The authors concluded that their review demonstrates a very high lifetime prevalence of AUD among people with a PD. They estimate that, approximately three quarters of people with antisocial PD and about half of those with borderline PD will experience an AUD at some point in the life course…


Nikki Guy, Giles Newton‐Howes (email: giles.newton‐, Hannah Ford, Jonathan Williman and James Foulds


Guy, N., Newton-Howes, G., Ford, H., Williman, J. and Foulds, J. (2018). The prevalence of comorbid alcohol use disorder in the presence of personality disorder: Systematic review and explanatory modelling. Personality and Mental Health, 12(3), pp.216-228.

Personality and Mental Health
Release date

Prevalence of Comorbid Personality Disorder and Alcohol Use Disorder

Research article



Seventy years ago, alcoholism was defined by the first edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM) as a type of sociopathic personality disturbance or more simply, a “personality disorder” (PD) (American Psychiatric Association, 1952; Newton-Howes & Foulds, 2018). While today PDs and alcohol-use disorders (AUDs) are understood to be separate constructs, there remains considerable evidence that they are related. Beyond their chronic and highly stigmatised nature, they share common causal risk factors, such as childhood abuse (Fergusson et al., 2013) which has been hypothesised to bring about a shared pathophysiology in the endogenous opioid system (Bandelow & Wedekind, 2015).


The authors used a broad search strategy across Medline, Cochrane, Embase and PsycINFO databases from 1980 until 24 November 2015 as well as a hand search of the references in papers they cited.

Two authors independently extracted data and conducted quality assessments with a 5-item tool adapted from a Newcastle Ottowa design.

The primary outcome was the lifetime prevalence of any AUD among individuals with a PD and secondary outcomes included lifetime prevalence of alcohol dependence (AD), 12-month AUD prevalence and 12-month AD prevalence. Prevalence outcomes are non-normally distributed and so Freeman-Tukey transformations were applied.

Random effects meta-analysis models were then fitted with a multilevel framework to account for some studies reporting more than one type of PD. Heterogeneity was assessed using the I2 statistic and by examining outcome variance within studies (τ2 at level 2) and between studies (τ2 at level 3). After model fitting, pooled prevalence estimates and confidence limits were calculated.

For lifetime AUD prevalence, the analysis was extended to include moderators such as sample type (clinical vs general population) and type of PD (antisocial (ASPD) vs borderline (BPD) vs other PD).

The contribution of these predictors was assessed by using the Bayesian Information Criterion to assess model fit and by comparing the total variance at levels 2 and 3 (τ2) in these models with the τ2 in the original model with no predictors.

To assess for publication bias, the authors performed a univariate regression of lifetime AUD prevalence against its variance.

The authors performed a sensitivity analysis to determine the influence of excluding studies with special populations or those of low quality.


The literature search generated 3,421 records. 293 were screened in full for eligibility and 20 were included in the quantitative synthesis.

All studies reported using DSM criteria to diagnose PDs and AUDs.

Thirteen articles (from 10 unique studies) reported Lifetime AUD prevalence. Since some articles reported outcomes for >1 type of PD, 16 outcomes in total were meta-analysed, giving a pooled estimate of 58.7% (95% CI: 46.4 to 70.6) for lifetime AUD prevalence among people with a PD.

When the type of PD reported was added to the model, the model fit improved and type of PD was found to be a statistically significant moderator (Q=20.9, df = 2, p<0.0001).

When dividing participants by PD type, lifetime AUD prevalence estimates were:

  • Antisocial personality disorder (ASPD): 76.7% (95% CI: 65.7 to 86.1)
  • Borderline personality disorder (BPD): 52.2% (95% CI: 41.6 to 62.5)
  • Other (remaining types including combined and undifferentiated PD): 38.9% (95% CI: 28.9 to 49.4).

There was no evidence of a significant difference in lifetime AUD prevalence between the general population and clinical populations.

Eight articles (from 6 unique studies) reported lifetime AD prevalence. Three were in clinical populations, two in prisons and three studies reported on the United States Epidemiologic Survey of Alcohol and Related Conditions (NESARC).

In clinical and general population samples, lifetime AD prevalence estimates ranged from 21% to 47% (median 35%). Among prison samples, estimates of lifetime AD prevalence were considerably higher (45% to 97%). These estimates were not meta-analysed due to very high heterogeneity (I2 = 98.5%).

Five papers from 4 unique studies measured 12-month AUD prevalence, which ranged from 12.9% to 31.1%. Outcomes were not meta-analysed due to heterogeneity (I2 = 98.5%).

Four papers discussed 12-month AD prevalence. Three of these reported NESARC data and the fourth had only 11 participants, so these findings were not meta-analysed. Estimates of 12-month AD prevalence among the NESARC papers ranged from 12.9% in patients with obsessive compulsive PD to 31.1% in patients with BPD.

The assessment of bias demonstrated that smaller studies reported higher prevalence estimates. Authors reviewed studies with the highest figures and noted that these were reported in treatment seeking ASPD samples or in special populations. They then excluded both low quality studies and those in offender or indigenous populations and found a reduced estimate of lifetime AUD prevalence (49.6%; 95% CI: 34.7 to 64.6).


The authors concluded that their review demonstrates a very high lifetime prevalence of AUD among people with a PD.

They estimate that, approximately three quarters of people with antisocial PD and about half of those with borderline PD will experience an AUD at some point in the life course.

Implications for practice

The finding that AUDs (alcohol use disorders) and PDs (personality disorders) are highly comorbid has significant implications for how adult mental health and substance misuse services should be designed and delivered. The authors rightly note that the most obvious step to take is to thoroughly screen PD-presenting patients for AUDs, and vice versa. Perhaps more difficult is determining how to manage the comorbidity, which is associated with greater disability than either individual disorder.

Therapies do exist for comorbid substance-use and personality disorders, including dialectical behaviour therapy (DBT) (Linehan et al., 1999), dynamic deconstructive therapy (Gregory et al., 2008) and dual focus schema focused therapy (Ball et al., 2011). However, their results are variable and DBT, perhaps the most studied intervention for PDs has only been trialled specifically for AUDs once (Maffei et al., 2018). Further research is needed to determine the best course of action when these chronic, debilitating conditions present together.

Source Website: National Elf Service