A, Nonparametric estimate of the hazard function (and 95% CIs) of first AUD registration in wives as a function of a first registration for AUD in their husbands at time 0 compared with the risk when the husband had no AUD history. B, The hazard ratio for first AUD registration in wives who have vs have not been exposed to the onset of AUD in their husbands. The curve represents the ratio between the 2 curves in panel A.
A, Nonparametric estimate of the hazard function (and 95% CIs) of first AUD registration in husbands as a function of a first registration for AUD in their wives at time 0 compared with the risk when the wife had no AUD history. B, The hazard ratio for first AUD registration in husbands who have vs have not been exposed to the onset of AUD in their wives. The curve represents the ratio between the 2 curves in panel A.
eTable 1. Rates of AUD in Total Cohort as a Function of Marital History
eTable 2. Tetrachoric Correlations and Odds Ratios for Ascertainment for AUD Across Criminal, Medical and Pharmacy (Drug) Registries - Males Above the Diagonal and Females Below
Customize your JAMA Network experience by selecting one or more topics from the list below.
Kendler KS, Lönn SL, Salvatore J, Sundquist J, Sundquist K. The Origin of Spousal Resemblance for Alcohol Use Disorder. JAMA Psychiatry. 2018;75(3):280–286. doi:10.1001/jamapsychiatry.2017.4457
What causes spouses to resemble one another in their risk for alcohol use disorder?
In this population-based registry study, the increase in risk for a first onset of alcohol use disorder in a married individual after the onset of alcohol use disorder onset in his or her spouse was large and rapid. When an individual was married in either order to serial partners with vs without alcohol use disorder, the risk for alcohol use disorder was substantially increased when the partner had an alcohol use disorder registration and decreased when the partner did not have an alcohol use disorder registration.
A married individual’s risk for alcohol use disorder is likely directly and causally affected by the presence of alcohol use disorder in his or her spouse.
Although spouses strongly resemble one another in their risk for alcohol use disorder (AUD), the causes of this association remain unclear.
To examine longitudinally, in first marriages, the association of a first registration for AUD in one spouse with risk of registration in his or her partner and to explore changes in the risk for AUD registration in individuals with multiple marriages as they transition from a spouse with AUD to one without or vice versa.
Design, Setting, and Participants
Population-wide Swedish registries were used to identify individuals born in Sweden between 1960 and 1990 who were married before the end of study follow-up on December 31, 2013. The study included 8562 marital pairs with no history of AUD registration prior to their first marriage and an AUD registration in 1 spouse during marriage and 4891 individuals with multiple marriages whose first spouse had no AUD registration and second spouse did or vice versa. Final statistical analyses were conducted from August 15 to September 1, 2017.
A spousal onset or history of AUD registration.
Main Outcomes and Measures
Alcohol use disorder registration in national medical, criminal, or pharmacy registries.
Among the 8562 marital pairs (5883 female probands and 2679 male probands; mean [SD] age at marriage, 29.2 [5.7] years) in first marriages, the hazard ratio of AUD registration in wives immediately after the first AUD registration in their husbands was 13.82, which decreased 2 years later to 3.75. The hazard ratio of AUD registration in husbands after the first AUD registration in their wives was 9.21, which decreased 2 years later to 3.09. Among the 4891 individuals with multiple marriages (1439 women and 3452 men; mean [SD] age at first marriage, 25.5 [4.2] years), when individuals transitioned from a first marriage to a spouse with AUD to a second marriage to a spouse without AUD, the hazard ratio for AUD registration was 0.50 (95% CI, 0.42-0.59) in women and 0.51 (95% CI, 0.44-0.59) in men. After a first marriage to a spouse without AUD, the hazard ratio for AUD with a second marriage to a spouse with AUD was 7.02 (95% CI, 5.34-9.23) in women and 9.06 (95% CI, 7.55-10.86) in men. These patterns were modestly attenuated when moving from second to third marriages. Controlling for AUD registration prior to first marriage or between first and second marriages produced minimal changes in risk.
Conclusions and Relevance
The increase in risk for AUD registration in a married individual following a first AUD registration in the spouse is large and rapid. When an individual with serial spouses is married, in either order, to partners with vs without an AUD registration, the risk for AUD registration is substantially increased when the partner has an AUD registration and decreased when the partner does not have an AUD registration. These results suggest that a married individual’s risk for AUD is directly and causally affected by the presence of AUD in his or her spouse.
Spouses resemble one another regarding the amount of alcohol they consume1-6 and their risk for alcohol use disorder (AUD).7-14 However, the degree to which this resemblance results from assortative mating or spousal interaction remains unclear. Two approaches have been taken to discriminating these mechanisms: examining spousal resemblance over time1,6,15-17 and modeling cross-sectional data with twins and their spouses.3,8 Longitudinal studies, which have focused on alcohol consumption, have found some evidence for partner selection18 but have found more support for spousal interaction.1,6,15-17,19 Prior modeling studies for alcohol consumption and dependence have, by contrast, found evidence for assortative mating.3,8
In this article, we take 2 novel approaches to this question. First, we examine AUD registration in married couples in Sweden to determine how the first registration for AUD in one spouse is associated with the risk of a first registration for AUD in his or her marital partner. Second, we use a within-person design examining the risk of AUD registration in men and women in Sweden who had multiple marriage partners in which at least one spouse had a lifetime history of AUD and one spouse lacked such a history. Registration for AUD was assessed in national medical, criminal, and pharmacy registries, where the disorder is associated with expected risk factors including prior externalizing behaviors and heavy alcohol consumption, low educational attainment, and genetic risk.20
We linked nationwide Swedish registers via the unique 10-digit identification number assigned at birth or immigration to all Swedish residents. The identification number was replaced by a serial number to ensure anonymity. The following sources were used to create our data set: the Total Population Register, containing information about year of birth, sex, and family and marital status from 1968 onward; the Swedish Census, containing household information from 1960, 1965, 1970, 1975, 1980, 1985, and 1990; the Multi-Generation Register, linking individuals born after 1932 with their parents; the Hospital Discharge Register, containing hospitalizations for Swedish inhabitants from 1964 to 2012; the Prescribed Drug Register, containing all prescriptions in Sweden picked up by patients from 2005 to 2013; the Outpatient Care Register, containing information from all outpatient clinics from 2001 to 2012; the Crime Register, including complete national data on all convictions in lower court from 1973 to 2012; and the Swedish Suspicion Register, including national data on individuals strongly suspected of crime from 1998 to 2012. The study was approved by the ethics committee of Lund University in Malmö, Sweden, who waived participant consent because data were deidentified. All procedures contributing to this work complied with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.21
We included individuals born in Sweden between 1960 and 1990 who were married before the end of follow-up on December 31, 2013. We identified spouses individuals getting married on the same date and obtaining the same family or household identification number after marriage. By our definition, therefore, a spouse needed to reside with his or her partner. However, if an individual was married multiple times and we were not able to unambiguously identify the spouse in all marriages, we included the marriages for which we had good information. For our first analysis, we identified couples where neither had a history of AUD before marriage and one spouse had a first registration for AUD during their first marriage. Spousal AUD was defined as an AUD registration during the current marriage that did not occur on the same day as that of their partner. We identified a comparison control group of individuals who were not exposed to an affected spouse. For our second set of analyses, we identified individuals with multiple marriages. For these analyses, spousal history of AUD was defined as prior to or during the marriage under consideration.
Individuals’ AUD status was identified from Swedish medical registries by the following International Classification of Diseases, Eighth Revision, codes: 571.0, 291, 303, and 980; the following International Classification of Diseases, Ninth Revision, codes: V79B, 305A, 357F, 571A, 571B, 571C, 571D, 425F, 535D, 291, 303, and 980; the following International Classification of Diseases, 10th Revision, codes: E244, G312, G621, G721, I426, K292, K700, K701, K702, K703, K704, K709, K852, K860, O354, T510, T512, T511, T513, T518, T519, F101, F102, F103, F104, F105, F106, F107, F108, and F109; and from the Prescribed Drug Register if they received disulfiram (Anatomical Therapeutic Chemical Classification System N07BB01), acamprosate (N07BB03), or naltrexone (N07BB04). In addition, we identified individuals with AUD as those convicted for or suspected of at least 2 alcohol-related crimes according to law 1951:649, paragraph 4 and 4A, and law 1994:1009, Chapter 20, paragraph 4 and 5 from the Swedish Crime Register and code 3005 and 3201 in the Suspicion register.
Statistical analysis was conducted from August 15 to September 1, 2017. We identified the first individual with AUD within a marriage and then followed his or her spouse until the spouse’s first AUD registration or censoring. For comparison, we identified a control group matched on sex, birth year, year of marriage, family history of AUD (in a sibling or parent), and parental educational level. One unexposed (without spousal AUD) control individual was selected for each exposed (with spousal AUD) proband. We estimated smoothed hazard functions nonparametrically for the exposed individual and the unexposed control using the R package bshazard.22 To obtain an empirical estimate of how the hazard ratio (HR) changed during follow-up, we divided the 2 smoothed hazard curves (exposed and unexposed) at each estimated bin. Confidence intervals for this curve cannot be obtained in a corresponding way. However, comparing the CIs for the 2 original hazard functions will provide an indication of the level of significance of their difference. Nonoverlapping 95% CIs implies P < .05.
Second, we focused on within-individual contrasts by modeling serial marriages and divorces using a frailty model for survival data, assuming a gamma distribution. We included up to 3 marriages and divorces per proband and observed the proband until the first registration for AUD within each marital phase. Marital status, including spousal history of AUD, was included as a time-dependent covariate and we censored at death, widowhood, migration, or end of follow-up (year 2013), whichever came first. We allowed the risk of AUD to depend on the spouse’s history of AUD and to differ between first, second, and third marriages and divorces. In a second step, we allowed these associations to depend on the spouse’s AUD status before first marriage, and, finally, we accounted for AUD registration between 2 marriages (ie, while divorced). Men and women were analyzed separately and we present only the contrasts of interest. The statistical analyses were performed using the survival package in R, version 22.214.171.124
Our first analyses were of pairs of spouses in which neither individual had a prior AUD registration. When the first individual in the pair had an AUD registration, we then examined the pattern of risk for a first registration for AUD in the spouse, compared with the spouse of a control individual whose partner remained unaffected by AUD.
We identified 5883 marital pairs in which the husband first developed AUD and 2679 marital pairs in which the wife first developed AUD, as well as matching samples of control pairs (Table 1). First onsets of AUD in both wives and husbands were much greater in pairs in which the spouse had an AUD onset than when the spouse did not. Furthermore, that increase was especially concentrated in the first year after spousal onset of AUD.
Using nonparametric survival methods, the hazard function (with 95% CIs), representing the absolute risk of a first registration for AUD for the wife conditional on her husband’s first registration for AUD, is seen in Figure 1A and can be compared with the corresponding risk in controls. The 95% CIs do not overlap, indicating that they are significantly different. More easily interpretable is the HR, which is seen in Figure 1B. Immediately after the husband’s first AUD registration, the HR for AUD registration in the wife equaled 13.82. This risk then fell rapidly over the first 2 years to approximately a 4-fold excess (HR, 3.75), which then stayed relatively constant from 2 to 6 years after the husband’s first registration for AUD.
The pattern of results was broadly similar when we examined the effect of wives’ first AUD registration on the subsequent risk for husbands’ first AUD registration (Figure 2A and B). However, the short-term increase was slightly smaller (HR, 9.21) in husbands and the risk fell more slowly than in wives, stabilizing at an approximately 3-fold increase (HR, 3.09) after 3 years.
We next examined the risk for onset of AUD during marriage in individuals who had multiple marriages, one of which was to a spouse with AUD and one to a spouse without AUD. The proportion of individuals married multiple times in our population sample and the associated frequency of AUD registration is seen in eTable 1 in the Supplement. The sample size of such individuals is given in Table 2. More individuals with multiple marriages went from a spouse without AUD to a spouse with AUD than vice versa.
Table 2 presents 5 complementary analyses, always comparing the later marriage with the earlier one. We estimated that when individuals move from a marriage to a spouse with AUD to a marriage with a spouse without AUD, that move would be protective and such individuals would have a lower risk for AUD. By contrast, when individuals move from a marriage to a spouse without AUD to a marriage with a spouse with AUD, that move would be a predisposing marital transition and increase the risk for AUD.
Starting with husbands, the initial contrast compares first and second marriages, controlling only for the husband’s AUD prior to his first marriage. Moving from a spouse with AUD to a spouse without AUD reduced the husband’s risk for AUD by half (HR, 0.50; 95% CI, 0.42-0.59) (Table 2). By contrast, transitioning from a spouse without AUD to a spouse with AUD increased his risk for AUD 7-fold (HR, 7.02; 95% CI, 5.34-9.23). We then repeated the same analyses, comparing second with third marriages. The protective effects of moving from a second wife with AUD to a third wife without AUD were similar to those found with first and second marriages. The predisposing effect of moving from a second wife without AUD to a third wife with AUD was, however, weaker than parallel effects seen with first vs second wives.
Next, we examined first and second marriages in husbands allowing for an interaction between AUD registration before their first marriage and subsequent risk. When husbands had no prior AUD registration, the protective effect of a transition from a first wife with AUD to a second wife with AUD was similar to that seen in all husbands. However, the predisposing effect of having a second wife with AUD after marriage to a first wife without AUD was greater (HR, 11.85; 95% CI, 8.57-16.39) (Table 2). For husbands who had an AUD registration prior to their first marriage, the protective effect of marrying a second wife without AUD was considerably stronger than in all husbands, reducing risk by more than two-thirds (HR, 0.31; 95% CI, 0.70-1.23). However, the predisposing effect of moving from a first wife without AUD to a second wife with AUD was weaker than in the entire sample (HR, 1.84; 95% CI, 1.11-3.04).
In our final comparison, we focused on husbands who did not have an AUD registration in the intermarital interval (between their first and second marriages). In this subgroup, the expected predisposing effect of moving from a wife without AUD to a wife with AUD declined slightly from an HR of 7.02 in the entire sample to an HR of 6.77 (95% CI, 5.20-8.81) (Table 2). The protective effect of moving from a first marriage to a wife with AUD to a second marriage to a wife without AUD was effectively unchanged.
Table 2 also presents parallel analyses of wives with multiple marriages to husbands with vs without a history of AUD. The pattern was very similar to that seen for husbands and none of the HRs differed significantly across sexes.
We used 2 different analytic strategies to clarify the causes of spousal resemblance for AUD that produced consilient results. The risk for AUD in a married individual appears to be causally affected by the onset and/or presence of AUD in his or her spouse.
Longitudinal designs have previously been used in clarifying the correlations in drinking patterns in spouses, demonstrating evidence for causal effects of an individual’s alcohol consumption (and sometimes the consumption of that individual’s peers) on drinking in the husband or wife.1 To our knowledge, there have been no similar designs applied to the study of AUD. Using Swedish national registries, we showed that among spouses—neither of whom had a prior registration for AUD—the registration of one spouse produced a transient, large increase in the risk of AUD risk in his or her marital partner. These effects were relatively similar whether it was the wife or the husband who was first registered for AUD.
Within 2 to 3 years of the first spouse’s registration for AUD, the large increase in risk for AUD in his or her partner declined substantially to a relatively stable baseline risk of approximately 4 times that seen in controls. Although tentative, we suggest that the large increased risk for AUD in an individual temporally closely associated with AUD registration in his or her spouse is a result of spousal interaction—the causal effect of an individual’s drinking on the drinking of his or her spouse. This association may reflect several processes including maladaptive coping to the acute stress of a spouse’s AUD24 and motivation to reduce discrepancies between one’s own drinking and a spouse’s drinking, as discrepancies in partners’ drinking patterns are associated with marital dissatisfaction.25 However, the elevated and relatively stable excess risk for AUD that follows the large increase likely reflects assortative mating—the tendency of spouses to select mates who are similar to themselves in alcohol consumption and other risk factors for AUD.
Our second set of analyses use a within-individual design to compare the risk for AUD registration in individuals who have had multiple marriages, where at least 1 spouse has had AUD and at least 1 other spouse has not had AUD. This design has the important advantage of controlling completely for all genetic, temperamental, and environmental risk factor exposures prior to the first marriage. When an individual moved from a marriage to an individual with AUD to a marriage to an individual without an AUD history, the HR for AUD registration declined around 50%. This is true for both husbands and wives, when comparing a first to second marriage or a second to third marriage, whether including or censoring spouses with an AUD registration prior to their first marriage, and whether including or censoring spouses with an AUD registration between their first and second marriage. However, for both men and women, this protective effect became appreciably stronger when their spouse had an AUD registration before the first marriage.
When an individual with multiple marriages transitions from marriage to an individual without AUD to marriage to an individual with an AUD history, the HR for AUD registration increases 7-fold to 9-fold when comparing a first marriage with a second marriage and 4-fold to 5-fold when comparing a second marriage with a third marriage. This effect was modestly stronger in wives than in husbands, meaning that a husband with AUD may have a stronger effect on his wife’s risk for AUD than a wife with AUD may have on her husband’s risk for AUD. This finding is consistent with an early assertion that the transmission of AUD is stronger from husbands to wives,26 although prior studies of the transmission of drinking behavior between husbands and wives have provided mixed evidence on this point.16,27-29 This predisposing effect is appreciably stronger in both men and women when the identified spouse at risk lacks a prior AUD history and is much weaker when the married individual had an AUD registration before his or her first marriage. However, excluding proband spouses with AUD registrations between their first and second marriages produced no appreciable change in the predisposing effect. This finding is important because the increased risk for AUD of moving from a first marriage to a spouse without AUD to a second marriage to a spouse with AUD could arise because the proband spouse developed AUD after the first marriage and sought a spouse with a similar problematic pattern of alcohol consumption. Our results suggest that this mechanism, if it occurs, produces at most a modest bias in our estimates.
These results should be interpreted in the context of 4 possible methodological limitations. First, we ascertained individuals with AUD from multiple registries. While this approach does not require cooperation or accurate recall from respondents, this method does not reproduce results that might be obtained from studies using structured psychiatric interviews. Our population prevalence of AUD is lower than that found in an interview survey in nearby Norway.30 Our ascertainment methods likely detect severely affected individuals who have experienced medical and legal complications of their drinking. However, the validity of our AUD definition is supported by the high rates of concordance observed across our ascertainment methods (mean [SD] odds ratio, 24.5 [11.7]) (eTable 2 in the Supplement) and by the pattern of resemblance in relatives,31,32 which is similar to that found in studies based on personal interviews.33,34
Second, we do not know how accurately a first AUD registration indexes the onset of serious drinking problems. The lag time could be quite variable. Early phases of problematic drinking in a married individual might increase risk both for divorce and selection of another marital partner with heavy drinking. The chances that this problem produces biases in our findings is reduced by our showing that controlling for AUD registration between marriages had little effect on the predisposing effect of moving from a first spouse without AUD to a second spouse with AUD.
Third, our sample with multiple marriages was atypical and had elevated rates of AUD (eTable 1 in the Supplement). However, our findings are likely valid and generalizable, as the logic of these analyses depended only on within-individual comparisons.
Fourth, our results do not address the mechanisms through which AUD registration in married individuals increases spousal risk. In couples concordant for AUD in our first analyses, sources of AUD registration were weakly correlated (χ24 = 8.24, P = .08; κ [SE] = 0.10 [0.05]). These results suggest that registry-specific effects, which would arise from specific correlations in ascertainment of AUD in couples from criminal or medical sources, played little role in AUD marital concordance.
The risk for AUD is substantially correlated in marriage partners. We have produced results from Swedish epidemiologic samples using 2 different methods that both suggest that an appreciable proportion of that correlation is a result of causal factors acting via spousal interaction rather than entirely through the effects of assortative mating. Alcohol use disorder in one spouse, through a range of psychosocial mechanisms, appears to directly increase the risk for AUD in the partner. Although genetic and biological factors contribute strongly to the predisposition to alcohol dependence, these findings complement our prior work on marriage35 and divorce36 in showing how close social bonds such as marriage can also powerfully influence, for better or worse, the risk for AUD.
Accepted for Publication: December 7, 2017.
Corresponding Author: Kenneth S. Kendler, MD, Virginia Institute for Psychiatric and Behavioral Genetics, Virginia Commonwealth University, Box 980126, Richmond, VA 23298 (email@example.com).
Published Online: February 7, 2018. doi:10.1001/jamapsychiatry.2017.4457
Author Contributions: Dr K. Sundquist had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kendler, K. Sundquist.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Kendler, Salvatore.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Kendler, Lönn.
Obtained funding: Kendler, J. Sundquist, K. Sundquist.
Administrative, technical, or material support: Kendler, J. Sundquist, K. Sundquist.
Supervision: Kendler, K. Sundquist.
Conflict of Interest Disclosures: None reported.
Funding/Support: This project was supported by grant R01AA023534 from the National Institutes of Health; grants K2012-70X-15428-08-3, 2012-2378, and 2014-10134 from the Swedish Research Council; and grants 2013-1836 and 2014-0804 from the Swedish Research Council for Health, Working Life and Welfare, as well as ALF (Avtal om Läkarutbildning och Forskning) funding from Region Skåne.
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.