FT12 Sample

Twins enrolled into FT12 were born 1983 through 1987; they were identified through Finland’s Central Population Registry (CPR). Ascertainment was exhaustive, and twins in all pairs with both twins alive, resident in Finland and enrolled in normal public schools, were invited to participate. To date, four waves of questionnaire data have been collected from FT12 participants. Key information regarding the participation of the twins and parents in the epidemiological, questionnaire-based study waves is provided in Table 1, while Figure 2 shows the numbers of twins who have replied to all or fewer of the survey waves.

Table 1.

Questionnaires and interviews of the FinnTwin12 twin cohort in five waves of data collection. At baseline, a family questionnaire was completed by the parents (n = 2705 families out of 3136 contacted) about the twins’ birth and childhood and provided consent for the study. In addition, parents replied to a questionnaire about themselves and their own parents and jointly completed a questionnaire on the twins’ behavior

Note: MZ, monozygotic; DZ, dizygotic; SSDZ, same-sex dizygotic; OSDZ, opposite-sex dizygotic.

Fig. 2.

Overlap in responses to wave 1 to 4 studies for the FinnTwin12 twins. The young adult study (YAQ) consists of the intensive sample study (interview, in-person assessments and questionnaires) and the remaining epidemiology sample (questionnaire only).

Stage 1 (Epidemiological) Study

A total of 2705 families (87% of all identified twins in the five birth cohorts, living with one or both biological parents and eligible for study) returned the initial family questionnaire mailed late in the year in which successive twin birth cohorts reached age 11. The family questionnaire was mailed to all families of all eligible twins, using updated residential postal addresses obtained from the Population Registry. Return of the completed family questionnaire gave us permission to contact the twin children, and questionnaires for twin boys and girls, and those for their parents, were then mailed. Parents living apart from their twins were included in these postal mailouts. Parent and teacher ratings of all twins were collected at baseline, with high participation rates; 92% of the parents and 93% of the twins’ classroom teachers returned ratings. Cooperation at these initial stages was not associated with family structure, area of residence within Finland, parental age, or sex or zygosity of twin-pairs. All twins in the FT12 sample were invited to participate in the first follow-up survey at age 14 (with 92% retention). Questionnaires for the second follow-up at age 17 (75% retention) were sent only to those who responded earlier, but in young adulthood (mean age 22, 66% retention), questionnaires were again sent to all (the now adult twins) to permit them to withdraw from further study if they chose to do so.

Stage 2 (Intensive) Study

Nested within the full epidemiological sample described above is a subset of intensively studied twins (n = 1854), for whom we have collected much more detailed phenotypic and genetic information. In creating the invited subsample (1035 families), random sampling preceded selective enrichment for familial alcoholism risk. A random sample (of predetermined n) was selected from each birth cohort; to that random sample, all remaining twins in that cohort for whom one or both parents exceeded the threshold on a questionnaire screen for alcoholism (the Malmö-Modified Michigan Alcoholism Screening Test; Mm-MAST; Kristenson & Trell, Reference Kristenson and Trell1982) were then added. The subsample of twins selected for familial risk comprised 28% of all families selected for intensive study from the five successive birth cohorts. Parents in the intensively studied subsample were interviewed using SSAGA during the year in which their twin children reached age 12. In the early months of that year, twins in the intensive subsample completed an in-school assessment that included a student questionnaire and peer nominations (Pulkkinen et al., Reference Pulkkinen, Kaprio and Rose1999) in classrooms across Finland. At follow-up at age 14, twins in the intensive sample were invited to complete an in-person, adolescent version of the SSAGA and a set of neurocognitive tests.

In young adulthood, the intensive sample of twins was invited to in-person testing in Helsinki. This wave 4 data collection in 2006–2009 was the most comprehensive assessment of the twins conducted to date. It included structured psychiatric interviews at an average age of 21.9 (SD 0.8, range 21–26) years: adult SSAGA interviews in person or via telephone if a twin could not travel to Helsinki. The in-person testing also included neuropsychological assessments and collection of DNA and serum samples. In total, 73% completed young adult SSAGAs and 1260 individuals provided DNA [526 of whom were monozygotic (MZ) twins]; 812 also completed a laboratory assessment that included a test battery assessing neurocognitive functioning.

Measures

We highlight below relevant measures collected from the twins across the categories of the theoretical model illustrated in Figure 1.

Alcohol and other substance use. Beginning in midadolescence, twins in the full epidemiological sample responded to survey questions about their frequency of alcohol use and intoxication, as well as their use of tobacco, cannabis and other substances. In follow-up in young adulthood, twins also reported on their alcohol problems, as measured with the Mm-MAST. Intensively studied twins completed an in-school questionnaire in the spring of the year they turned age 12; it included simple yes/no questions on personal experience with alcohol and cigarettes and that of their peers, and at age 14, they were administered the adolescent version of SSAGA in face-to-face format, yielding symptom counts and diagnoses for DSM-IV alcohol dependence and abuse and nicotine dependence. At age 22, the intensive sample provided information about their substance use, abuse and dependence (smoking, alcohol, cannabis and other drug use) based on questionnaires and interviews for comorbid DSM-IV diagnoses (alcohol, illicit drugs, depression, eating disorders and antisocial behavior), while all twins in the full epidemiological sample were asked about their alcohol and tobacco use in questionnaires.

Internalizing and externalizing behaviors. At age 12, teachers, parents and twins in the full FT12 sample reported on twins’ emotional problems, behavioral problems and social adjustment using the Multidimensional Peer Nomination Inventory (MPNI) and teacher ratings (a different teacher, different school setting) were obtained again on the full sample at age 14. Peer nominations were obtained at age 12 for the intensive sample with the MPNI in parallel teacher, parent and peer format (Pulkkinen et al.,Reference Pulkkinen, Kaprio and Rose1999). For more information on the development and predictive value of the MPNI, see the recent book by Lea Pulkkinen (Reference Pulkkinen2017). Twins responded to questions about their depressive symptoms with the Children’s Depression Inventory at ages 12 and 14 (Kovacs, Reference Kovacs1991) and the General Behavior Inventory (GBI) at ages 17 and 22 (Depue, Reference Depue1987). As part of the SSAGA psychiatric interviews, the intensively studied subsample of twins was assessed for depression, anxiety, suicidal behavior, conduct disorder/adult antisocial behavior and oppositional defiant disorder at age 14 and 22.

Person-level factors. In the age 17 follow-up questionnaire, twins completed the Karolinska Scale of Personality (Klinteberg et al., Reference Klinteberg, Magnusson and Schalling1986), with scales tapping predispositions to psychopathology. Additional person-level data, including genotyping and neurocognitive functioning, were collected from participants in the intensively studied subsample at wave 4. The neurocognitive functioning battery included measures of executive functioning (Delis et al., Reference Delis, Kaplan and Kramer2001; Närhi et al., Reference Närhi, Räsänen, Metsäpelto and Ahonen1997; Reitan & Wolfson Reference Reitan and Wolfson1993), general cognitive ability (Wechsler, Reference Wechsler1997) and working memory (Wechsler, Reference Wechsler1987). The intensive sample at age 22 was also asked about adjustment (12-item General Health Questionnaire GHQ-12), personality (NEO-Five Factor Inventory, NEO-FFI), schizotypy (Schizotypal Personality Questionnaire, SPQ) and sense of coherence.

Environmental/contextual factors. In adolescence, twins reported on their recent life events (Salvatore, Meyers et al., Reference Salvatore, Meyers, Yan, Aliev, Lansford, Pettit and Dick2015), their parents’ knowledge about their plans and activities, how they were spending their money, where and with whom they spent leisure time (Chassin et al.,Reference Chassin, Pillow, Curran, Molina and Barrera1993), and their affiliations with deviant peers. In young adulthood, they reported on whether they were involved in a committed romantic partnership; the number of their social support providers and those with whom they could share feelings and concerns (Barr et al., Reference Barr, Salvatore, Maes, Korhonen, Latvala, Aliev and Dick2017); years of completed education and student status (Barr et al., Reference Barr, Salvatore, Maes, Aliev, Latvala, Viken and Dick2016; Latvala et al., Reference Latvala, Dick, Tuulio-Henriksson, Suvisaari, Viken, Rose and Kaprio2011); their employment status and experiences with unemployment and parenthood.

Physical health. At all ages, participants reported on their self-rated overall health and 10 somatic symptoms (e.g., headache, pain, and tension/nervousness; Silventoinen et al., Reference Silventoinen, Posthuma, Lahelma, Rose and Kaprio2007). Information on illnesses was also asked in most detail at age 22 in the intensive sample, including general health (illnesses, body composition, metabolism, and experienced health). Weight and height were asked at all ages. They also reported on lifestyle factors (physical activity and food habits). The twins were tested for taste and smell, and they were asked about chemosensory preferences in the intensive sample at age 22.

Genetics, epigenetics and metabolomics. Genotyping was conducted using the Human670-QuadCustom Illumina BeadChip (Illumina, Inc., San Diego, CA, USA) and DNA methylation using the 450-k Illumina array. A nuclear magnetic resonance (NMR)-based metabolomics panel was used (Bogl et al., Reference Bogl, Kaye, Ramo, Kangas, Soininen, Hakkarainen and Pietiläinen2016). We have also collected DNA samples from some of the nonintensive twins, and Table 2 gives the numbers with genotypes and epigenetic data.

Table 2.

Number of deaths, and the numbers genotyped or with methylation data by survey wave. The number of pairs genotyped on genomewide array

Note: MZ, monozygotic; DZ, dizygotic.

Register follow-up. We update vital status from the CPR as needed, and Table 2 shows the numbers of twins who have died by participation status in the survey waves. Over time, we will be able to examine causes of death by linkage to Statistic Finland Cause-of-Death Register. In wave 4 and earlier, we have not consented the twins for linkage to other medical registers in Finland, but that is an option we will consider in the next wave of data collection. For details on data availability through the National Institute for Health and Welfare Biobank for those participants with biosamples, please see Kaprio et al. (Reference Kaprio, Bollepalli, Buchwald, Iso-Markku, Korhonen, Kovanen and Waller2019) for our review paper on the older Finnish Twin Cohort.

What follows is an overview of major areas of scientific knowledge to which FT12 data analyses have contributed.

Analyses of Gene-Environment Correlations and Interactions

FT12, in combination with FinnTwin16, has been pivotal in advancing our understanding of how genetic and environmental influences contribute to alcohol use and related behavioral health challenges across adolescence and into young adulthood. The genetically informative longitudinal data from a population-based sample with high response rates and multiple ratings, including parent, teacher and peer reports, have provided unprecedented opportunity to characterize the etiology of alcohol use and drinking-related problems across the critical developmental period from adolescence to young adulthood. These data have demonstrated important changes in genetic and environmental influences across adolescence. A fundamental finding is that alcohol initiation in early adolescence is largely influenced by shared environmental factors, with little or no evidence of genetic effects (Rose, Dick, Viken, Pulkkinen et al., Reference Rose, Dick, Viken, Pulkkinen and Kaprio2001). But as individuals progress across adolescence, transitioning from initiation and experimentation to more regular patterns of drinking, genetic effects become increasingly important, accounting for nearly 50% of the variance in frequency of alcohol use by age 18. Concurrently, common environmental influences steadily decline in importance across adolescence (Pagan et al., Reference Pagan, Rose, Viken, Pulkkinen, Kaprio and Dick2006). In contrast to adult alcohol dependence, which shows significant genetic influence (Verhulst et al., Reference Verhulst, Neale and Kendler2015), symptoms of alcohol dependence in very early adolescence are largely environmentally influenced, indicating that the etiological factors underlying alcohol problems at early developmental stages are often to be found in environmental experiences.

Our data have also demonstrated dramatic modulation of genetic effects as a function of multiple environmental contexts across adolescence, with genetic influences on substance use assuming greater importance under conditions of reduced parental monitoring (Dick, Viken et al., Reference Dick, Viken, Purcell, Kaprio, Pulkkinen and Rose2007), high peer deviance (Cooke et al., Reference Cooke, Meyers, Latvala, Korhonen, Rose, Kaprio and Dick2015), more urban settings (Dick, Rose, Viken et al., Reference Dick, Rose, Viken, Kaprio and Koskenvuo2001) and as a function of related sociodemographic factors (Dick, Rose, Viken et al., Reference Dick, Rose, Viken, Kaprio and Koskenvuo2001; Rose, Dick, Viken & Kaprio, Reference Rose, Dick, Viken and Kaprio2001). Furthermore, we found that the importance of relevant moderating environmental factors changes across developmental stages (Dick, Pagan, Viken et al., Reference Dick, Pagan, Viken, Purcell, Kaprio, Pulkkinen and Rose2007). In young adulthood, genetic influences on alcohol use vary as a function of romantic relationship status and social support networks (Barr et al., Reference Barr, Salvatore, Maes, Korhonen, Latvala, Aliev and Dick2017). We have extended these studies to include measured polygenetic risk scores, characterizing the mediating pathways by which genetic risk is operating, and the environments that moderate associations of polygenic risk scores with adverse outcomes (Barr et al., Reference Barr, Kuo, Aliev, Latvala, Viken, Rose and Dick2019; Li et al., Reference Li, Savage, Kendler, Hickman, Mahedy, Macleod and Dick2017; Salvatore et al., Reference Salvatore, Aliev, Edwards, Evans, Macleod, Hickman and Dick2014; Salvatore et al., Reference Salvatore, Savage, Barr, Wolen, Aliev, Vuoksimaa and Dick2018; Savage, Salvatore et al., Reference Savage, Salvatore, Aliev, Edwards, Hickman, Kendler and Dick2018). These studies underscore the important role of multiple environmental contexts in exacerbating or buffering the impact of genetic influences on alcohol use outcomes emerging during early adulthood.

A series of FT12 studies have characterized effects of early pubertal maturation (Dick et al., Reference Dick, Rose, Viken and Kaprio2000; Dick, Rose, Pulkkinen et al., Reference Dick, Rose, Pulkkinen and Kaprio2001; Savage, Rose et al., Reference Savage, Rose, Pulkkinen, Silventoinen, Korhonen, Kaprio and Dick2018), behavioral characteristics of peer friendships (Dick, Pagan, Holliday et al., Reference Dick, Pagan, Holliday, Viken, Pulkkinen, Kaprio and Rose2007) and the nature of parental influences on initiation and trajectories of substance use (Latendresse et al., Reference Latendresse, Rose, Viken, Pulkkinen, Kaprio and Dick2008, Reference Latendresse, Rose, Viken, Pulkkinen, Kaprio and Dick2009, Reference Latendresse, Rose, Pulkkinen, Kaprio and Dick2010; Salvatore, Aliev et al., Reference Salvatore, Aliev, Bucholz, Agrawal, Hesselbrock, Hesselbrock and Dick2015). One finding of these efforts (Dick, Pagan, Holliday et al., Reference Dick, Pagan, Holliday, Viken, Pulkkinen, Kaprio and Rose2007; Dick et al., Reference Dick, Rose, Viken and Kaprio2000) has been that females may be more strongly affected by environmental influences than are males; and data from peer nominations at age 12 offer evidence of a protective role of social anxiety on trajectories of substance use, with differential effects for males and females (Savage et al., Reference Savage, Kaprio, Korhonen, Pulkkinen, Rose, Verhulst and Dick2016). We briefly review some of that recent work below.

Sex Differences

The multiple sources of adolescent assessment in FT12 permitted us to demonstrate shared environmental variance in measured characteristics of home environments (reduced parental monitoring and less supportive household atmosphere). Analyses of drinking patterns at age 14 illustrated cooperative social interaction on abstinence and the peer nomination data from FT12 document significant similarities of peer-assessed behavior of twins and their best friends. Behavioral resemblance was greater in the peer assessments of best friends of MZ cotwins, suggesting that adolescent friendship selection is an active process guided, in part, by our genetic dispositions. Patterns of substance use in early adolescence are especially sensitive to both active selection (adolescents seek to befriend those with behaviors and attitudes matching their own), and reciprocal effects of socialization experiences with the selected peers. A robust predictor of drinking/abstaining in adolescence is drinking or smoking friends, and having friends who drink, smoke, use drugs or engage in delinquent behaviors is more predictive of drinking among girls than among boys.

Our data show that girls are more vulnerable to reduced parental modeling, accelerated pubertal maturation and to effects of associating with substance-using peers (Dick, Pagan, Holliday et al., Reference Dick, Pagan, Holliday, Viken, Pulkkinen, Kaprio and Rose2007). Recent analyses of FT12 data provide more evidence of gender differences in vulnerability to risk factors and in trajectories of use and abuse into adulthood. Sex differences in patterns of the association of pubertal development with substance use (Savage, Rose et al., Reference Savage, Rose, Pulkkinen, Silventoinen, Korhonen, Kaprio and Dick2018) found shared environmental factors contributed to early puberty and heavier substance use for females, and longitudinal effects of differences in pubertal development at age 12 were generally greater among females; a result shown in the association of puberty development at age 12 with symptom counts for alcohol use disorder (AUD) at ages 14 and 22. In contrast, the association of puberty development with illicit drug use was greater for males.

A negative relationship (i.e., a buffering effect) between social anxiety and alcohol dependence was found using ratings of social anxiety from classmate peers at age 12. The protective effect of social anxiety against alcohol problems at age 22 was greater for boys than for girls. But the association of social anxiety at age 12 with smoking frequency at 14 was more strongly negative (more protective) among girls. Meyers et al. (Reference Meyers, Salvatore, Vuoksimaa, Korhonen, Pulkkinen, Rose and Dick2014) calculated estimates of genetic risk from symptom counts of AUD obtained in interviews of the cotwin (at age 22) and the parents (when twins were 12) and related these genetic risk estimates to twins’ alcohol use across their development. Results showed significant sex differences: Genetic risk for AUD influenced early adolescent alcohol use more in girls than in boys. In parallel analyses of the association with drinking from genetic risk for externalizing behaviors at ages 14 and 17, sibling interaction effects were found. The associations were greater in females with twin brothers, evidence of positive sibling interaction effects. And in a recent analysis of FT12 data, Barr et al. (Reference Barr, Kuo, Aliev, Latvala, Viken, Rose and Dick2019) asked whether associations of genomewide polygenic scores for alcohol consumption predict alcohol misuse, whether those associations are moderated by romantic relationships, and, if so, whether the gene–environment interactions are consistent across sex. Alcohol misuse was assessed by frequency of drinking, frequency of intoxicating and AUD symptom counts in young adulthood. Genomewide polygenic scores predicted all three measures of misuse, and having a romantic relationship negatively influenced each of the three outcomes. But, interestingly and importantly, the protective effect of a romantic relationship on frequent intoxication was found only in males, replicating other evidence that males are more protected from high-risk drinking by involvement in romantic relationships.

Another recent analysis used both teacher and repeated self-report measurements across the FT12 waves of assessment to show that alcohol use behaviors at ages 12 and 14 predicted lower educational achievement at later time points, and those associations remained even after previous achievement and measured confounders were taken into account (Latvala et al., Reference Latvala, Rose, Pulkkinen, Dick, Korhonen and Kaprio2014). In another study, the genetics of perceived family environment was evaluated with data from the full FT12 sample at 12, 14, and 17, yielding evidence that additive genetic effects explained increasingly more of the variation across that 5-year period of development while the role of the environment shared by cotwins decreased (Silventoinen et al., Reference Silventoinen, Su, Barr, Rose, Dick and Kaprio2019).

Jointly, these FT12 studies have importantly contributed to our understanding the interplay of genetic and environmental risk factors for alcohol use and problems across adolescence and into young adulthood. With future follow-up of this informatively rich twin cohort, we are developing research protocols to characterize alcohol use patterns, predictors and consequences into early midlife.

Testing the ‘Gateway’ Hypothesis in Substance Use

FT12 data have contributed insights into debate surrounding the gateway hypothesis of substance use/abuse – the hypothesis that early exposure to ‘milder’, legal substances, such as tobacco and alcohol, leads to use of ‘stronger’, ‘illegal’, more addictive substances, including cannabis. In an initial test of that hypothesis (Korhonen et al., Reference Korhonen, Huizink, Dick, Pulkkinen, Rose and Kaprio2008), we explored factors predicting use of cannabis and other illicit drugs (baseline at age 12 with follow-up at ages 14 and 17) in an epidemiological longitudinal study of >4100 twins. The potential predictors were age 12 baseline measures reported by the twins, their parents or teachers. As individual factors, we tested smoking, alcohol use, behavioral and emotional problems; as peer factors: the number of smoking friends and acquaintances with drug experience; and as family factors: parental substance use, socioeconomic status and prenatal exposure to nicotine. When adjusted for within-family confounds, the significant predictors were early smoking onset, drinking to intoxication, having peers with smoking and drug experience, father’s weekly drinking to intoxication and aggressive behavior. Early-onset smoking was the most powerful predictor among individuals and within exposure-discordant twin-pairs, but whether that association is causal remains uncertain. Accordingly, a follow-up study (Huizink et al., Reference Huizink, Levälahti, Korhonen, Dick, Pulkkinen and Kaprio2010) compared two models: (1) one that describes a direct impact of liability to tobacco use on cannabis and other illicit drug use with (2) a model that assumes shared underlying liability for tobacco and illicit drug use. Tobacco and illicit drug use were assessed at age 17. The multivariate model, including a direct impact of the initiation of tobacco use on illicit drug use, provided the best fit to the data, suggesting that liability to initiate smoking directly affects illicit drug use, findings supporting the gateway hypothesis, although shared underlying liabilities cannot be excluded.

To evaluate the role of externalizing behaviors in the pathway from smoking to drug use, we analyzed whether hyperactivity–impulsivity, aggressiveness, and inattention at 12 predict illicit drug use at age 17 independently, or whether their associations with drug use are mediated through cigarette smoking at age 14 (Korhonen et al., Reference Korhonen, Levälahti, Dick, Pulkkinen, Rose, Kaprio and Huizink2010). The association of hyperactivity–impulsivity with drug use was mostly mediated through cigarette smoking. Concerning aggressiveness and inattention, consistent evidence on mediation was seen only among boys. Consistently in all models, the direct association of cigarette smoking on drug use was strong and highly significant, while the associations of externalizing problem behaviors with illicit drug use are partially mediated through cigarette smoking. Because both smoking and externalizing behavior appeared to be important risk factors of illicit drug use, we estimated genetic and environmental influences common to externalizing behavior (at age 12), smoking (at age 14) and initiation of drug use (at age 17). Multivariate Cholesky models were fit to data from 737 MZ and 722 dizygotic (DZ) twin-pairs. Heritability of externalizing behavior was 56%, that of smoking initiation/amount 20/32%, and initiation of drug use 27%. In the best-fitting model, common environmental influences explained most of the covariance between externalizing behavior and smoking initiation (69%) and amount (77%). Covariance between smoking initiation/amount and drug use was due to additive genetic (42/22%) and common environmental (58/78%) influences. Half of the covariance between externalizing behavior and drug use was due to shared genetic and half due to the environments shared by cotwins. Our results indicated that early observed externalizing behavior provides significant underlying genetic and environmental influences common to later substance use (Korhonen et al., Reference Korhonen, Latvala, Dick, Pulkkinen, Rose, Kaprio and Huizink2012).

Finally, we examined the role of overlapping genetic and environmental factors in the relationship between early adolescent conduct problems and substance use in young adulthood (Verweij et al., Reference Verweij, Creemers, Korhonen, Latvala, Dick, Rose and Kaprio2016). We found modest to moderate phenotypic correlations between early adolescent conduct disorder symptoms and substance use in young adulthood. In males, the phenotypic correlations of conduct disorder symptoms with all three substance use variables were explained largely by overlapping genetic influences. In females, overlapping shared environmental influences predominantly explained the phenotypic correlation between conduct disorder symptoms and tobacco and cannabis use. Conduct disorder symptoms in early adolescence appear to moderately predict substance use in early adulthood. In males, genetic influences seem to be most important in explaining the relationship between conduct disorder symptoms and substance use, whereas in females, shared environmental influences seem to be most important.

Smokeless tobacco in the form of snus, a moist powdered pouch tobacco held between lips and gums, is legal in Sweden and has become a popular form of early nicotine exposure in Finland, leading us to test the association between snus experimentation in late adolescence and daily cigarette smoking in early adulthood among men. At age 17, we assessed lifetime use of cigarettes and snus with other potential predictors of cigarette smoking status at young adult follow-up. Analyses focused on 375 men who were never smokers at baseline; when adjusted for age, monthly alcohol intoxication, maternal smoking and peer drug use, snus experimentation increased the likelihood of later daily cigarette smoking nearly four-fold. Results (Araneda et al., Reference Araneda, Korhonen, Laatikainen, Haukkala, Rose and Kaprio2019) cannot confirm a causal association, but they do suggest that snus experimentation might constitute a risk for initiating use of other tobacco or nicotine products.