We investigated the frequency of AUDs in patients after BS based on a large patient sample with a post-bariatric history of up to twelve years by analyzing survey data as well as health claims data from the same patient population. This approach enabled us to at least partially compensate for the inherent disadvantages of one data source (health claims data: no information on lifestyle or subthreshold diseases; survey data: e.g. inherent potential recall bias, inaccuracy of medical information) with the strengths of the other data source (health claims data: e.g. accurate depiction of comorbidities and cross-sectoral health care resource utilization over a long observation period; survey data: e.g. documentation of lifestyle behaviors, trend in post-surgical body weight, evaluation of subthreshold mental disorders). This helped us to get a more complete picture of the research subject.

Based on health claims data, we found a proportion of 2–3% who had been diagnosed with AUDs at least once during the observation period. According to AUDIT, one in ten patients reported risky drinking, with almost six percent meeting the criteria for hazardous alcohol consumption and almost four percent meeting the criteria for alcohol dependence. Almost twice as higher proportions were found when increasing the weighting or the scoring of the AUDIT items that relate to the consumption patterns of alcohol, as well as decreasing the original cut-off scores of the instrument. According to these sensitivity analyses which have been introduced as to our knowledge the AUDIT has not been validated in post-bariatric patients before, one in three male patients and one in seven female patients revealed at least risky consumption patterns.

The performance of our study participants is comparable to previous findings reported from the general population and from patients attending to general practices [25, 26]. Unlike the AUDIT screening, which was applied only once in our study after surgery, the use of health claims data allowed us to count the frequencies of AUDs pre-/post-surgery. Notably, up to 70% of patients with AUDs were diagnosed post-surgery but not before. This is in line with previous studies which reported an increased risk of incident AUDs after BS in patients without an AUD history, based on established psychometric instruments [8, 12, 27, 28]. However, we wish to emphasize that the interpretation of AUDs based on health claims data warrants great caution, since miscodings cannot be fully excluded. While inpatient discharge diagnoses usually can be considered valid, the accuracy of single outpatient diagnoses can be questionable. We attempted to overcome this shortcoming by requiring outpatient diagnoses to be confirmed by at least another outpatient diagnosis (either consecutive or from different physicians). Nonetheless, we attribute a higher diagnostic validity to the group of patients with at least one inpatient diagnosis available in our data than to the group with outpatient diagnoses without the necessity of an inpatient diagnosis. However, it must be clearly stated that our results are of a purely descriptive nature and that a direct, causal relationship between bariatric surgery and the occurrence of AUDs cannot be assessed with our study design. The lack of an otherwise comparable control group but without bariatric intervention is a clear limitation of our study.

In essence, the frequencies we found for severe alcohol disorders are of the same order of magnitude as those reported for non-bariatric populations. In a German nationwide health study, Jacobi et al. [29] found a 12-month prevalence for alcohol dependence of 3%, based on a structured clinical interview (with men affected about four times more often than women). Similarly, in a re-analysis of several European epidemiology studies, Wittchen et al. [30] reported a 12-month prevalence of 3.4% for alcohol dependence. It should be considered, however, that these figures were derived from the general population. In contrast, our data refer to a population for whom AUDs are actually an exclusion criterion for BS and for whom reduced alcohol consumption (if not abstinence) is indicated after surgery [7]. Thus, these figures still give rise to concern. In this context, however, it is remarkable that, although the frequency of alcohol dependence is comparable, the proportion of those without manifest dependence yet with latent alcohol abuse is three times higher than in the general population [29]. That is, although patients in our sample were not more often alcohol dependent than the general population according to the AUDIT, the proportion of patients with “risky drinking” was substantially increased which might indicate that BS puts patients at risk of substance abuse (although, as mentioned before, this cannot be verified with our study design). The incorporation of our study results into the previously reported frequencies of post-bariatric AUDs is complicated due to the methodological heterogeneity of the studies. A recently published review which included 18 articles revealed inconclusive findings with studies reporting both worsening and improvement of drinking behaviors after surgery [31]. Svensson et al. [32] compared the development of AUDs in bariatric patients and in non-bariatric controls with obesity in a prospective cohort study. Almost seven percent of patients after bariatric interventions reported an alcohol consumption pattern beyond low risk, which is slightly lower than our estimates in the main analyses and corresponds to our estimates in the sensitivity analyses, which replaced missing values on the item regarding the quantity of drinks. However, while this work is among the few with a longer follow-up, these results have limited comparability with ours because they are derived exclusively from inpatient data and from patients without a pre-surgical history of AUDs, potentially limiting the generalizability of their findings. Suzuki et al. [33] reported AUDs two times more often (12%) in inpatients after BS, also using the AUDIT instrument for the detection of alcohol disorders, and they found no association between post-surgical weight loss and AUDs. However, patients with a history of AUDs and specific types of surgery were more likely to have post-surgical AUDs. The authors themselves acknowledge that the precision of their study is limited by the small number of patients included (n = 51). The validity might have been compromised by a low response rate (11–22%), which potentially introduced bias to their study. According to structured clinical diagnostic interviews of 200 patients up to three years after a RYGB, eight percent developed an AUD and almost half of these patients had no pre-surgical history of AUDs [12]. Similar results were reported for other addictive disorders covered by the interviews. Also, when comparing pre- and post-surgical prevalence rates of high-risk drinking according to the AUDIT-C instrument, Wong et al. [28] reported an almost two times higher proportion of affected patients one year after surgery (23% vs. 13%). Another recently published study investigating AUDs in patients after bariatric surgery by using the AUDIT-C was conducted by White et al. [34]. Based upon a multicenter prospective cohort study on 217 adolescent patients (aged 13—19 years) undergoing BS, the authors concluded that nearly half of all patients screened positive for AUDs during the follow-up of up to eight years. While this study is among the few covering a longer observation period, these findings are difficult to align with ours due to the different age ranges under study. However, especially regarding the age range covered by White and colleagues, which is of high public health relevance, these findings are important and point out to further studies.

We found higher rates of AUDs in patients whose surgery had occurred a longer time ago compared with patients with a shorter post-surgical history. While this finding has to be treated with caution since longer observation periods per se allow a higher cumulative incidence, this result is consistent with several studies suggesting a slow rather than a rapid development of AUDs over several years after surgery, putting emphasis on the need for long-term follow-up and care. For instance, results from a previously published multicenter cohort study covering two thousand patients after BS and a maximum follow-up of seven years [9], prevalence of substance disorders, including AUDs, ranged between 7% at baseline and up to 16% seven years after surgery. Notably, the prevalence as well as the cumulative incidence of AUDs varied by type of bariatric intervention and AUDs were more than twice as high for patients with RYGB as for patients with gastric banding. Again, no data were available for patients with sleeve gastrectomy, as at the time of study conduction this intervention was less common than it is today. As more than 40% of our patients had this type of intervention, our study provides further insights.

However, while the aforementioned studies generally suggest a steady increase in AUDs after surgery, there is also evidence of a more uneven or even reversible course. Wee et al. [35] presented findings from a multicenter study on 375 patients with a post-surgical follow-up of two years. In addition to a proportion of 7% of patients with incident high-risk drinking, more than half of patients with pre-surgical high-risk drinking discontinued this behavior after surgery. Similarly, findings from a re-analysis of ten studies on the risk of post-surgical AUDs demonstrated an increased risk from the third year on but not during the first two years after surgery [13]. However, these findings warrant caution as the majority of the considered studies included comparably fewer participants (eight studies covered 800 patients or fewer), had shorter durations (three years or less in six studies) or offered cross-sectional data only (four studies). The authors emphasize the need for long-term investigations to determine if there is a true increase in AUD prevalence in the context of BS. Similarly, Sen et al. [36] investigated the post-bariatric risk of AUDs as estimated with AUDIT in 183 patients up to 6 years after sleeve gastrectomy and reported a reduction in AUDIT scores in the first 3-year follow-up and an increase in the 4–6-year follow-up. Therefore, regarding the sample size and the long-term follow-up in our study, our data might contribute to these findings. In this regard, it is a limitation that we used the AUDIT screening only once after surgery. Thus, we could not determine a change of drinking patterns with this instrument. However, with all due caution the results from the health claims data discussed earlier might hint at an increasing rather than a declining trend.

We found increased health care resource utilization in patients with risky drinking, even in those patients who showed hazardous consumption but not (yet) alcohol dependency. This held true for the number of hospitalizations as well as for the number of inpatient days. Interestingly, for both measures, these differences only emerged from the second year after surgery, but not during the first year or before. This supports the previously mentioned studies, which suggest a slow rather than an immediate development of AUDs. It is important to note that these differences in our data cannot be explained by higher comorbidities. Although some obesity-related comorbidities (e.g. hypertension, diabetes) were less frequent after surgery than before, the comorbidities between patients with low-risk consumption and hazardous alcohol consumption/AUDs were largely comparable at both points in time. This was true with the exception of smoking and treatment with antidepressants, which can be explained by a higher association of AUDs with depressive disorders and which has been reported in the literature [37]. Against this background, it is noteworthy that according to health claims data, patients with AUDs do not appear to receive or seek psychotherapy more often than patients without. Undertreatment of AUDs, although already a known problem [38], would be of particular concern in this patient population for whom close follow-up care is indicated. However, in the questionnaire, patients with AUDs reported being treated by psychiatrists or psychotherapists more often. This obvious contradiction in our data cannot be satisfactorily resolved. It is not unlikely that laypersons do not know the difference between psychiatrists and psychotherapists. However, since in Germany, the services of both professions can be reimbursed equally by statutory health insurance providers, this unawareness would not explain the difference.

This study was the first to investigate AUDs in patients after BS using survey data and health claims data from the same study population. In addition to the long post-surgical period of up to twelve years, the large sample size also facilitated stratification by sex, enabling us to investigate AUDs in male patients also. This is important, since on the one hand, women with BS surgery clearly outrank men in sample size in most studies, while on the other hand in most populations, alcohol-related disorders are more common in men. Moreover, the use of health claims data—available from each patient irrespective of survey response status—also allowed us to assess whether non-responders and responders differed in terms of sociodemographic and clinical parameters, which was not the case [17]. It can therefore be assumed that the problem of responder bias, which occurs very frequently in field studies, was considerably lower in this study. In addition to the need for caution in the interpretation of our health claims data, as already discussed, a further limitation is that only the F-diagnoses were used. We did not use diagnoses that indicate consequences of AUDs or severe drinking (e.g. fat liver, liver cirrhosis). Furthermore, it can generally be assumed that the extent of alcohol disorders is significantly underestimated in both data sources: only cases of diagnosed AUDs are documented in the health claims data, while clinically significant but not yet full-blown AUDs cannot be identified here. In the survey data, an underestimation can occur due to untrue statements about drinking behavior on the part of the participant and/or a lack of understanding of the “drinking units” as queried in the AUDIT. Against this background, the prevalences we found must be critically scrutinized. This also holds true as the standard cut-off score of the AUDIT might underestimate the true prevalence since it has not been validated for this specific population, which might require lower cut-off recommendations due to altered metabolism of alcohol. Another methodological limitation stems from the gap of at least three years between BS and the survey, as only patients with BS between 2004 and 2018 could be included, but the survey was conducted in 2021. Therefore, it is not possible to use the AUDIT to infer drinking patterns during the first two years after surgery. However, in the context of previous research, we estimate this limitation to be minor because, as discussed earlier, many studies already considered the early years after surgery.

In conclusion, the proportion of patients with BS and with alcohol disorders gives rise to concern as alcohol consumption should be restricted after BS. The results suggest the necessity for close monitoring and post-surgical care.