Prisoners co-infected with tuberculosis and HIV: a systematic review

Introduction Almost from the beginning of the HIV epidemic in 1981, an association with tuberculosis (TB) was recognized. This association between HIV and TB co-infection has been particularly evident amongst prisoners. However, despite this, few studies of TB in prisons have stratified results by HIV status. Given the high prevalence of HIV-positive persons and TB-infected persons in prisons and the documented risk of TB in those infected with HIV, it is of interest to determine how co-infection varies amongst prison populations worldwide. For this reason we have undertaken a systematic review of studies of co-infected prisoners to determine the incidence and/or prevalence of HIV/TB co-infection in prisons, as well as outcomes in this group, measured as treatment success or death. Methods A literature search was undertaken using the online databases PubMed, Embase, IBSS, Scopus, Web of Science, Global Health and CINAHL Plus. No restrictions were set on language or publication date for article retrieval, with articles included if indexed up to 18 October 2015. A total of 1975 non-duplicate papers were identified. For treatment and outcome data all eligible papers were appraised for inclusion; for incidence/prevalence estimates papers published prior to 2000 were excluded from full text review. After full text appraisal, 46 papers were selected for inclusion in the review, 41 for incidence/prevalence estimates and nine for outcomes data, with four papers providing evidence for both outcomes and prevalence/incidence. Results Very few studies estimated the incidence of TB in HIV positive prisoners, with most simply reporting prevalence of co-infection. Co-infection is rarely explicitly measured, with studies simply reporting HIV status in prisoners with TB, or a cross-sectional survey of TB prevalence amongst prisoners with HIV. Estimates of co-infection prevalence ranged from 2.4 to 73.1% and relative risks for one, given the other, ranged from 2.0 to 10.75, although some studies reported no significant association between HIV and TB. Few studies provided a comparison with the risk of co-infection in the general population. Conclusions Prisoners infected with HIV are at high risk of developing TB. However, the magnitude of risk varies between different prisons and countries. There is little evidence on treatment outcomes in co-infected prisoners, and the existing evidence is conflicting in regards to HIV status influence on prisoner treatment outcomes. PROSPERO Number: CRD42016034068


Introduction
Almost from the beginning of the AIDS epidemic in 1981, an association with tuberculosis (TB) was recognized. This relationship is attributed both to the suppression of host immune systems by human immunodeficiency virus (HIV) [1] as well as environmental risk factors shared by the two diseases [2]. People infected with HIV are 20 to 30 times more likely to develop active TB than HIV-negative persons, with an estimated one in three deaths among HIV-positive persons due to TB in 2014 [3]. Although deaths from TB in those living with HIV have fallen throughout the last decade, 360,000 deaths were attributed to co-infection in 2013 [4].
This association between HIV and TB co-infection has been particularly evident amongst prisoners. Between 1976 and 1986 there was a fivefold increase in the incidence of TB among inmates in the New York prison system, attributed in part to infection with HIV [5]. Studies in other high-risk populations such as intravenous drug users added to evidence linking the two conditions [6], as did a 1990 study of the prison population in Maryland that documented an increased risk of infection with TB among prisoners who were HIV positive. This increased risk fell just short of significance, but this study may have lacked adequate power to detect a difference [7].
There are several reasons for this association. First, immunosuppression from HIV infection both predisposes individuals to TB reactivation [8,9] and increases the risk of progression from infection to disease [8,10]. Second, there is some limited evidence from the USA that prisoners with a history of TB may be more likely to engage in behaviours that place them at a high risk of contracting HIV [11]. Third, the two infections share a number of socio-demographic and behavioural risk factors with each other and with the probability of incarceration, such as injecting drug use [2].
Fourth, conditions in prisons, such as poor ventilation and overcrowding, increase the risk of transmission of TB. Fifth, racial and ethnic groups that tend to experience proportionately higher TB rates, such as blacks and Hispanics in the USA, are often vastly over-represented, by up to 30 to 40 times [12], within prison populations [13Á15]. Yet, despite the coalescence of all of these factors, knowledge of the epidemiology of HIV/TB co-infection among those who are incarcerated remains fragmentary.
There is already a large body of empirical evidence showing that history of imprisonment is an important risk factor for TB among those with AIDS [16,17] and those who engage in high-risk behaviours for HIV infection [18]. However, coinfection with HIV and TB among prisoners poses risks not only to the prisoners themselves, but also to prison staff [19,20] and contacts following release.
Risks associated with imprisonment extend beyond the prison walls. Numerous studies have documented TB increases in the broader population that can be attributed, at least in part, to contacts with prison populations [21Á23]. More recently, studies using genetic epidemiology have provided important new insights into the role that prisons play in community outbreaks of TB [24,25].
In part, this is because prisons act as incubators of TB [26]. Research in Eastern Europe and the former Soviet Union has documented a close association between rates of incarceration and of both TB and multidrug resistant TB (MDR-TB), an association exacerbated by a high prevalence of HIV [27]. The aforementioned Russian study [25] found that TB spoligotypes (a means of genotyping Mycobacterium tuberculosis based on patterns of repeat units in DNA) were much less heterogeneous amongst prisoners than the general population, homeless or HIV-infected groups outside prison, suggesting that TB infection is transmitted especially easily amongst this confined group. Research in 1999 documented cases of HIV-positive prisoners co-infected with more than one M. tuberculosis strain [28]. Imprisonment also offers increased potential for TB outbreaks to run rife amongst closely confined inhabitants and prison staff, with numerous TB outbreaks having been recorded amongst correctional facilities in the USA. In 1991 a New York State correctional facility experienced a MDR-TB outbreak from January to November 1991. Eight persons were identified as having MDR-TB, seven of whom were HIV-positive inmates and one a correctional facility worker, immunocompromised as a result of recent chemotherapy [29]. In South Carolina (1999 to 2000) contact tracing of 323 prisoners housed in the same dormitory as a TB index case, found 31 HIVpositive inmates infected with TB, alongside a medical student who examined the source case [30]. Meanwhile, between 1995 and 1996, TB outbreaks occurred amongst HIV housing units in two separate Californian correctional facilities. In prison A, a 500-person HIV housing unit, 14 inmates and the wife of the index case were diagnosed with drug-susceptible TB. In prison B, a 180-person HIV housing unit, an index case resulted in 15 further cases of TB amongst inmates [31].
Given the high prevalence of HIV-positive persons and TB infections in prisons [32Á34] and the documented risk of TB in those infected with HIV, it is of interest to determine how co-infection varies amongst prison populations worldwide. In 2005 it was estimated that around one-third of co-infected persons in Central Asia were found in penitentiaries [35]. However, definitive estimates of the prevalence and incidence of co-infection in prisoners worldwide remains sparse, with evidence suggesting that prisons often fail to implement simple case-finding and screening procedures for infectious disease [36]. For this reason we have undertaken a systematic review of studies of co-infected prisoners, to determine the incidence and/or prevalence of HIV/TB co-infection in prisons, as well as outcomes in this group, measured as treatment success or death. In this review, the terms prisoners and inmates are used interchangeably and refer to both convicted and pretrial (on remand) persons held in prisons, jails, detention and other penal institutions. This resulted in 3791 papers, and 2016 remained after duplicates were removed by EndNote. All further library management was undertaken using EndNote software. Further duplicates were found when reviewing paper titles and abstracts. These were manually removed, leaving 1975 papers. Titles and abstracts from these papers were reviewed by two independent reviewers and were selected for further appraisal according to the following criteria for inclusion: data must be from empirical research; data must relate to either incidence and prevalence of co-infection or outcomes of coinfection; diagnostic methods employed in HIV and TB detection must be specified, including HIV antibodies and TB skin testing (TST) or direct microscopy but not x-ray alone; research must be undertaken in a prison, jail, pretrial detention centre or compulsory drug detention centre; data must be available on the population denominator; and information must be available on the timing of the study. Case definition for TB can include latent TB infection (LTBI) and active TB.

Methods
After papers that failed to meet inclusion criteria were removed, 175 papers remained for full appraisal and the full text was retrieved. After the full text appraisal, 46 papers were selected for inclusion in the review, 41 for incidence and prevalence estimates and nine for outcomes data, with four in both categories. For full details of paper selection, see  Table 1 presents illustrative examples of global research on TB/HIV co-infection. These studies provide further evidence that prisoners infected with HIV are at high risk of developing TB.

Results
However, the magnitude of risk varies between different prisons and countries.
Few studies reported the incidence of TB amongst HIV positive inmates [15,55,56,65], with most simply reporting results from cross-sectional surveys of TB/HIV prevalence or assessing HIV status amongst those that develop TB, as opposed to observing the incidence of TB amongst prisoners of known HIV status. Of the four papers that did provide measures of incidence, three found incidence of TB greater amongst those who were HIV positive [15,56,65], whilst one did not provide a comparison between HIV-negative andpositive prisoners [55]. This lack of reported incidence measures may be reflective of the challenges of calculating these rates in a dynamic, transient prison population, such as issues with accurate calculation of person time at risk, and exclusion of prevalent TB cases, which may or may not be apparent on entry [77].
Estimates of co-infection prevalence, as mentioned previously, tended to favour measuring HIV prevalence amongst TB cases, although some studies measured prevalence of TB in HIV-positive prisoners. Amongst TB cases, prevalence of HIV positivity varied from 0.1% [47] to 73.1% [49]. There were few clear geographical patterns within the results; studies performed in African countries often reported co-infection prevalence over 50% in TB cases [49,60,61,68]. As with measures of incidence, this wide variation in prevalence estimates may be in part due to the settings and characteristics of the prisoners themselves but also in part to methodological differences in calculation of prevalence.
Measures of association between TB and HIV positivity were variable amongst those studies that reported it, with some finding no significant association [40,58,60,65,73,75] whilst others reported highly significant associations between the two infections, with relative risks of co-infection ranging from 2.0 to 10.75 [51,58].
There was a high risk of bias in many of the studies included (Table A1), most often because studies were performed in those with TB or HIV, as opposed to the entire prison population. Few studies explicitly set out to measure coinfection, often including HIV positivity in a list of characteristics common to patients with TB. Even in studies that did not specify whether an infected group was targeted, most often a subset of the prison population was used, such as those with respiratory symptoms. Amongst the studies case definitions varied. In total, two examined LTBI diagnoses only, six at both active TB and LTBI and 35 at active TB only. The bias inherent in many of these studies means it is hard to draw any definitive conclusions as to the size of the association or increased risk between TB and HIV in prisoners. For this reason a formal meta-analysis was not undertaken on the retrieved results,  Table A2.
Only six of the studies retrieved [2,50,53,62,69,70] provided any comparison of a prisoner's risk of co-infection to that of the general population. All of these studies found the risk or prevalence of co-infection to be greater among prison residents than that of the general community. However, given the small numbers of studies reporting these comparative data, it is hard to ascertain a definitive value of increased risk. Therefore, we must simply conclude that the risk of developing TB amongst HIV-positive inmates is high and appears higher than for those outside of prison.
Few studies retrieved were recorded as performed in jails (used to hold individuals awaiting trial or sentencing) [39,40,47,73,74], meaning most of the conclusions drawn were based on those resident in prisons, who are likely to be less transient than jail or remand populations. One jail study found HIV positivity predictive of TB (odds ratio (OR) 3.07; p B0.01) [73], whilst another found very few cases of coinfection (n02/2103) [47], making it hard to draw conclusions on co-infection risk in jail. Those in jails may potentially be more likely to represent people who are incarcerated with TB, as short length of stay may preclude TB development; however this association is unproven. In addition, short-stay prisoners may represent a population with high rates of recidivism [78], therefore returning as a prevalent case if TB treatment completion was complicated by the short stay nature of jails [73].
While noting the caveats concerning TST, research suggests that prisoners are no more likely to be infected with TB [79] but are more likely to develop active disease [44,56,60,73] than people residing outside of prison. Indeed, one study found no association between HIV positivity in prisoners and LTBI [40]. One of the few studies among female prisoners, from Brazil, found that while the TB infection prevalence was similar among women with or without HIV infection, with a purified protein derivative test conversion rate of approximately 30% in both, the incidence of TB was 9.9 per 100 person-years among those who were HIV positive compared with 0.7 per 100 person-years in those who were negative [80]. This same study found that the development of TB was most likely within the first 12 months of incarceration, as was also the case in a Texan prison [15].
Some studies looking at activation of disease have found an association between TB and CD4 counts below 200 cells/mm 3 [81,82], whilst others suggest a significantly increased risk below 350 cells/mm 3 , with risk rapidly increasing as CD4 count decreases [16]. Therefore, differing rates of active disease co-infection between settings may in part be explained by the stage of AIDS in HIV-positive prisoners. Receipt of antiretroviral   therapy (ART) by prisoners could also impact on the observed heterogeneity in co-infection rates, with those receiving therapy less likely to succumb to opportunistic infections.

Prevention and treatment
Low rates of TB treatment completion amongst the prison population in general are well documented, and the reasons are multifactorial. Many low-and middle-income countries [83] have severely limited prison health facilities, and prisoners with symptoms suggestive of TB may go undiagnosed. In addition, access to appropriate chemoprophylaxis in prisons can be problematic due to reasons such as affordability, circulation of suboptimal drugs and the low priority given to prison healthcare by policymakers [84,85]. In addition, the movement of prisoners, both within and out of the prison system, exacerbates the rate of treatment default.
Evidence on the quality and effectiveness of treatment in those with co-infection is fragmentary. However, the available evidence indicates that, even in general populations, there are substantial geographical variations. For example, patients in Eastern Europe are significantly less likely to receive adequate treatment for both conditions than those in other parts of Europe or Latin America [86]. For prisoners with coinfection, treatment completion is complicated by both the nature of this transient population and the complex drug regimes required for co-infected patients, sometimes resulting in adverse drug-drug interactions [15,87]. HIV positivity is not generally associated with TB treatment adherence [88]. However, one US study of prisoners with latent TB found that they were less likely to complete treatment if they were HIV positive, with defaults attributed to movements within the correctional system, refusal of treatment and adverse drug effects [74]. The results for treatment outcomes have been summarized in Table 2.
There have been a few studies of TB outbreaks among HIV-positive prisoners. One review concluded that, although consideration must be given to drug-drug interactions involving ART and anti-TB medication, shorter drug regimes remain preferable, due to the transitory nature of the prison population [91].
The available evidence provides several other important findings. First, in HIV-positive prisoners found to have evidence of infection by TST, latent TB infection treatment can reduce the incidence of TB [65]. Second, where effective treatment is provided, outcomes of TB infection among those infected with HIV, at least where immunosuppression is mild or ART is also provided, are as good as in HIV-negative prisoners [92], consistent with other evidence that treatment for drugsensitive TB is effective in the early stages of immunodeficiency [93]. Third, in the presence of sensitivity testing and continuity of treatment, those with HIV infection are at no greater risk of developing MDR-TB, beyond their risk for TB in general [94,95], and, in one Russian study, prisoners with coinfection were significantly less likely to develop resistance, although the explanation was unclear [96].
Isolation of infectious prisoners to prevent further spread is not consistently documented. Moreover, many of those isolated may not be receiving treatment, conditions are frequently appalling and these facilities may be used to hold  uninfected juveniles thought to be at risk from older, violent prisoners elsewhere in the prison [97]. As mentioned previously, the segregation of prisoners into HIV-infected units provides an ideal opportunity for TB to spread rapidly amongst this vulnerable population [31]. Even in high-income countries there may be problems. For example, in US prisons, Hispanics are systematically less likely to be tested for HIV and TB than white prisoners [98].

Discussion
This review provides evidence that individuals infected with HIV who reside in a prison setting are at high risk of developing active TB, with this risk varying vastly by factors such as prison location and prison population density. This is consistent with literature reporting higher rates of TB and HIV individually, amongst prison populations worldwide, as noted in a recent paper from which this systematic review arose [99]. It has summarized an extensive body of literature on its determinants and the responses to it. It is clear that, if services appropriate to the scale of the problem are to be developed, further research is needed to accurately quantify the burden of co-infection in prison populations today and the differences between and within countries, as well as the outcomes prisoners can expect during and after treatment. In this review most of the studies citing numbers of co-infected prisoners did not have this as their primary objective, more often recording HIV positivity as one of numerous variables in patients with TB, so that these estimates of prevalence/ incidence are merely a by-product of different research objectives. As noted by Rieder et al., new prison entrants may be more likely to participate in screening processes, meaning that prevalence estimates may be higher if made at a time when the proportion of new entrants was high. Prevalence can also vary according to the case definition used for inclusion [77]. With most studies reporting only crosssectional results, it remains hard to disentangle whether imprisonment increases the risk of developing TB over the course of imprisonment, whether those with TB are more likely to be imprisoned and found to be HIV positive during prisonbased screening, or whether in fact high prevalence estimates result from a combination of both these factors[A8]. Overall, it is apparent that there is a need for accurate quantification of the burden of co-infection in high-risk prison settings. Consideration must be given to the funding and administration in resource-poor settings. In England, which maintains a comprehensive infectious disease surveillance programme based in Public Health England, the national public health agency, there are problems in ascertaining rates of coinfection. Although [A9]TB is a notifiable disease and authorities are required to report it upon detection, the same is not true of HIV status, despite the associated benefits for public health, such as contact tracing and treatment [100]. Without robust surveillance systems to confirm numbers of co-infected prisoners, it will be nearly impossible to monitor and appraise the effectiveness of interventions to reduce the rates of coinfection. Studies of treatment outcomes in those co-infected were even more limited, with results conflicting as to the impact of HIV positivity on treatment outcomes. Prisoners were often lost to follow-up post-release, thereby reinforcing the need for a robust continuum of care to ensure treatment completion as well as monitor and evaluate treatment effectiveness. Further research is needed to determine optimum treatment regimes, delivery models and expected outcomes in this population. This reflects the relative paucity of studies on TB in prisons [101], few of which have stratified results by HIV status [102]. It also reflects the scarcity of studies of HIV/TB co-infection in general. A recent international systematic review of coinfection studies, excluding China, found 47 studies, and just one on prisoners [103]. However, this shed light on several important issues. First, those studies combining chest radiography with serology produced higher estimates, consistent with what is known about the lower sensitivity of sputum culture and the problems of interpreting TST in individuals who are immunosuppressed and therefore less likely to react [52,64,104,105]. Second, it confirmed an earlier study showing that co-infection was more likely in those communities where the prevalence of HIV was highest [106]. However, the use of different methodology precluded comparison.
Heterogeneity observed in study results is likely due in part to study design but also to the differences between prisons themselves. Prisons are diverse, both in terms of the accommodation style and the numbers and demographic characteristics of their inhabitants. Prisons may be mixed or single sex, with some previously segregating HIV-positive inmates into separate housing facilities, although this is now rare [31]. Rooms range from well-equipped single cells to cold, dark, damp concrete boxes accommodating large numbers of prisoners, with one study from Ethiopia reporting an average number of 333 prisoners per cell [49]. As one UN inspector visiting a predetention centre in Russia noted, it would need the literary skills of Dante or the artistic skills of Hieronymus Bosch, to describe fully the horrors with which she was confronted [107]. Recent data showed that, in 40% of 193 countries, prison occupancy was over 100% of stated prison occupancy capacity; more than half of overcrowded prisons had occupancy rates between 150 and 200% of capacity and 15 were above 200% of capacity. Some approached 400% [108,109]. A study that modelled the transmission dynamics of outbreaks of TB originating in institutions such as prisons and mines concluded that the most important factors were the size of the population within the institutions and movement between them and the community [110]. Therefore, it is almost inevitable that the scale of HIV/TB co-infection will vary between countries and, within them, between prison settings, regardless of study methodology and associated bias.
A recurring theme in the quest to improve treatment outcomes was loss to follow-up of patients released from prisons [61,74,89]. This topic has been the subject of numerous studies, particularly in regard to continuing HIV therapy [111,112]. A recent systematic review found improvements along the pathway of care for HIV during incarceration, often beyond what was achieved in the general population, but problems rapidly arose as soon as subjects were released from jail [111]. Recommendations to address this problem are complex [112,113], acknowledging the competing demands in prisoners' lives once they are released back into the community. As re-incarceration has been known to reach up to 76% within five years of release [113], inadequate treatment of LTBI [74], which later manifests as active TB in patients with HIV, may result in the reintroduction of TB both into the community and back into the prison system. In addition, loss to follow-up and therefore non-compliance with the full TB treatment regime contributes to the development of drug-resistant strains of TB [114,115]. Particular attention must therefore be paid to developing models of treatment continuum that are acceptable to prisoners post-release.
This systematic review is subject to certain limitations. Although there was no restriction on publication date for paper retrieval, we did not include data on co-infection incidence and/or prevalence from papers published prior to 1999, as their relevance to the current situation is dubious given the changing epidemiology of TB and HIV since this period. Prevalence or numbers of co-infected prisoners documented only in conference abstracts were not presented, as it is unknown whether this information was subjected to critical peer review, and many of these abstracts failed to document clearly the methods of ascertaining infection status. Several high quality studies documented rates of HIV, TB and other infections amongst a specific prison population but failed to give any data on numbers with co-infection. It is plausible that direct contact with these authors may have yielded further information. However, there may be doubt about whether the results provided would be accurate if co-infection was not specifically recorded and reported during the original study. Many of the papers presented did not set out purposefully to record levels of co-infection, but instead measured numbers of HIV-positive people amongst TB cases, or vice versa, meaning prevalence estimates within the entire prison population could not be calculated. International comparisons must also be undertaken with care because of the very different nature of prisons, the analyses undertaken in these studies and the methods used to identify TB infection. Moreover, rates of co-infection will be influenced by what is happening in the general population, which also varies greatly.

Conclusions
If the tide of HIV/TB co-infection in prisons is to be stemmed, there is a clear need to develop and implement specific, coordinated policies for active case-finding and treatment of co-infected prisoners, for both those in prison settings and those transitioning back into the community. This will bring benefits to the individuals concerned but also reduce the reservoir of disease, both within the prison population and the community at large. The difference in the care provided to prisoners and the general population, as well as the priority it achieves on the policy agenda, has long been recognized.
Authors' contributions CE defined the search terms and performed the initial literature search. CE and EK reviewed and appraised the titles returned from the search. MM conceived the project. CE and MM wrote the manuscript. KD commented on the final draft. All authors have read and approved the final version.