variable was coded as yes if at least one conviction was committed between age 10 and 56. For cannabis users, only convictions that were committed subsequent to cannabis use were counted.
Self-reported violence
Self-reported violence (SR-V) was measured based on report of the persons involvement in assaults, fights, and use of a weapon in physical fights and also included two DVs as for violent convictions.
DV1SR-V [cumulative number of (subsequent) SR-V]: Data on violence (yes/no) was collected at three different time points, including T5: age 18 (violence between 15 and 18); T7: age 32 (involvement in fights between 27 and 32); and T8: age 48 (involvement in fights between 43 and 48).
DV2SR-V [risk of (subsequent) SR-V]: This dichotomized dependent variable was coded as yes if a subject admitted to violence at either T5, T7 or T8. Cannabis use was considered to have preceded SR-V if its use was reported either at the same or a time-point prior to the time-point under consideration for assessment of SR-V.
Cannabis use (Independent variable, IV)
At each face-to-face data collection, i.e. at ages 14 (T3), 16 (T4), 18 (T5), 32 (T7), and 48 (T8) years, the respondents were asked about their use of cannabis during the preceding 5 years.
IV 1 (Ever cannabis use): Three dichotomized cannabis variables were computed, including (1) cannabis at T5 or before (yes/no if cannabis has been used at age 18 or before); (2) cannabis at T7 (yes/no if cannabis has been used at least once between the age 27 and 32); and (3) cannabis at T8 (yes/no if cannabis has been used at least once between the age 43 and 48). IV 1 was coded as yes if a subject was classified as a cannabis user in at least one of the three variables. No separate cannabis use variables were computed for T3 and T4 as very few individuals (n=2 and n=25 respectively) reported use at these assessment points, which were therefore all considered together while estimating use at the T5 assessment (cannabis at 18 or before).
IV 2 (Continuity of cannabis use): An ordinal independent variable was computed based on three cannabis variables, classifying subjects either as (1) never cannabis user, (2) cannabis user at 1 time point only (e.g., at T5 only but not T7 or T8), (3) cannabis user at 2 time points (e.g., cannabis use at T5 and T7 but not T8), or (3) cannabis user at all 3 time points.
Covariates
Antisocial traits were assessed at age 10 based on teachers, peers, or parents ratings using the antisocial personality scale (AP) as described in detail by ADDIN EN.CITE Farrington199137373717Farrington, David PAntisocial personality from childhood to adulthoodThe PsychologistThe Psychologist389-39441991(HYPERLINK \l "_ENREF_3" \o "Farrington, 1991 #37"Farrington 1991)and included: troublesomeness, conduct problems, difficult to discipline, dishonest, has stolen, gets angry, daring, lacks concentration/restlessness, impulsive, and truants.
Alcohol use defined as presence of binge drinking (>13 units per evening in the last month yes/no) was assessed at T5, T7, and T8 and a continuous variable was computed with a score ranging from 0 = never binge drinker at T5, T7 or T8; 1= binge drinker 1 time-point only; 2= binge drinker at 2 time-points; and 3= binge drinker at all 3 time-points.
Other drug use (yes/no) was assessed at T7 and was coded as yes if the subjects had tried drugs other than cannabis.
Childhood risk factors
Childhood risk factors that may independently contribute to both violence and drug use were included in this analysis ADDIN EN.CITE Farrington20061351351356Farrington, DaviidCoid, Jeremy W.Jolliffe, DarrickSoteriou, NadineTurner, Richard E.West, Donald JCriminal careers up to age 50 and life success up to age 48: New findings from the Cambridge Study in Delinquent Development2006Home Office Research, Development and Statistics Directorate1847261086West19733535356West, Donald JamesFarrington, David PWho becomes delinquent? Second report of the Cambridge Study in Delinquent Development1973LondonHeinemann(HYPERLINK \l "_ENREF_1" \o "Farrington, 2006 #135"Farrington et al. 2006, HYPERLINK \l "_ENREF_14" \o "West, 1973 #35"West and Farrington 1973):
Social class (socioeconomic status): This variable was dichotomized, with 2 indicating that the family breadwinner (usually the father) had an unskilled manual job. All of the rest were coded as 1.
Family history of criminal/delinquent behaviour was measured up to the boys tenth birthday and referred only to biological relatives, with 2 indicating the presence of delinquent sibling and/or criminal parent. All those without a delinquent sibling and/or parent were coded as 1.
Supplementary Appendix S3. Analysis
Data was analysed using R ADDIN EN.CITE R Core Team201558585812R Core Team,R: A Language and Environment for Statistical Computing201509/092015Vienna, Austriahttps://cran.r-project.org/doc/manuals/fullrefman.pdf(HYPERLINK \l "_ENREF_9" \o "R Core Team, 2015 #58"R Core Team 2015) comprising four main statistical approaches:
We used the Kruskal-Wallis Test followed by Bonferroni correction for multiple testing to make comparisons among the different cannabis trajectory groups (never use vs. use at 1 point vs. use at 2 points vs. use at 3 points) on the average number of total convictions committed by age 56/ average number of self-reported violence by age 48. This non-parametric test was chosen as the Shapiro-Wilk Normality Test statistic was highly significant for number of convictions (p<0.001) and SR-V (p<0.001).
Secondly, univariate logistic regression analysis was employed to estimate the uncontrolled effect of cannabis use and other potential risk factors on violence outcome to identify those variables that are significantly associated with risk of VC/SR-V. The variables were chosen based on previous research ADDIN EN.CITE Farrington20065151516Farrington, DaviidCoid, Jeremy W.Jolliffe, DarrickSoteriou, NadineTurner, Richard E.West, Donald JCriminal careers up to age 50 and life success up to age 48: New findings from the Cambridge Study in Delinquent Development2006Home Office Research, Development and Statistics Directorate1847261086Resnick200431313117Resnick, Michael DIreland, MarjorieBorowsky, IrisYouth violence perpetration: what protects? What predicts? Findings from the National Longitudinal Study of Adolescent HealthJ Adolesc HealthJ Adolesc Health424. e1-424. e1035520041054-139X(Farrington et al. 2006, Resnick et al. 2004). Subsequently, odds ratios (ORs) were computed using multiple logistic regression analysis to examine the relationship between cannabis use and violence outcome, while accounting for the covariates retained from the initial bivariate models (all factors with p<0.10 were included). The violence (yes/no) dichotomization based on conviction data and the self-reported violence served as the dependent variable.
In the third stage, fixed-effects logistic regression models were fitted using the R package lme4 for binary outcome data in order to extend the ordinary logistic regression by adjusting for time-invariant non-observed fixed factors that vary across individuals, such as family background, genetic profile, personality or pre-existing violent traits. This approach allows the estimation of effect of within-person changes over time t for cannabis use [t=C1(age 18), C2(age 32), C3(age 48)] on SR-V[t=V1(age 18), V2(age 32), V3(age 48)] and VC[t=V1(age 10-18), V2(age 19-32), V3(age 32-56)]. In order to minimize any effects that may be at play in the reverse direction (reverse causation, i.e. violence predisposing to cannabis use) we (i) implemented fixed-effects models that used lagged outcome, i.e. examined whether changes in cannabis use (C1-C2) were associated with subsequent changes in violence (V2-V3) and (ii) tested a competing reverse causation model in which we estimated the effects of changes in violence (V1-V2) on changes in cannabis use (C2-C3). Alcohol use and cigarette use were included as time-dynamic covariates in the models.
Cross-lagged structural equation models were estimated using the lavaan package ADDIN EN.CITE Rosseel201254545417Rosseel, Yveslavaan: An R package for structural equation modelingJournal of Statistical SoftwareJournal of Statistical Software1-364822012(Rosseel 2012). The cannabis (C2-C3) and violence variables (V2-V3) were treated as dependent variables, allowing to test for reciprocal changes in the association at different stages of the life span. We tested both an unconstrained model and a model in which the parameters (Ct and Vt) were constrained to be equal across time points. The models were fitted using the robust weighted least squares (WSL) approach. Model goodness of fit was assessed on the basis of a number of fit indices, including the model chi-squared goodness of fit statistic (non-significant or small chi-square value indicates that the model fits the data well), the root mean squared error of approximation (RMSEA, for which values of .05 indicate good fit and values up to .08 represent reasonable errors of approximation) ADDIN EN.CITE MacCallum199649494917MacCallum, Robert C.Browne, Michael W.Sugawara, Hazuki M.Power analysis and determination of sample size for covariance structure modelingPsychological methodsPsychological methods1301219961939-1463(HYPERLINK \l "_ENREF_7" \o "MacCallum, 1996 #49"MacCallum et al. 1996) and the Comparative Fit Index (for which values of .95 are acceptable and of .95 or higher are indicative of good fit ADDIN EN.CITE Hu199950505017Hu, Li t z e <