Transmission Dynamics in Tuberculosis Patients With Human Immunodeficiency Virus: A Systematic Review and Meta-analysis of 32 Observational Studies

Abstract

Background

There are large knowledge gaps on the transmission dynamics of Mycobacterium tuberculosis in settings where both tuberculosis and human immunodeficiency virus (HIV) are endemic. We aimed to assess the infectiousness of tuberculosis patients coinfected with HIV.

Methods

We systematically searched for studies of contacts of both HIV-positive and HIV-negative tuberculosis index cases. Our primary outcome was Mycobacterium tuberculosis infection in contacts. Data on sputum smear and lung cavitation status of index cases were extracted from each study to assess effect modification. Secondary outcomes included prevalent tuberculosis and HIV in contacts of HIV-positive and HIV-negative index cases.

Results

Of 5255 original citations identified, 32 studies met inclusion criteria, including 25 studies investigating M. tuberculosis infection (N_participants = 36 893), 13 on tuberculosis (N_participants = 18 853), and 12 on HIV positivity (N_participants = 18 424). Risk of M. tuberculosis infection was lower in contacts of HIV-positive index cases (odds ratio [OR], 0.67, 95% confidence interval [CI], .58–.77) but was heterogeneous (I² = 75.1%). Two factors modified this relationship: the lung cavitary status of the index case and immunosuppression (measured through CD4 counts or HIV or acquired immunodeficiency syndrome diagnoses) among index people living with HIV. Rates of HIV were consistently higher in contacts of coinfected index cases (OR, 4.9; 95% CI, 3.0–8.0). This was modified by whether the study was in sub-Saharan Africa (OR, 2.8; 95% CI, 1.6–4.9) or in another global region (OR, 9.8; 95% CI, 5.9–16.3).

Conclusions

Tuberculosis patients coinfected with HIV are less infectious than HIV-uninfected cases when they have severe immunosuppression or paucibacillary disease. Contacts of coinfected index cases are almost 5 times more likely to also have HIV.

Tuberculosis patients coinfected with human immunodeficiency virus (HIV) are less infectious than HIV-uninfected people when they have severe immunosuppression or paucibacillary disease. Contacts of coinfected index cases are almost 5 times more likely to also have HIV.

(See the Editorial Commentary by Kendall on pages e3456–8.)

The coepidemic of tuberculosis and human immunodeficiency virus (HIV) began more than 3 decades ago and has led to millions of deaths, the vast majority of which occur in impoverished settings with few health resources [1–3]. Individuals living with HIV are at high risk for both primary progressive disease and reactivation disease [1, 4]; moreover, people with HIV who develop tuberculosis experience accelerated HIV disease progression and severity [5, 6]. Because of this, morbidity and mortality of both diseases are substantial. The heavy toll of tuberculosis and HIV continues because of ongoing Mycobacterium tuberculosis transmission, from both HIV-seropositive and HIV-seronegative tuberculosis cases. But our understanding about transmission in the context of HIV infection is incomplete.

There are large gaps of knowledge regarding how transmission of M. tuberculosis occurs in settings where both HIV and tuberculosis are endemic [1]. The infectiousness of people with tuberculosis coinfected with HIV is unclear. A previous meta-analysis on the infectiousness of HIV-positive people with tuberculosis concluded that coinfected patients were as infectious as people who are HIV-negative and have tuberculosis after pooling 4 studies with tuberculin skin test results from HIV-negative contacts [7]. This meta-analysis displayed substantial between-study heterogeneity and was unable to explain this variability because of few available studies [7–9]. A reassessment of the infectiousness of people with tuberculosis and HIV is needed to improve our understanding of M. tuberculosis transmission dynamics where both HIV and tuberculosis are burdensome and to inform health policy decision-making.

To fill this knowledge gap, we conducted a systematic review and meta-analysis of studies evaluating M. tuberculosis infection rates in contacts of HIV-positive and HIV-negative index cases. We attempted to explain heterogeneity by stratification when possible. We also evaluated the risk of coprevalent tuberculosis and HIV-positivity in contacts exposed to HIV-positive and HIV-negative index cases.

METHODS

Search Strategy and Study Selection

This review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement (Supplementary Appendix). The study protocol is registered with the International Prospective Register of Systematic Reviews (protocol #CRD42019120273). Briefly, our search strategy aimed to identify all studies that assessed the number of M. tuberculosis infections, tuberculosis, or HIV found when contact investigation was done and stratified by the HIV status of the index case. We searched journal articles of any study design in Medline, Web of Science, Biosis, and Embase electronic databases. The search was conducted with the help of a librarian database consultant and was conducted until October 2014. An identical search was then conducted for the time period of October 2014 through August 2018. We used the following search terms, adapted for each database when appropriate: “Mycobacterium tuberculosis,” tuberculosis, TB, “human immunodeficiency virus,” HIV, tuberculin, transmission, contact*. We did not restrict articles by publication date or language. Bibliographies of reviews, both systematic and descriptive, were also searched and evaluated for eligibility [1, 7, 9–15].

After the search and exclusion of duplicate articles, 2 authors (L.M. and H.W.) independently screened articles by title, abstract, and full text. If reviewers disagreed about inclusion, a consensus of authors determined manuscript eligibility. Corresponding authors were contacted for additional data if a study met eligibility criteria but exact counts were unavailable. From each article, information was collected on the study, index cases, and from contacts. We were unable to extract study-level data on outcomes from 1 eligible study [16] because only adjusted relative risks were presented in the manuscript. We therefore used an online database of this cohort from another published study [17] to re-create measures of association for the differing outcomes.

Study Definitions

Exposure

Exposure was defined as close contact, either in the same household or through the community, to a tuberculosis case. This was defined by each set of authors. We stratified by level of tuberculosis exposure (household vs community). We did not include studies with exposed healthcare workers as a previous meta-analysis [7] did because these individuals have highly distinct populations compared with other types of contacts and are at high-risk for repeated exposures and reinfections. Index cases were considered source cases and were eligible if diagnosis was confirmed either bacteriologically (sputum smear or culture positive) or radiographically.

Outcome

The primary outcome was M. tuberculosis infection in contacts of HIV-positive and HIV-negative index cases. Studies using the tuberculin skin test or an interferon gamma assay to diagnose M. tuberculosis infection were included. Various tuberculin induration cutoffs were used to define a positive tuberculin skin test. We used each study’s specified definition for a positive tuberculin skin test. A cutoff of 5- and 10-mm induration reactions to define a positive tuberculin skin test were used in 9 and 10 studies. Three studies used an induration of 10 mm or greater for HIV-negative contacts and 5 mm or greater for HIV-positive contacts. One study used QuantiFERON Gold In-Tube tests [18] and a cutoff of ≥0.35 IU/mL was used to define a positive test.

Secondary outcomes included tuberculosis and HIV in contacts of HIV-positive and HIV-negative index cases. Tuberculosis was defined as the presence of tuberculosis regardless of tuberculin skin test results. Most studies used microbiological confirmation of diseased contacts, whereas some studies used a combination of symptom screening, radiographical examinations, and tuberculin skin test results. Two studies did not specify the diagnostic method used [19, 20]. The discretion of each study’s authors to define tuberculosis was used.

HIV in contacts was defined as a positive laboratory test for HIV. The study was ineligible if HIV status was obtained through participant self-report alone. Several HIV tests (enzyme-linked immunosorbent assay [ELISA], Western blot, Determine, and Unigold) were used and we used each study’s definition of a positive HIV test.

Assessment of Study Quality

The quality of each eligible study was assessed. We used a modified version of the Newcastle-Ottawa scale for assessment of observational studies. Studies were evaluated based on 3 characteristics: adequate selection of participants (4 points), comparability of studies based on the design and analysis (1 point), and adequate ascertainment of study outcomes (3 points). This scale awards a maximum of 8 points to each study. We defined studies of high quality as those that scored ≥66.6%, moderate for 33.3%–66.6%, and low for those <33.3%.

Statistical Analysis

We estimated the odds ratio (OR) for M. tuberculosis infection and disease in contacts of HIV-positive and HIV-negative tuberculosis people for each study. We then pooled these ORs using a random effects model with DerSimonian and Laird weights. We used a random effects model, equalizing the weight of the studies to the pooled estimate, a priori because we included only observational studies [21]. We stratified the analysis by prespecified characteristics of eligible studies when available. These included contact characteristics (age, HIV status, bacillus Calmette–Guérin [BCG] vaccination status), index case characteristics (lung cavitary disease, sputum smear status, extrapulmonary disease, HIV immunosuppression, cough duration), and study characteristics (year of implementation, region, closeness of tuberculosis contact [ie, household or community]).

95% confidence intervals (CI) were used to assess statistical significance in all models. We used the I² statistic to evaluate heterogeneity between studies [22, 23]. The I² statistic represents the proportion of variability in included studies resulting from heterogeneity alone rather than random error. A threshold of I² > 50% was used as indicating statistically significant heterogeneity. We assessed publication bias through inspecting funnel plot symmetry and by using the Harbord test for publication bias [24, 25]. All statistical analyses were performed with Stata, version 14.0 (StataCorp LP, College Station, Texas, USA) and R statistical software (R Foundation for Statistical Computing).

RESULTS

Study Selection and Characteristics

From our database searches, we found a total of 5255 original titles, of which 32 studies [2, 16–19, 26–52] met eligibility requirements and were included in the meta-analysis (Figure 1). Of these studies, 25 investigated prevalent M. tuberculosis infection [2, 16–18, 26–46], 13 on tuberculosis [2, 17–19, 26, 30, 31, 33, 34, 39, 47, 48, 52], and 12 on HIV infection among contacts [2, 17, 26, 27, 33, 37, 46, 49–52] (Table 1). Of these 3 outcome measures, 19 studies evaluated only 1 outcome, 9 studies evaluated 2, and 4 studies evaluated 3. Two studies [30, 49] had the same study population; the most recent publication [25] reported prevalent M. tuberculosis infection and disease, whereas the oldest [45] reported these 2 outcomes plus HIV infection in contacts. We took data from Klausner and colleagues [30] on M. tuberculosis infection and disease because data were most recent; data on HIV infection of contacts were taken from Baende and colleagues [49]. Among the 32 studies, 22 recruited only household contacts of tuberculosis cases, 9 studies had both household and community contacts, and 1 study did not specify the type of contact [19].

Figure 1.

Flow chart detailing literature search results for studies on the association between the HIV status of tuberculosis cases and clinical tuberculosis-related outcomes in case-contacts. *Final numbers of studies may not add to previous totals and exclusions because multiple studies investigated more than 1 outcome variable. Abbreviation: HIV, human immunodeficiency virus.

Table 1.

Summary of the 32 Observational Studies Included in the Meta-analysis on the Association Between HIV Status of Tuberculosis Cases and Clinical Tuberculosis Outcomes in their Case Contacts

						Contacts of HIV-positive TB Cases		Contacts of HIV-negative TB Cases
First Author, Year of Publication	Yeara	Contactb	Country	Diagnosis	N Index	N Contacts	% Yield Outcome	N Contacts	% Yield Outcome
Mycobacterium tuberculosis infection
Martinez, 2016	1995–2006	HHC	Uganda	TST, ≥10 mm	499	878	65.7	974	73.6
Alseda, 2003	1991–1997	All	Spain	TST, ≥5 mmc	437	199	31.7	1962	36.5
Aibana, 2016d	2009–2012	HHC	Peru	TST, ≥10 mme	4500	405	42.5	11 590	43.4
Kenyon, 2002	1997	HHC	Botswana	TST, ≥10 mm	51	174	13.2	29	17.2
Kifai, 2009	2007	HHC	Tanzania	TST, ≥10 mme	57	125	61.7	112	62.5
Suggaravetsiri, 2003	2000–2002	HHC	Thailand	TST, ≥10 mm	499	487	46.2	705	62.1
Lienhardt, 2002	1999–2001	HHC	Gambia	TST, ≥10 mm	315	83	71.1	1397	78.0
Gustafson, 2008	1999–2000	HHC	Guinea B.	TST, ≥10 mm	220	285	39.2	738	42.5
Naing, 2005	2000–2004	HHC	Malaysia	TST, ≥10 mm	215	84	33.3	320	51.9
Cailleaux-Cezar, 2009	2000–2002	All	Brazil	TST, ≥10 mmc	276	110	51.4	480	46.5
Kasambira, 2011	2006–2009	HHC	SA	TST, ≥5 mm	167	233	26.2	7	57.1
Godoy, 2013	2005–2006	All	Spain	TST, ≥5 mm	1079	198	33.3	5173	28.4
Biraro, 2014	2011–2012	HHC	Uganda	QFT, 0.35 IU/mL	101	56	51.8	207	67.6
Fatima, 2004	1997–1999	HHC	Brazil	TST, ≥10 mm	297	177	38.4	22	45.5
Reichler, 2003	1996	All	USA	TST, ≥5 mm	349	29	37.9	146	60.3
Pitchenik, 1987	1985–1986	HHC	USA	TST, NS	71	54	35.2	108	43.5
Cauthen, 1996	1985–1989	All	USA	TST, ≥5 mm	956	1095	31.2	2158	43.4
Espinal, 2000	1994–1995	HHC	Brazil	TST, ≥5 mm	174	252	60.7	551	75.9
Klausner, 1993	1989–1990	HHC	Zaire	TST, ≥10 mme	169	521	60.1	692	63.0
Elliott, 1993	1989	HHC	Zambia	TST, ≥5 mm	71	207	52.2	141	70.9
Nunn, 1994	1989–1990	HHC	Kenya	TST, ≥5 mm	82	80	61.3	223	58.3
Garcia Ordonez, 1999	1995–1997	HHC	Spain	TST, ≥5 mmc	249	152	20.4	516	48.8
Manoff, 1988	1988	All	USA	TST, NS	491	392	21.2	1703	30.4
Carvalho, 2001	1995–1997	All	Brazil	TST, ≥10 mm	86	104	26.9	256	35.2
Khan, 2017	2013–2015	HHC	Malawi	TST, ≥10 mm	150	132	22.0	170	44.1
Tuberculosis
Martinez, 2016	1995–2006	HHC	Uganda	Micro	499	878	4.0	974	4.3
Aibana, 2016d	2009–2012	HHC	Peru	Micro, Symp	4500	424	0.9	12 094	1.7
Suggaravetsiri, 2003	2000–2002	HHC	Thailand	Micro, CX	499	490	2.9	710	4.4
Cailleaux-Cezar, 2009	2000–2002	All	Brazil	Micro, CX	276	110	4.5	480	2.5
Rodrigo, 1997	1990–1993	All	Spain	Micro	1079	163	4.9	916	2.4
Pitchenik, 1987	1985–1986	HHC	USA	NS	71	54	1.9	108	0.9
Standaert, 1989	1985–1986	HHC	Burundi	Micro	NS	48	12.5	28	0.0
Klausner, 1993	1989–1990	HHC	Zaire	Micro, Sympf	174	521	3.1	692	4.0
Elliott, 1993	1989	HHC	Zambia	Micro, Symp	71	207	3.9	141	2.8
Nunn, 1994	1989–1990	HHC	Kenya	Micro, Symp	82	87	6.9	248	4.8
Garcia Ordonez, 1999	1995–1997	HHC	Spain	Micro, CX	249	152	1.3	516	5.0
Topley, 1996	1993–1994	HHC	Malawi	Sym, CX, TST	206	105	31.4	37	35.1
Cayla, 1993	1990–1991	NS	Spain	NS	136	225	8.0	216	3.2
HIV infection
Martinez, 2016	1995–2006	HHC	Uganda	ELISA	499	915	16.8	1018	4.6
Suggaravetsiri, 2003	2000–2002	HHC	Thailand	ELISA	499	376	13.8	514	2.5
Aibana, 2016d	2009–2016	HHC	Peru	NS	4500	419	2.4	11 959	0.3
Nunn, 1994	1989–1990	HHC	Kenya	NS	82	101	13.9	250	0.8
Elliott, 1993	1989	HHC	Zambia	ELISA	71	133	13.5	69	7.2
Carvalho, 2001	1995–1997	All	Brazil	ELISA and WB	86	75	10.7	179	1.7
Kifai, 2009	2007	HHC	Tanzania	ELISA	57	115	8.7	103	8.7
Baende, 1990	1989	HHC	Zaire	ELISA and WB	100	323	5.9	410	2.4
Hirsch-Moverman, 2015	2002–2006	All	USA, Can.	NS	651	184	23.4	806	2.7
Reichler, 2003	1996	All	USA	NS	29	30	53.0	147	2.1
Standaert, 1989	1985–1986	HHC	Burundi	ELISA	NS	48	8.3	28	7.1
Mutsvangwa, 2010	2002–2004	HHC	Zimbabwe	Det., Unigold	129	172	28.5	50	12.0

Abbreviations: Can., Canada; CC, community contact; CX, chest radiographical exam; Det., determine; DRC, Democratic Republic of Congo; Guinea B., Guinea–Bissau; HHC, household contact; HIV, human immunodeficiency virus; Micro, microbiological testing; NS, not specified; QFT, QuantiFERON Gold In-Tube Test; SA, South Africa; Sym, symptom screening; TST, tuberculin skin test; WB, Western blot.

^a Year of implementation of the study. If dates are not given for the study implementation the study publication year is given.

^b All refers to studies that collected data on both household contacts and community contacts and grouped their results together. Only 1 study presented stratified results based on differing types of contacts.

^c In Alseda and colleagues (2010), for all contacts unvaccinated with the BCG the definition of a positive tuberculin skin test was ≥5 mm. For all BCG-vaccinated contacts, a positive test was defined as ≥15 mm. In Cailleaux-Cezar and colleagues (2009), for all contacts unvaccinated with the BCG, the definition of a positive tuberculin skin test was ≥10 mm. For all BCG-vaccinated contacts, a positive test was defined as ≥15 mm.

^d Aibana and colleagues (2016) is presented here rather than Huang and colleagues (2014), which is part of the same cohort. We present here the Aibana and colleagues study because study-level data on outcomes from contacts of HIV-positive and HIV-negative contacts were not extractable from the Huang and colleagues study. We used an online database of this Peruvian cohort and analyzed the specific outcomes (Mycobacterium tuberculosis infection, active tuberculosis, HIV infection) among contacts of HIV-positive and HIV-negative index cases for the subsequent analysis.

^e For all HIV-negative participants the definition of a positive tuberculin skin test was ≥10 mm. For all HIV-positive contacts, a positive test was defined as ≥5 mm.

^f A portion of the contacts diagnosed with tuberculosis were confirmed with microbiological testing.

Risk of M. tuberculosis Infection in Contacts

From the 25 studies investigating M. tuberculosis infection, the total number of household contacts of HIV-positive and HIV-negative index cases from all studies was 6513 (median, 177 [interquartile range [IQR], 104–285]) and 30 380 (median, 480 [IQR, 146–974]), respectively. One study used QuantiFERON-TB Gold in Tube [16]. Two studies did not specify the criteria for a positive tuberculin skin test but were included in the analysis because they stipulated whether subjects had either positive or negative skin test.

Risk of M. tuberculosis infection was lower in contacts of HIV-positive index cases (pooled OR, 0.67 [95% CI, .58–.77]; Table 2) but was heterogeneous (I² = 75.1%). To evaluate the observed heterogeneity, we stratified from studies with available information (Table 2). When the tuberculosis index case had lung cavitary disease, the risk of M. tuberculosis infection was elevated in contacts of HIV-positive index cases (OR, 1.37 [95% CI, 1.05–1.78]; I² = 0%) compared with HIV-negative index cases, whereas when the index case did not have cavitary lung lesions and contacts of HIV-positive cases had less M. tuberculosis infection (OR, 0.69 [95% CI, .47–1.04]; I² = 83.9%). When the tuberculosis index case was stratified by sputum smear status, the relationship between risk of M. tuberculosis infection and the HIV status of the index case diverged (sputum smear positive: OR, 0.69 (95% CI, .59–.80), I² = 63.5%; Sputum smear negative: OR, 0.41 [95% CI, .13–1.28], I² = 93.5%). Three studies (Figure 2) also showed that people with HIV and either a low CD4 count or with acquired immunodeficiency syndrome (AIDS) status modified this relationship.

Table 2.

Pooled Random Effects Logistic Regression for the Influence of the Tuberculosis Case’s HIV Status on Mycobacterium tuberculosis Infection in Contacts, Stratified by Secondary Risk Factors

Characteristic	No. Studies	Pooled OR (95% CI), P Value for Effecta	I 2
Crude
All studies	25	0.67 (.58–.77), <.001	75.1
Age of contact, y
0–4	3	0.69 (.51–.94), .018	0.0
5–14	10	0.61 (.45–.82), .001	74.9
≥15	4	0.74 (.59–.92), .008	0.0
BCG vaccination status of contact
Vaccinated	3	0.87 (.58–1.31), .504	82.0
Unvaccinated	3	0.78 (.58–1.06), .116	0.0
HIV-negative contacts
HIV-negative contacts onlyb	6	0.60 (.39–.93), <.001	0.86.1
Form of contact with tuberculosis case
Household	19	0.63 (.53–.75), <.001	67.5
Community	1	0.46 (.30–.71), <.001	–
Both	5	0.86 (.61–1.22), .040	85.0
Definition of Mycobacterium tuberculosis infection c
≥10-mm induration for all contacts	9	0.58 (.42–.79), .001	55.0
≥5-mm induration for all contacts	10	0.71 (.58–.86), <.001	85.8
≥10 mm for HIV-neg.; ≥5 mm for HIV-pos.	3	0.93 (.81–1.08), .342	0
Region
Asia	2	0.51 (.41–.63), <.001	0.0
Africa	11	0.69 (.57–.85), <.001	57.0
Europe	3	0.66 (.29–1.48), .309	94.0
Americas	9	0.69 (.55–086), .001	72.9
Year d
1990 and before	2	0.63 (.49–.80), <.001	0
Post-1990	23	0.67 (.57–.78), <.001	77.0
2000 and before	8	0.59 (.46–.75), <.001	78.8
Post-2000	17	0.72 (.60–.86), <.001	69.6
Sputum smear status of tuberculosis case
Positive	13	0.69 (.59–.80), <.001	63.5
Negative	4	0.41 (.13–1.28), .124	93.5
Both	11	0.75 (.58–.97), .031	73.9
Cavitary lung disease of tuberculosis case
Lung cavitary disease	3	1.37 (1.05–1.78), <.018	0
Noncavitary disease	3	0.69 (.47–1.04), .074	83.9
Both	19	0.69 (.58–.83), <.001	76.1
Extrapulmonary tuberculosis
Pulmonary cases only	16	0.68 (.58–.79), <.001	64.2
Pulmonary and extrapulmonary casese	6	0.62 (.42–.91), .015	87.5
Not specified	1	0.73 (.25–2.11), .562	-

^a In all analyses, the reference category for the measure of association are contacts of HIV-negative tuberculosis cases.

^b Cauthen (2004) assumed that children under 14 in their study population were likely not HIV-infected contacts; however, this was not confirmed with HIV testing.

^c Two studies did not specify their definition for a positive tuberculin skin test.

^d Year of publication.

^e No studies stratified both HIV status and the presence of extrapulmonary tuberculosis in the index case.

Figure 2.

Individual studies demonstrating evidence for effect modification on Mycobacterium tuberculosis infection in case contacts by clinical characteristics of the HIV-tuberculosis coinfected index case. We present here results from individual studies that show modification of the infectiousness of HIV-positive and HIV-negative index cases. These studies are stratified by the severity of tuberculosis in the index case (as measured through smear positivity and cavitary status) and the severity of the HIV status of the index case (as measured through CD4 count and AIDS status). Huang and colleagues (2014) is presented here rather than Aibana and colleagues (2016) which is part of the same cohort. We present here Huang and colleagues’ study because they include CD4 count. The study by Aibana and colleagues is presented and analyzed in the rest of the paper because study-level data on outcomes from contacts of HIV-positive and HIV-negative contacts were not extractable from the Huang and colleagues study. Abbreviations: HIV, human immunodeficiency virus; RR, relative risk; TST, tuberculin skin test. * All index cases have tuberculosis. This column stratifies tuberculosis index cases by their HIV status and other secondary modifying variables related to the severity of HIV or tuberculosis. • Adjusted for age, sex, smoking status, alcohol intake, nutritional status, number of BCG scars, household smoke exposure, relation to the tuberculosis case from household contacts; age, sex, cavitary lung disease, smear status, and treatment delay from tuberculosis cases. ^†Adjusted for age, education level, and alcohol status of the household contact; sputum smear and cavitary status of the tuberculosis case; and the number of individuals in the household. ^¥Adjusted for female sex and sputum smear grade of the tuberculosis case, and household clustering. ^¶Smear negative, culture negative index cases were grouped together with smear negative, culture positive cases in this group. The prevalence of positive skin tests was roughly equivalent in both their sets of contacts. ^‡New-onset AIDS was diagnosed before contact testing while previously diagnosed AIDS was diagnosed before contact testing.

When we pooled rates from other subgroups, including different age groups, BCG vaccination, household versus community contacts, region, and study year of implementation, we found little mediation of the relationship between M. tuberculosis infection and the HIV status of the index case. For example, among contacts 0–4, 5–14, and ≥15 years old, the OR of M. tuberculosis infection among contacts of index cases living and not living with HIV did not appreciably differ and were 0.69 (95% CI, .51–.94), 0.61 (95% CI, .45–.82), and 0.74 (95% CI, .59–.92), respectively. Similarly, when stratifying by the type of contact, there was little difference in the relationship between the HIV status of the index case and M. tuberculosis infection among contacts. Among studies with household and community contacts, the odds of M. tuberculosis infection among contacts of index cases living and not living with HIV was 0.63 (95% CI, .53–.75), 0.46 (95% CI, .30–.71), and 0.86 (95% CI, .61–1.22), respectively.

Two studies investigated effect modification by the index case’s cough duration [2, 17]; both found that cough duration did not influence the relationship between M. tuberculosis infection and the differing HIV status of index cases.

Risk of Active Tuberculosis in Contacts

Of 13 studies on tuberculosis, the total number of contacts of HIV-positive and HIV-negative index cases from all studies was 2623 (median, 152 [IQR, 87–225]) and 16 230 (median, 248 [IQR, 108–692]), respectively. Nine studies confirmed tuberculosis with a microbiological examination (sputum smear or culture); 1 study used a combination of symptom screening, chest radiographical examinations, and tuberculin skin tests; and 2 studies did not specify their method of diagnosis (Table 1). Only 1 [19] of the 13 studies demonstrated a statistically significant difference in rates of tuberculosis among contacts of HIV-positive and HIV-negative tuberculosis cases. There was no statistical difference between the groups (OR, 1.07 [95% CI, .74–1.56]) and a low level of between-study heterogeneity (I² = 39.0%) (Supplementary Appendix).

HIV Infection in Contacts

Of 12 studies on HIV infection among contacts, the total number of contacts of HIV-positive and HIV-negative index cases from all studies was 2891 (median, 153 [IQR, 88–350]) and 15 533 (median, 215 [IQR, 86–660]), respectively (Table 1). Seven of these studies were implemented in Africa, 4 from the Americas, and 1 in Asia. To diagnose HIV in contacts, 4 studies used an ELISA test, 2 used a combination of the ELISA and Western blot tests, and 3 studies did not specify.

The odds ratio among contacts of HIV-positive and HIV-negative index cases ranged from 1.0 in Kifai and colleagues [27] to 48.0 in Reichler and colleagues [46]. The pooled random effects OR was 4.9 (95% CI, 3.0–8.0; I² = 77.0%), indicating that the HIV serostatus of the index case is a marker of HIV infection among contacts. All studies had more HIV-positive contacts among HIV-positive index cases compared with HIV-negative index cases except for 1 study [27], which had an equal amount. Nine of the 12 studies had statistically significantly more HIV-positive contacts among HIV-positive index cases compared with HIV-negative index cases (Figure 3). The pooled increased odds of HIV infection among HIV-positive index cases was substantially higher in studies from countries outside of Africa (N_studies = 5; OR, 9.80; 5.88–16.34) compared with African countries (N_studies = 7; OR, 2.79; 1.60–4.86) (Figure 3).

Figure 3.

Risk of HIV infection among contacts of HIV-positive and HIV-negative tuberculosis cases, stratified by the study location†. Abbreviation: HIV, human immunodeficiency virus. †In all analyses, the reference category for the measure of association are case contacts of HIV-negative tuberculosis cases.

Study Quality and Publication Bias Assessment

Generally, studies were either of moderate or high study quality (Supplementary Appendix). For studies investigating M. tuberculosis infection, 13 studies were of high quality, 9 were of moderate quality, and 3 were of low quality. For studies investigating tuberculosis, 6, 5, and 2 studies were of high, moderate, and low quality. For studies investigating HIV in contacts, 6 studies of high and moderate quality, respectively.

When we assessed publication bias through inspection of funnel plots for each outcome, we found no evidence of publication bias. Harbord’s test gave nonsignificant P values of 0.56, 0.17, and 0.29 for the outcomes of M. tuberculosis infection, tuberculosis, and HIV.

CONCLUSION

Our results suggest that differential M. tuberculosis infection rates in contacts of HIV-infected and uninfected tuberculosis cases is driven predominantly by the level of immunosuppression among persons living with HIV and the lung cavitation of tuberculosis of that case. Upon stratifying our results on index case characteristics such as lung cavitary status, CD4 count, or AIDS status, we found that heterogeneity among studies reduced considerably. The rate of tuberculosis among contacts of HIV-positive index cases did not significantly differ from contacts of tuberculosis cases without HIV; however, HIV infection was 5 times more common in contacts of HIV-positive tuberculosis cases. This was modified by whether the study was in or outside of sub-Saharan Africa, likely because of the background HIV rate.

A previous systematic review on this topic was performed in 2001 [7]. After applying a random effects model to 4 contact studies, this review concluded that tuberculosis cases were not more infectious than HIV-negative tuberculosis cases. In this updated meta-analysis including 25 studies investigating M. tuberculosis infection, our results suggest that tuberculosis patients with HIV are less infectious than HIV-uninfected tuberculosis cases and that this was modified by the severity of the tuberculosis case and/or the immunosuppression of index cases with HIV. This may partially explain heterogeneity in past studies in which some studies restrict index cases to smear-positive disease, whereas others also allow paucibacillary index cases. Our results also suggest that contacts of well-controlled HIV among people with tuberculosis may also have a lower risk of M. tuberculosis infection. In Kenyon et al and Huang et al, contacts of HIV-positive people with tuberculosis and high CD4 counts had a similar risk as HIV-negative people with tuberculosis. However, Cauthen et al found that HIV-positive people with tuberculosis had similar infection risk compared with new and previously diagnosed people with AIDS (see Figure 2). Further clarification is needed to confirm these findings on well-controlled HIV and its relation to M. tuberculosis infection risk in contacts. Smear, HIV, and lung cavitation among index cases are likely to be correlated in many of these studies. Many of the paucibacillary cases in these studies may represent HIV-seropositive patients with a low CD4 counts and therefore may be investigating similar patients. Cough duration of tuberculosis cases may be an important factor because people who are HIV-positive and have tuberculosis usually present to care earlier than people who are HIV-negative [53, 54]. Only 2 studies investigated effect modification by the index case’s cough duration [3, 17]. Both of these individual studies found that it did not influence rates of M. tuberculosis infection rates in contacts of tuberculosis cases with differing HIV status, however.

An important finding of this meta-analysis is the paucity of data found on the impact of antiretroviral therapy (ART) on the infectiousness of tuberculosis patients. ART has become increasingly accessible over the past 20 years and is associated with increasing CD4 counts in individuals living with HIV [55]. Two studies [17, 29] showing decreased transmission from HIV-positive tuberculosis patients with low CD4 counts may indirectly suggest higher infectiousness in coinfected patients on ART. A recent report [45] did not show ART as an influential variable; however, this research question remains unclear and requires further studies with larger sample sizes [56].

Our results also have important implications for policy on active case finding and HIV testing among contacts of tuberculosis cases. Current evidence graded by the World Health Organization for HIV testing in contacts of tuberculosis cases is considered of “low quality” [57]. We found that contacts of HIV-positive index cases were almost 5 times more likely to also have HIV infection. Although these studies were heterogeneous, the range of estimates consistently demonstrated a marked increased risk in HIV among contacts exposed to HIV-positive index cases. This association was modified by whether the study was set in sub-Saharan Africa, where HIV prevalence is high. Contacts from sub-Saharan African countries had a lower measure of association because the background HIV prevalence is much higher and therefore contacts of HIV-negative index cases are also at high-risk to acquire HIV. These results suggest that HIV testing of all contacts of tuberculosis cases in sub-Saharan Africa regardless of the HIV status of the index should be considered. This would support current global recommendations [57]. In areas with a lower background HIV prevalence and minimal resources, HIV testing only contacts of HIV-positive index cases is likely to be a resource efficient and effective method of finding new cases of HIV.

The results presented in this meta-analysis should be taken in context with several limitations. First, because of the epidemiology of tuberculosis and the inability of the tuberculin skin test to measure recent infection, M. tuberculosis infection may not be the ideal measure of tuberculosis infectiousness. M. tuberculosis infection in a population increases with age [58–60], and nondifferential misclassification may be present if transmission occurred before the exposure event investigated. We searched for molecular clustering studies and studies using tuberculin skin test conversion in contacts; however, few have been performed with data on this topic [17, 28, 61, 62]. Second, although several studies measured effect modification between infectiousness of HIV-positive tuberculosis cases and M. tuberculosis infection in contacts, few studies used similar measurements, which limited our ability to group studies. Third, very few studies stratified their disease results by initial M. tuberculosis infection status; therefore, we could not see if our tuberculosis results were modified by infection status. Fourth, although we adjust our results by several important secondary characteristics relevant to M. tuberculosis transmission, we acknowledge that other factors may be relevant and not widely measured among studies. Last, in the prevalent tuberculosis outcome analysis, we are not able to confirm the direction of transmission between index case and contacts. Potentially, an individual with HIV may progress more quickly to symptoms and diagnosis, and thus may be misclassified as the index case.

In conclusion, our results suggest people with tuberculosis coinfected with HIV are less infectious than HIV-uninfected cases when they have severe immunosuppression or paucibacillary disease. Contacts of coinfected index cases are almost 5 times more likely to also have HIV infection strongly indicating an urgent need for integrated HIV and tuberculosis policy and intervention.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author contributions. L. M. conceived the study, analyzed the data, wrote the first draft of the manuscript, and was the main investigator responsible for interpretation of the results. L. M., H. W., C. C., B. H., and M. E. C. contributed to study and data management. H. W., C. C., B. D. H., M. E. C., Q. L., C. C. W., and L. Z. assisted L. M. with the analytical plan and data analysis. All authors were involved in interpretation of the study results, reviewing and editing of the manuscript, and approval of the final version of the manuscript.

Acknowledgments. The authors thank the University of Georgia Interlibrary Loan office for their tireless efforts at retrieving hard-to-reach manuscripts, abstracts, and books on this topic. The authors also thank the research group at Jiangsu Province Tuberculosis Control Division at China’s Centers for Disease Control and Prevention for their useful and instructive comments on presentations of this project. Last, L. M. thanks Fernando Martinez and Maria Ines Pantoja for their constant support.

Financial support. C. C. W. and L. M. were supported in part by an investigator-initiated grant (AI 093856) and a diversity supplement grant (3R01AI093856–05W1) from the National Institutes of Allergy and Infectious Diseases. L. M. was supported by the Ruth L Kirschstein National Research Service Award. M. E. C. was supported by the Schlumberger Foundation Faculty for the Future Fellowship.

Potential conflicts of interest. The authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.