The association between alcohol use, alcohol use disorders and tuberculosis (TB). A systematic review

AMSTAR, a measurement tool to assess the methodological quality of systematic reviews [10], and the overview by Egger and colleagues [11] were used as a guide to conduct this systematic literature review.

A systematic literature search on individual epidemiological studies reporting a prevalence of a) heavy alcohol use/AUD among TB patients and b) TB among patients with AUD was performed in multiple electronic bibliographic databases including: Ovid MEDLINE, PubMed, EMBASE, Web of Science (including Science Citation Index, Social Sciences Citation Index, Arts and Humanities Citation Index), PsycINFO, CABS (BIDS), WHOLIST, SIGLE, ETOH, Google Scholar, and the Cochrane Database of Systematic Reviews. The available published and unpublished literature was searched up to September 2008 inclusive. The search was conducted using multiple combinations of the following key words: tuberculosis, alcohol, abuse, misuse, dependence, alcohol use disorders, prevalence and incidence. In addition, a manual search of the bibliographic pages of selected articles and reviews as well as the content pages of major epidemiological journals [including The International Journal of Tuberculosis and Lung Disease (IJTLD)] were conducted. The search was not limited geographically and to English language publications.

Studies were excluded from the analysis for any of the following reasons: article did not have sufficient data on prevalence of heavy alcohol use/AUD among TB patients or of TB among patients with AUD; a meta-analysis or systematic review; duplicate publication of the same study; and articles available in abstract form only.

The titles and abstracts, where available, were independently reviewed by two researchers to identify potentially relevant papers. The papers were obtained and independently read in full by two researchers. Differences were resolved by discussion and a third party, if necessary. Reasons for exclusion were identified. The data were extracted based on the inclusion and exclusion criteria defined above.

Training of coders to achieve sufficient (> 0.80) interrater reliability was conducted. Interrater reliability gives a score of how much homogeneity, or consensus, there is in the ratings given by different raters. In order to calculate interrater reliability, Fleiss' kappa statistics using the attribute agreement analytic method was used [12].

Using a standardized Excel spreadsheet, each study was coded for the following variables: reference, setting [country/study years], study sample, type of TB, percent of patients with: a) heavy alcohol use/AUD among TB patients and b) TB in patients with an AUD diagnosis, and definition of heavy alcohol use/AUD. A second member cross-checked the table entries for accuracy, against the original article.

The quality of the data from individual articles was assessed based on the following domains as recommended by U.S. Department of Health and Human Services [13]: comparability of subjects, exposure, outcome measurement, and funding or sponsorship.

In the literature search for different causal criteria and pathways, priority was given to systematic reviews and meta-analyses. Special search strategies [14] were used. The same data bases as for the other research were used, with the key words alcohol, TB, plus one of the following: longitudinal, consistency, dose-response relationship, biological pathways, confounding, and reversibility. Only the most comprehensive reviews were included for the discussion of different causal criteria.

Comparison of the standardized mortality rates in Russia and the Ukraine was based on the WHO mortality data bank (European standard population).

AAFs for TB deaths in 2002 in selected countries were calculated based on relative risk (RR) information from the most recent meta-analysis [9]. AAFs were calculated using the formulas of Walter [15, 16], using a disaggregated approach with 5 age groups (15-29; 30-44; 45-59; 60-69; 70+ years). For exposure, we used the estimates of heavy drinking from the 2002 Comparative Risk Analyses for WHO [17]. This model is restricted to adults and only captures the effect of alcohol on oneself, not potential effects on others in spreading the disease.

Even though these studies [14] controlled for some confounding, residual confounding of social and other risk factors for TB such as smoking or drug taking cannot be ruled out [18]. Unfortunately, there is no sufficient basis for the quantification of confounding. To be conservative, we use the standard method for potential confounding of tobacco risks (i.e. halving the effect) [19].

The underlying work is based on systematic reviews of published data, and thus did not require ethical review.

The electronic searches yielded a total of 1,537 publications regarding epidemiological prevalence studies (Figure 1).

After 321 duplicate articles were removed, 1,216 were screened using titles and abstracts. 432 full-text articles were retrieved for further consideration. Cross-checking of references and hand-searches of the IJTLD did not yield any additional studies to be included. A total of 53 studies met inclusion criteria: 47 studies reported a prevalence of heavy alcohol use/AUD among TB patients and 6 studies reported a prevalence of TB among patients with AUD. Studies were in English (50), Russian (2) and Romanian (1) languages. A list of excluded studies is available from the authors. The studies were found for the following countries: USA (29 studies), Russia (5), Canada (4), Denmark (3), India (3), Brazil (2); and one study in each of Australia, Belarus, Finland, Germany, Kazakhstan, Romania, Slovenia, Sweden, and the UK.

At least one meta-analysis or systematic review was found for each of the causal criteria, except for establishing the dose-response relationship and reversibility. In the following, the results with respect to the causal criteria will be presented and discussed under the respective subheading. A general discussion will follow.

There have been clear associations found between heavy alcohol use/AUD, variously defined, and TB. Additional file 1 provides an overview of the prevalence of heavy alcohol use/AUD in various samples of TB patients. The prevalence of heavy alcohol use/AUD in most samples of men and women is markedly higher than in the general population, or than in matched control groups. As expected, the prevalence of TB among people with AUD was also elevated (see additional file 2). In addition, TB was found to be a major cause of death among people with AUD [20].

A formal meta-analysis, conducted by one of the co-authors of the present manuscript, was based on 3 cohort and 18 case-control studies with individual-level data on alcohol exposure and TB [11]. The objective was to systematically review alcohol use as a risk factor for active TB. The pooled RR across the 11 studies that used an exposure cut-off level set at 40 g alcohol per day or above, or defined exposure as a clinical diagnosis of an AUD, was 3.50 (95% CI: 2.01-5.93). After exclusion of 3 small studies, because of suspected publication bias, the pooled RR was 2.94 (95% CI: 1.89-4.59). Several sensitivity analyses confirmed this substantial risk of TB for heavy alcohol use or AUD [9]. The pooled effect size across studies that had controlled for infection status (OR 4.21, 95% CI: 2.73-6.48), suggest that one possible causal pathway is through increased risk of progression from infection to disease. This meta-analysis did not find an association between consumption of less than 40 g alcohol per day and risk of TB.

Two recent meta-analyses have assessed risk factors for TB clustering (as a marker for recent TB transmission), and both found that alcohol abuse (not further defined) was a risk factor for clustering. The first, which was based on 36 genotyping studies in 17 countries [21], found significant effects of alcohol abuse on TB clustering for both low TB incidence (OR = 2.6; 95% CI: 2.1-3.3) and high TB incidence (OR = 1.4; 95% CI: 1.1-1.9) studies. The significant effects remained after adjustment [21]. The other meta-analysis, based on 30 studies, reported that recent transmission of TB was associated with heavy alcohol use (OR 2.27; 95% CI: 1.69-3.06) [18].

In summary, the evidence indicates a clear and strong association between heavy alcohol use/AUD and both risk of developing active TB and risk of recent TB transmission.

Most of the studies in [9] were case-control studies, where either newly detected or recurrent cases of active TB were compared to control populations. Exposure (either heavy use or AUD) was measured retrospectively in these studies, sometimes for the past year only, or for longer time periods. However, since AUD takes usually years to develop, the temporal sequence of AUD before TB seems to be confirmed in case-control studies with newly detected TB as the outcome variable. It is hardly conceivable that the AUD could have developed after the onset of active TB and before the subsequent interview (usually within weeks), for example as a reaction of the patient to being diagnosed with TB. Moreover, there are a few cohort studies where the temporality has been clearly established (see [22, 23]).

Thus, the available evidence clearly supports a temporal sequence with heavy alcohol use/AUD preceding TB.

Overall, the association between heavy alcohol use/AUD and TB is quite consistent (see additional files 1 and 2; and the results of the three meta-analyses). The association has been replicated in studies in different countries, using different settings and different methodologies. Formal tests of consistency have been conducted in the meta-analysis of Lönnroth and colleagues [9] for the subset of studies which could give indicators of effect strength in the form of RRs. Marked heterogeneity was found when all measures of exposure were included, that is if dichotomous measures of any alcohol consumption at a time before incidence of TB (present/absent) were included [9]. This original heterogeneity decreased considerably after subdividing studies into low- and high-exposure levels (high-exposure defined as a measure of AUD and/or average consumption of ≥ 40 g/day). However, statistically significant heterogeneity persisted in the high-exposure category, which may be due to either the rather wide definition of exposure (including definitions based on average volume as well as on psychiatric classification), to misclassification of exposure across studies, to different types of control groups used, or to the different endpoints included (i.e., pulmonary TB vs. all TB; first incident TB vs. recurrent TB). While heterogeneity was statistically significant, it should be stressed that no study for heavy alcohol use/AUD showed a RR below 1, and all tests, including on various subsets of studies controlled for different confounders, showed pooled RRs higher than 3 [9].

In summary, there is a consistent relationship between heavy alcohol use/AUD and TB, and more formal measures of risk stemming from controlled studies were consistently positive with relatively large effect sizes.

There are few studies that allow an explicit test of a dose-response relationship between alcohol use and TB. The work of Brown and Campbell [24] is the exception, and from the data provided in that paper, the following odds ratios can be calculated for different levels of exposure: 10-25 ml/day: 1.60 (95%CI: 0.57-4.55); 26-50 ml/day: 2.38 (95% CI: 0.89-6.44); 51-75 ml/day: 9.27 (95% CI: 2.77-32.58); 76-100 ml/day: 8.50 (95% CI: 1.93-40.54); >100 ml/day: 35.55 (95% CI: 6.41-260.70). However, these findings are based on a small sample of men, and the confidence intervals are wide. A large case-control study of Russian mortality data [25] published as these analyses found, in comparison to light drinkers, the following dose-response relationship beyond an apparent threshold: <1 half-litre bottle of vodka or equivalent per day relative risk 1.01 for men (95% CI 0.83-1.22), 0.93 for women (95% CI 0.64-1.35); one to less than 3 bottles 1.97 for men (95% CI 1.64-2.36), 4.06 for women (95% CI 2.97-5.56); 3 bottles and above 4.14 for men (95% CI 3.44-4.98), 5.32 for women (95% CI 3.70-7.65).

Most other studies use dichotomous exposure with some variant of alcohol "abuse", either operationalized as average heavy drinking or by using AUD as the defining criterion. Assuming an exponential dose-response relationship like that found in the studies cited above is true, one would of course expect that any dichotomous variable associated with heavy drinking would be significant. AUD are strongly associated with frequency of heavy drinking occasions, and thus with overall volume of alcohol consumption [26, 27]. Thus, the current evidence does not contradict an exponential dose-response relationship between average volume of alcohol use and TB. Clarifying the exact relationship between alcohol exposure and TB should be one of the priorities of future research, including a test for gender-specific relationships. But also, independent of a diagnosis of AUD, patterns of drinking have been identified as a risk factor for a number of disease categories over and above volume of consumption [28], and this line of investigation should be extended to TB as an outcome as well.

In summary, there is some evidence for a dose-response relationship between level of alcohol use and risk for TB.

About one-third of the people in the world are infected with M. tuberculosis [29]. Ninety percent of individuals in the general population who become infected with M. tuberculosis will never develop clinical disease [30]. That means that the majority of people infected with M. tuberculosis, those who have an adequate immune system, are able to fight off the infection and do not develop the disease. Only about 10% of those infected develop active TB, where the immune system is not able to fight off the infection [31].

Research to date has shown that heavy alcohol use/AUD is a risk factor for an impaired immune system, and increases a person's susceptibility to active TB infection as well as to the reactivation of latent disease [5, 31‐34]. Alcohol-consuming hosts are considered "immuno-compromised" because the incidence and severity of infectious diseases among them are greater than for abstainers [5, 33‐36]. Individuals with alcohol dependence are particularly susceptible to lung infections such as TB and pneumonia [5, 32‐34].

Greater than 90% of inhaled M. tuberculosis bacteria are normally destroyed by alveolar macrophages [37]. If alveolar macrophages are not able to kill M. tuberculosis, the bacteria multiply within macrophages, and tubercles form in the lungs. In vitro studies have shown that intracellular survival and growth of M. tuberculosis within human macrophages is enhanced by exposure to alcohol [38‐40]. It has also been shown that alcohol use exacerbates murine pulmonary TB, which is associated with alterations in the region-specific CD4+- and CD8+-lymphocyte responses and defective lung granuloma formation [41]. In vivo and in vitro studies have demonstrated that alcohol significantly hinders antimycobacterial defenses by suppressing mobilization, adherence, phagocytosis, and superoxide production of alveolar macrophages [42‐45]. Chronic ethanol ingestion in rats significantly decreases membrane expression of the GM-CSF receptor in alveolar macrophages and, in parallel, decreases cellular expression and nuclear binding of the transcription factor that activates GM-CSF-dependent macrophage functions [46].

Acute and chronic alcohol exposure may decrease the activation of antigen-specific T-cells by inhibiting the macrophage's capacity to present mycobacterial antigen to lymphocytes [47]. In addition, alcohol has been shown to reduce macrophage response to immune system modifiers (e.g., cytokines, including interleukin-6 (IL-6), IL-1β, TNF-α, and IL-8) and to prevent the protective effect exerted by the cytokines [40, 48]. Moreover, acute and chronic alcohol exposure may suppress the capacity of monocytes to produce cytokines, which directly inhibit bacterial growth and play a critical role in cellular communication, activation, proliferation, and migration, as well as regulating inflammation and healing mechanisms [48, 49]. Finally, alcohol may adversely affect antigen-specific T-cell activation so that the Th2 population (humoral immunity) dominates the Th1 population (cell-mediated immunity, responsible for overcoming TB infection). This shift disturbs a balance between the two basic types of immune system, compromising the immune defense and increasing susceptibility to TB as a result of alcohol exposure [42‐45, 50].

The contribution of alcohol to TB through its effects on the immune system has been difficult to isolate from other adverse factors, as it is still not known whether alcohol per se or the other sequelae of heavy drinking and alcohol dependence such as liver damage, nutritional deficiency, or hygienic factors are primarily responsible for the impaired immunity associated with alcohol dependence [5, 33, 51, 52]. However, by whatever mechanism, alcohol exposure clearly has an impact on weakening the immune system and thus on the incidence of active TB.

In sum, there is good evidence for a biological plausibility of causal relationship between alcohol exposure and incidence of TB via weakening the immune system.

AUD and TB have both been labeled "diseases of poverty", and both can be consequences as well as causes of social marginalization [5, 53, 54]. TB incidence follows a socioeconomic gradient, where TB is more incident and prevalent in poorer areas, and among poorer households [54].

The relationship between socioeconomic status and heavy alcohol use/AUD is more complex. On the country level, higher GDP has been strongly associated with higher levels of alcohol use and lower abstainer rates [55, 56]. Within countries, while lower socioeconomic strata often have higher rates of abstention, rates of AUD are not necessarily lower; but stigma and health harm attached to alcohol use have been found higher in lower compared to higher strata [56‐59]. Furthermore, among specific groups such as the homeless or the imprisoned population, AUD rates are much higher than in the general population [60‐65].

Causal relations are not necessarily unidirectional: poverty may lead to AUD, and AUD may thus be on the causal pathway from poverty to increased TB risk. On the other hand, AUD can lead to downward social mobility, and such downward mobility can create or reinforce social conditions that increase the risk of TB.

Social marginalization and the associated conditions such as crowding, malnutrition, homelessness, and imprisonment increases the risk of TB independently of heavy AUD [9]. This means that we should not consider all of the relationship between heavy alcohol use/AUD and TB as causal. But in most studies heavy alcohol use/AUD emerges as a significant independent risk factor for TB, even if other risk factors are controlled for [9].

In sum, there is evidence of a social marginalization and drift pathway on how alcohol could lead to TB.

Unfortunately, our second search on causal criteria (see above) did not yield any controlled trials on the impact of removing heavy alcohol exposure/treatment for AUD on incidence or re-infection of TB. On the aggregate level, we can learn from the natural experiment of the Gorbachev-era campaign which reduced alcohol use in the Soviet Union [66]. In the mid-eighties, the Gorbachev government, through a variety of measures including a decrease in the state controlled production of alcohol, succeeded in markedly reducing per capita consumption of alcohol between 1984 and 1987 from 14.2 l to 10.7 l of pure alcohol (-24.6%). These numbers take into account unrecorded consumption such as illegally produced moonshine [67]. After the campaign, alcohol use went back to old levels (1993: 14.5 l; + 35.5% from 1987). At the same time TB mortality showed similar trends for both Russia and the Ukraine (see Table 1).

Of course, the above data are only ecologic, and alternative explanations cannot be excluded. It is reasonable to believe that dramatic changes in both socioeconomic conditions and gradients, as well as the functioning of public health care systems, contributed to the upsurge in TB incidence in Russia during the 1990s [68]. However, for the reduction of overall mortality between 1984 and 1987, there are not a lot of alternative explanations. In addition, the reduction of mortality was not uniform across all causes of death. In line with the explanation of a causal impact of alcohol, causes of death 100% attributable to alcohol (e.g. alcohol poisoning) showed higher mortality rate decreases, and causes where no impact was expected (e.g., cancer, where any impact of alcohol should be long term and no effect in the first years can be expected [69]) showed no change at all. In addition, other countries in Eastern Europe without a parallel reform did not show decreases between 1984 and 1987 (e.g., Romania showed an increase of 38.8% in men and of 13.0% in women; own calculations - see Table 1). On the other hand, for the increases in TB mortality after 1987 several explanations are possible. The Soviet Union broke apart, the economic situation in countries like Russia, Ukraine or Romania worsened, and unemployment and other indicators of social crisis increased. Thus, TB as a disease category linked to poverty would be expected to increase. It is no surprise in this situation that TB rates in a country like Romania also increased (99.0% in men, 128.1% in women).

In addition to alcohol's role in the onset of TB, there is also strong evidence of a negative influence of heavy drinking/AUD on the clinical course of TB [70‐74]. People drinking heavily or with AUD show higher relapse rates [75], a higher probability of an unfavourable clinical course, and a higher probability of experiencing the most destructive forms of TB [76‐78]. Recent studies have also shown that heavy drinking is positively associated with early relapse (OR 17.7) [79]. There are several reasons that the clinical course of TB is worse in those with AUD than in others - both clinical and social.

Based on the assumptions of Lönnroth and colleagues [9], the proportion of TB attributable to alcohol as a risk factor in different countries is substantial, even under conservative assumptions (right two columns of Table 2 and [102]).

Not surprisingly, given the level of consumption [103], the highest alcohol-attributable fractions were estimated for Russia, and also Nigeria, Thailand and South Africa, especially for men. Overall, about 10% of the TB globally was estimated to be attributable to alcohol.