Skip Navigation


American Journal of Epidemiology Advance Access originally published online on March 5, 2007
American Journal of Epidemiology 2007 165(10):1134-1142; doi:10.1093/aje/kwk116
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
165/10/1134    most recent
kwk116v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (1)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by López-Gatell, H
Right arrow Articles by Anastos, K
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by López-Gatell, H
Right arrow Articles by Anastos, K
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

American Journal of Epidemiology Copyright © 2007 by the Johns Hopkins Bloomberg School of Public Health All rights reserved; printed in U.S.A.

ORIGINAL CONTRIBUTIONS

Effect of Tuberculosis on the Survival of Women Infected with Human Immunodeficiency Virus

H López-Gatell1,2, SR Cole1, NA Hessol3, AL French4, RM Greenblatt3, S Landesman5, S Preston-Martin6 and K Anastos7

1 Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD
2 Dirección de Medicina, Instituto Nacional de Ciencias Médicas y Nutrición, Salvador Zubirán, México D.F., México
3 Department of Medicine, University of California, San Francisco, San Francisco, CA
4 CORE Center/Stroger Hospital of Cook County, Chicago, IL
5 SUNY Downstate Medical Center, State University of New York, Brooklyn, NY
6 Department of Preventive Medicine, School of Medicine, University of Southern California, Los Angeles, CA
7 Lincoln Medical and Mental Health Center, Bronx, NY

Reprint requests to Dr. Hugo López-Gatell, Dirección de Medicina, Instituto Nacional de Ciencias Médicas y Nutrición, Salvador Zubirán, Vasco de Quiroga 15, Tlalpan C.P. 14000, México D.F., México (e-mail: hlgatell{at}jhsph.edu).

Received for publication May 2, 2006. Accepted for publication October 26, 2006.


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 
Evidence regarding the effect of tuberculosis (TB) disease on progression of human immunodeficiency virus (HIV) disease is inconclusive. The authors estimated the effect of time-varying incident TB on time to acquired immunodeficiency syndrome (AIDS)-related mortality using a joint marginal structural Cox model. Between 1995 and 2002, 1,412 HIV type 1 (HIV-1)-infected women enrolled in the Women's Interagency HIV Study were followed for a median of 6 years. Twenty-nine women incurred incident TB, and 222 died of AIDS-related causes. Accounting for age, CD4 cell count, HIV-1 RNA level, serum albumin level, and non-TB AIDS at study entry, as well as for time-varying CD4 cell count, CD4 cell count nadir, HIV-1 RNA level, peak HIV-1 RNA level, serum albumin level, HIV-related symptoms, non-TB AIDS, anti-Pneumocystis jiroveci prophylaxis, antiretroviral therapy, and household income, the hazard ratio for AIDS-related death comparing time after incident TB with time before incident TB was 4.0 (95% confidence interval (CI): 1.2, 14). The effect of incident TB on mortality was similar among highly active antiretroviral therapy (HAART)-exposed women (hazard ratio = 4.3, 95% CI: 0.9, 22) and non-HAART-exposed women (hazard ratio = 3.9, 95% CI: 0.9, 17; interaction p = 0.91). Although results were imprecise because few women incurred TB, irrespective of HAART exposure, incident TB increases the hazard of AIDS-related death among HIV-infected women.

acquired immunodeficiency syndrome; causality; disease progression; HIV infections; models, statistical; survival analysis; tuberculosis; women

Abbreviations: AIDS, acquired immunodeficiency syndrome; CI, confidence interval; HAART, highly active antiretroviral therapy; HIV, human immunodeficiency virus; HR, hazard ratio; PCP, Pneumocystis jiroveci pneumonia; TB, tuberculosis; WIHS, Women's Interagency HIV Study


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 
Approximately 25 million people worldwide are coinfected with human immunodeficiency virus (HIV) and Mycobacterium tuberculosis (1). Tuberculosis (TB) disease is the most common opportunistic infectious disease among HIV-infected persons (2), and HIV infection is the strongest known risk factor for TB disease (3). A synergistic impact of TB disease and HIV infection may produce quicker depletion of CD4-positive T lymphocytes (CD4 cells) (4–6) and shorten survival. However, the epidemiologic evidence regarding the effect of TB disease on the progression of HIV disease is inconclusive (7).

Most prior studies of HIV-infected persons that have found TB disease to increase the hazard of death have had a prospective design but have determined participants' TB disease status only at study entry (8–13). Such simple categorization neglects the sequence of TB and non-TB acquired immunodeficiency syndrome (AIDS)-defining conditions. Conversely, investigators who have concluded that TB disease does not increase mortality among HIV-infected persons have conducted mainly retrospective studies or have compared groups defined by the first AIDS-defining condition, which may or may not have been TB disease (14–17). Such approaches prevent investigators from estimating the effect of TB disease among persons who already have AIDS. Finally, some researchers have controlled for confounding due to HIV stage by stratifying results on time-varying markers of immunosuppression, such as CD4 cell count (12, 13, 17). Researchers who stratify their data on such markers may underestimate the TB effect by estimating, at best, only the effects of TB disease that are not mediated through such markers.

If TB disease shortens HIV-related survival, it is imperative to improve efforts to prevent TB disease among HIV-seropositive persons through interventions such as treatment for latent TB infection. We estimated the effects of incident pulmonary and extrapulmonary TB disease on AIDS-related and all-cause mortality among HIV-infected women enrolled in a large prospective cohort study. We hypothesized that TB disease would increase the risk of death but have a reduced effect among women who had received highly active antiretroviral therapy (HAART).


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 References
 
Study population
The Women's Interagency HIV Study (WIHS) (18) is an ongoing, prospective multicenter study of HIV type 1 (HIV-1) infection among women in the United States. In 1994 and 1995, WIHS personnel recruited 2,056 HIV-seropositive women and 569 HIV-seronegative women at high risk of HIV infection from primary-care and hospital-based medical units, community support groups, and research and drug rehabilitation programs in six US locations: Brooklyn, New York City; the Bronx, New York City; Chicago, Illinois; Southern California and Hawaii; the San Francisco Bay Area; and Washington, DC. At study entry, participants at all sites had similar sociodemographic and immunologic characteristics (18).

Participants in the study undergo semiannual physical examinations and provide specimens for laboratory measurements, including lymphocyte subset counts and plasma HIV-1 RNA levels. In addition, participants answer interviewer-administered questionnaires about health status, medication use, use of health services, and health-related behaviors. The 259 HIV-seropositive participants from Washington, DC, were excluded because TB registry data were unavailable. The present analysis focused on 1,412 (79 percent) of the remaining 1,797 HIV-infected WIHS participants with no reported history of TB disease and no HAART use who had complete data available for the variables of interest at study entry and had attended at least two study visits.

Endpoint ascertainment
Participants were followed from the second WIHS visit until death, dropout, or the date of analysis on September 29, 2002—whichever came first. Thus, participants could contribute a maximum of 7 years to the analysis. Participants from Southern California were censored on December 31, 1999, because TB registry data were unavailable after this date. The endpoints of interest were AIDS-related mortality and all-cause mortality. Dates and causes of death were ascertained through death certificate abstraction via the National Death Index. Deaths were classified as AIDS-related if 1) the cause of death listed on the certificate was an AIDS-defining condition (19) or 2) the last recorded CD4 cell count was below 200 cells/mm3 and the declared cause of death was either a nonspecific infection or organ failure (e.g., pneumonia with an unidentified microorganism, sepsis, multiple organ failure) or AIDS without further specification (20).

Assessment of TB disease
The WIHS uses a mixed protocol of active and passive detection to ascertain clinical events (18). Self-reporting of selected AIDS-related medical conditions, including TB disease, triggers confirmation by medical record abstraction (21). TB disease is considered confirmed by a medical record upon isolation of M. tuberculosis in tissue or fluid. Independent of self-reports, personnel at all WIHS sites except Washington, DC, conduct periodic searches of the TB registries of state or local public health departments. The exposure of interest was confirmed incident TB disease, defined as 1) the first self-report of pulmonary or extrapulmonary TB disease at any semiannual visit undertaken 6 months beyond enrollment in the WIHS, confirmed by medical record abstraction, or 2) incident TB disease found by active surveillance of TB registries or death certificates. After exploration of various functional forms, incident TB disease was modeled as a time-varying dichotomous indicator that was set to 0 for all women at study entry and changed to 1 after onset of TB disease.

Assessment of covariates
The definition of HAART was based on the guidelines of a US Department of Health and Human Services panel (22) and has been previously published (23). Typical HAART regimens consisted of two or more nucleoside or nucleotide reverse transcriptase inhibitors in combination with at least one protease inhibitor or one nonnucleotide reverse transcriptase inhibitor. A binary time-varying indicator of HAART initiation was set to 0 for all women at study entry (before HAART was introduced), was changed to 1 at the first visit at which a woman reported using HAART, and was then changed back to 0 if a woman discontinued HAART. Antiretroviral therapies not meeting the definition of HAART were classified as either monotherapy or combination therapy and were represented by separate dichotomous time-varying indicators.

Age at study entry was obtained from each woman. Levels of T cells and their subsets were determined by immunofluorescent flow cytometry at laboratories participating in the National Institute of Allergy and Infectious Diseases quality assurance program (24). Plasma HIV-1 RNA level was measured using the isothermal nucleic acid sequence-based amplification (NASBA/Nuclisens) method (Organon Teknika Corporation, Durham, North Carolina), and the data were log10-transformed. CD4 cell count, CD4 cell count nadir, HIV-1 RNA level, peak HIV-1 RNA level, and serum albumin level (25) measured at each visit during follow-up were modeled as continuous time-varying variables. Self-reports of low income (annual household income below $12,000), use of anti-Pneumocystis jiroveci (formerly Pneumocystis carinii) pneumonia (PCP) prophylaxis (i.e., trimethoprim, cotrimoxazole, dapsone, or aerosolized pentamidine), HIV-related symptoms (i.e., persistent fever or night sweats, and unintentional weight loss), and AIDS-defining conditions were represented as binary indicators that were allowed to vary over follow-up. TB disease, low CD4 cell count/CD4 cell percentage and invasive cervical cancer were excluded from the definition of AIDS-defining conditions. For the 28 percent of longitudinal data points that were missing, the most recently observed values were carried forward.

Statistical analysis
We approximated the parameters of a joint marginal structural Cox proportional hazards model (26) using a weighted pooled logistic regression model (27). Assuming no unmeasured confounding, no informative censoring, and no model misspecification, this method allows estimation of the total effect of TB disease on mortality among HAART-exposed women and non-HAART-exposed women. These same assumptions apply to standard approaches (e.g., Cox regression). Moreover, the assumptions of the present analytical strategy are less restrictive than standard approaches when estimating the effects of time-varying exposures (26). The additional restriction of the standard approaches amounts to disallowing feedback between the time-varying exposure and time-varying confounders.

This joint marginal structural model included as regressors time-varying incident TB disease, time-varying HAART use, and a restricted cubic spline representing time since study entry, with knots at the 5th, 35th, 65th, and 95th percentiles, as well as age, CD4 cell count, HIV-1 RNA level, serum albumin level, and history of non-TB AIDS-defining conditions, all measured at study entry. Such a model appropriately controls for time-varying confounding by weighting each participant's person-time contribution proportionally to the product of 1) the inverse probability of her observed incident TB disease status and 2) the inverse probability of her observed HAART initiation and discontinuation status, each estimated from the available data as a function of the history of time-varying predictors (23, 28).

To construct the time-varying inverse probability of incident TB disease weights, we estimated the probability of incident TB disease at a given study visit using a separate pooled logistic regression model including CD4 cell count and HIV-1 RNA level, both measured at study entry, as well as time-varying CD4 cell count, HIV-1 RNA level, serum albumin level, fever or sweats, weight loss, non-TB AIDS-defining conditions, and the indicators of HAART initiation and discontinuation. To assure the correct time sequence, data on all time-varying covariates were lagged one visit and therefore were measured before TB disease onset.

To construct the time-varying inverse probability of HAART initiation weights, we included CD4 cell count, HIV-1 RNA level, serum albumin level, and the indicator of anti-PCP prophylaxis, all measured at study entry, as well as time-varying incident TB disease, CD4 cell count, CD4 cell count nadir, HIV-1 RNA level, serum albumin level, and indicators for anti-PCP prophylaxis, previous use of combination therapy, and any non-HAART antiretroviral therapy, in a similar separate pooled logistic regression model.

Finally, to construct the inverse probability of HAART discontinuation weights, we included age and HIV-1 RNA level, both measured at study entry, as well as time-varying incident TB disease, CD4 cell count, CD4 cell count nadir, HIV-1 RNA level, serum albumin level, and indicators for weight loss, low income, and previous use of non-HAART antiretroviral therapy.

Eight percent (111 of 1,412) of the participants were lost to follow-up before the date of analysis. To reduce the potential impact of informative censoring due to measured covariates, we calculated inverse probability-of-censoring weights using CD4 cell count, HIV-1 RNA level, and anti-PCP prophylaxis, all measured at study entry, as well as time-varying TB disease, HAART initiation and discontinuation, CD4 cell count, and peak HIV-1 RNA level. The final inverse probability weights were stabilized and taken as the fourfold product of incident TB disease, HAART initiation, HAART discontinuation, and censoring weights (28, 29). To ameliorate the impact of extremely influential values, weights were censored at the first and 99th percentiles.

Results were not appreciably altered by further inclusion of ethnicity, health insurance coverage, other HIV-related symptoms (i.e., diarrhea, oral candidiasis, acral dysesthesia), total lymphocyte count, CD8 cell count, hemoglobin level, platelet level, serum creatinine or blood urea nitrogen level, education, type of housing, depression, smoking, or use of alcohol, intravenous drugs, heroin, methadone, crack cocaine, amphetamines, or marijuana (all measured at study entry or semiannually during follow-up).

Throughout the analysis, the hazard ratio was used as a measure of effect and the 95 percent confidence interval was used as a measure of precision. Bootstrap confidence intervals (30) were used for the marginal structural models (26). We explored possible modification of the effect of incident TB disease on AIDS-related death by HAART exposure and CD4 cell count less than 200 cells/mm3 measured at study entry. The effect of TB disease on mortality was not modified by geographic region (East Coast vs. West Coast and Chicago) (robust p for homogeneity = 0.74). Neither a plot of log-log survival by time nor a term for the interaction between incident TB disease and follow-up time (robust p for homogeneity = 0.76) suggested a strong departure from the proportional hazards assumption. Analyses were performed with SAS, version 9 (SAS Institute, Inc., Cary, North Carolina).


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 References
 
The study population was demographically similar to all HIV-infected women in the United States; 53 percent of participants were African-American, 25 percent were Latina, and 20 percent were White (table 1). At study entry, 27 percent reported a history of non-TB AIDS-defining conditions. The median CD4 cell count was 336 cells/mm3 (interquartile range, 171–517), and the median quantitative HIV-1 RNA level was 21,000 copies/ml (interquartile range, <4,000–99,500). A total of 258 participants were observed in Brooklyn, 323 in the Bronx, 225 in Chicago, 337 in Southern California/Hawaii, and 269 in San Francisco.


View this table:
[in this window]
[in a new window]

  		TABLE 1. Characteristics of 1,412 HIV*-infected women at study entry (1995–1996) and averaged over a median of 6 years of follow-up (April 3, 1995–September 29, 2002), Women's Interagency HIV Study

 
During follow-up, 46 (3 percent) of the 1,412 participants self-reported incident TB disease. Four of these self-reports were confirmed by medical record abstraction, four by registry match, and 10 by both methods. Additionally, 11 unreported incident TB events were found through active surveillance (one in medical records, seven by registry match, two in death certificates, and one by both registry and death certificate). Thus, the total number of women who incurred confirmed incident TB disease was 29 out of 1,412 (2 percent), yielding an incident TB disease rate of 4.3 (95 percent confidence interval (CI): 3.0, 6.2) per 1,000 person-years (figure 1). Twenty of 29 TB events were pulmonary, six were extrapulmonary, and three were mixed. Women who incurred TB disease were evenly distributed across study sites; Brooklyn and the Bronx each had eight TB events, Chicago and Southern California/Hawaii each had five TB events, and the remaining three TB events occurred in San Francisco. During 3,365 HAART-free person-years of follow-up, 929 (66 percent) of the 1,412 women initiated HAART, yielding a HAART initiation rate of 276 (95 percent CI: 259, 294) per 1,000 person-years.


Figure 1
View larger version (11K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		FIGURE 1. Initiation of highly active antiretroviral therapy (HAART), total (all-cause) mortality, acquired immunodeficiency syndrome (AIDS)-related mortality, and incidence of tuberculosis (TB) disease per 100 women among 1,412 human immunodeficiency virus (HIV)-infected women followed for a median of 6 years in the Women's Interagency HIV Study, April 3, 1995–September 29, 2002.

 
Incident TB disease was independently predicted by variables measured both at study entry and over the course of follow-up (table 2). Incident TB disease was more frequent among women with lower CD4 cell counts during follow-up (comparing <50 cells/mm3 with ≥500 cells/mm3, hazard ratio (HR) = 9.0, 95 percent CI: 1.7, 48), women with higher HIV-1 RNA levels during follow-up (comparing ≥10,000 copies/ml with <4,000 copies/ml, HR = 2.1, 95 percent CI; 0.7, 7.0), women with lower albumin levels during follow-up (comparing the lower tertile with the upper tertile, HR = 8.2, 95 percent CI: 1.8, 37), and women who self-reported involuntary weight loss at follow-up visits (HR = 3.1, 95 percent CI: 1.4, 6.8). Incident TB disease was not independently associated with age, ethnicity, previous non-TB AIDS-defining conditions, or HAART use.


View this table:
[in this window]
[in a new window]

  		TABLE 2. Independent predictors of incident tuberculosis disease among 1,412 HIV*-infected women over a median of 6 years of follow-up, Women's Interagency HIV Study, April 3, 1995–September 29, 2002

 
Between April 3, 1995, and September 29, 2002, the 1,412 women were followed for a median of 6.0 years (interquartile range, 3.4–6.8), during which 355 deaths occurred; 222 of these deaths were AIDS-related and two were attributed to incident TB disease on the death certificate. Twelve of 29 women with TB disease died, 11 from AIDS-related causes; and 343 of 1,383 women without TB disease died, 211 from AIDS-related causes. Seventy deaths were recorded in Brooklyn, 77 in the Bronx, 73 in Chicago, 62 in Southern California, and 73 in San Francisco.

After accounting for age, CD4 cell count, HIV-1 RNA level, serum albumin level, and history of non-TB AIDS-defining conditions, all measured at study entry, as well as time-varying CD4 cell count, CD4 cell count nadir, HIV-1 RNA level, peak HIV-1 RNA level, serum albumin level, HIV-related symptoms (i.e., fever or night sweats, and weight loss), non-TB AIDS-defining conditions, anti-PCP prophylaxis, previous use of combination and any non-HAART antiretroviral therapy, HAART use, and low income in the joint marginal structural model, the hazard of AIDS-related death was 4.0 times (95 percent CI: 1.2, 14) larger among women who incurred incident TB disease than among women who did not (table 3). The effect of incident TB disease on AIDS-related mortality was similar for HAART-exposed women (HR = 4.3, 95 percent CI: 0.9, 22) and non-HAART-exposed women (HR = 3.9, 95 percent CI: 0.9, 17; robust p for homogeneity = 0.91). Analyses restricting the exposure to the 20 women with pulmonary TB disease provided a larger but less precise estimate of the effect on AIDS-related mortality (HR = 5.3, 95 percent CI: 1.1, 25). Having a CD4 cell count less than 200 cells/mm3 at study entry did not appear to modify the effect of incident TB disease on AIDS-related mortality (p for homogeneity = 0.64). For all-cause mortality, the hazard ratio from the joint marginal structural model was 2.3 (95 percent CI: 0.7, 7.4); hazard ratios were similar among HAART-exposed women (HR = 2.1, 95 percent CI: 0.4, 12) and non-HAART-exposed women (HR = 2.3, 95 percent CI: 0.6, 9.3; robust p for homogeneity = 0.90).


View this table:
[in this window]
[in a new window]

  		TABLE 3. Effect of incident tuberculosis on AIDS*-related and all-cause mortality among 1,412 HIV*-infected women according to their use of highly active antiretroviral therapy during follow-up, Women's Interagency HIV Study, April 3, 1995–September 29, 2002

 
Standard adjustment for the same set of time-fixed and time-varying covariates using a Cox proportional hazards model produced an attenuated hazard ratio for AIDS-related mortality of 2.5 (95 percent CI: 1.2, 5.3), with hazard ratios of 1.0 (95 percent CI: 0.2, 5.2) for HAART-exposed women and 3.5 (95 percent CI: 1.5, 8.0) for non-HAART-exposed women, whereas standard adjustment for only time-fixed covariates yielded a hazard ratio of 3.5 (95 percent CI: 1.8, 6.9), with hazard ratios of 3.5 (95 percent CI: 1.7, 7.5) for HAART-exposed women and 3.4 (95 percent CI: 0.8, 15) for non-HAART-exposed women. A similar pattern was observed for all-cause mortality (table 3).


   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
References
 
In this prospective study, we estimated the total effects of incident pulmonary and extrapulmonary TB on AIDS-related and all-cause mortality among HIV-1-infected women. Incident TB disease was associated with a fourfold increase in the hazard of AIDS-related mortality, which was largely obfuscated using standard analytical methods. Contrary to our hypothesis, exposure to HAART did not modify (i.e., attenuate) this effect. Some prior studies (12, 13) have found the TB effect on mortality in HIV-infected persons to be stronger among those with CD4 counts greater than 200 cells/mm3. We found a similar pattern, but because of the small number of women with TB, our estimates were imprecise.

The results of this analysis are consistent with those of previous studies showing shorter survival among HIV-infected persons who develop incident TB disease (9, 10, 12, 13), but the estimates of effect found here are larger. The primary outcome of this study was the effect of incident TB disease on AIDS-related mortality, rather than all-cause mortality as it was in many previous studies (12, 13). Progression of HIV disease is probably better represented by the more specific measure of AIDS-related mortality than by all-cause mortality, which, to the degree that non-AIDS-related mortality occurs, would dilute the impact of incident TB disease.

HIV disease progression is a predictor of both incident TB disease and mortality. Therefore, to reduce confounding when estimating the effect of incident TB disease on mortality, one needs to account for HIV disease progression. When using standard statistical methods, investigators adjust for or stratify on time-varying markers of HIV disease progression, such as CD4 cell count or HIV-1 RNA level (15–17). However, adjusting for or stratifying on these markers of HIV disease progression will not allow unbiased estimation of the total (i.e., direct and indirect) effect of incident TB disease on mortality (28) and may induce selection bias (31) whenever there is a time-varying confounder affected by previous exposure. In particular, CD4 cell count is probably an intermediate variable on the pathway from incident TB to death. This is evidenced by the fact that, in our study, fully adjusted standard methods, which did not allow the effect of incident TB to be transmitted through time-varying CD4 cell count, showed a markedly attenuated effect of incident TB on AIDS-related mortality. Therefore, we employed a joint marginal structural model, which, like standard methods, assumes no unmeasured confounding, no informative censoring, and no model misspecification; but unlike standard methods, it allows consistent estimation of the total effect of incident TB on mortality (28, 29).

Some of the effects of TB disease on mortality are probably mediated by anti-TB treatment. Unfortunately, we do not have detailed information on TB treatment for these women. With the exception of the two TB-related deaths, it is likely that these women received TB treatment because they survived for more than 6 months after TB onset. If all women were properly and promptly treated when TB arose, the effect of TB on mortality would probably be smaller than our estimate. Conversely, if no women were properly and promptly treated, the effect of TB on mortality would probably be larger than our estimate.

To our knowledge, this is the first study assessing the effects of incident TB disease on mortality among HAART-exposed and -unexposed women. Despite the known benefits of HAART in reducing AIDS incidence and extending the survival of HIV-infected persons (23, 32–34), we did not find modification of the effect of incident TB disease on mortality by HAART. We may have failed to detect true modification by HAART for the following reasons: 1) a relatively small number of women incurred incident TB disease during follow-up; 2) our assumption that women continuously adhere to HAART between initiation and discontinuation may have led to misclassification of HAART-exposed and non-HAART-exposed women; 3) anti-TB drugs (e.g., rifampicin and isoniazid) may impair the efficacy of HAART by reducing blood concentrations of some antiretroviral drugs (35); 4) resistance due to accumulated viral mutations may have reduced HAART effectiveness (36); or 5) some women may have experienced immune reconstitution disease, a paradoxical deleterious response to HAART (37) that may offset the beneficial effects of HAART and shorten the survival of women with TB disease.

The present study had several potential limitations. First, during the course of follow-up, only 29 women incurred confirmed incident TB disease. This small proportion of exposed participants is evident in our relatively wide confidence intervals. Second, the WIHS is a study of women, and the results may or may not be generalizable to men; however, we have no reason to believe that the results would differ appreciably by sex. Third, the WIHS study population is largely socioeconomically disadvantaged, and the results may or may not apply to persons at differing socioeconomic levels. Fourth, the WIHS cohort is an HIV-seroprevalent cohort; duration of HIV infection was unknown, and the comparison between women who incurred incident TB disease during follow-up and those who did not may have been influenced by differences in HIV disease beyond CD4 cell count and HIV-1 RNA level. However, it has been shown that time-varying information on CD4 cell count and HIV-1 RNA level suffices to control for differences in duration of infection (38). Fifth, although incident TB disease was confirmed through medical records or registries, misclassification of TB disease may have remained. Sixth and last, as with all observational studies, it is possible that the findings could be explained by unmeasured confounding. The combined impact of these potential limitations is difficult to predict.

The strengths of this work include the low rate of attrition and the use of the National Death Index for classifying endpoints, making emigrative selection bias and information bias due to misclassification of endpoints, respectively, unlikely explanations for the findings. Additionally, serial collection of biologic samples allows longitudinal control for markers of HIV disease progression. In conclusion, among these HIV-infected women, irrespective of HAART exposure, incident TB disease appeared to induce a fourfold increase in the hazard of AIDS-related death. Assuming that our findings have a causal interpretation, they emphasize the need for preventive interventions (e.g., treatment of latent TB infection) to avoid the occurrence of TB disease and the resultant increased mortality among HIV-infected persons, irrespective of HAART use.


   ACKNOWLEDGMENTS
 
The Women's Interagency HIV Study (WIHS) is funded by the National Institute of Allergy and Infectious Diseases, with supplemental funding from the National Cancer Institute and the National Institute on Drug Abuse (grants UO1-AI-35004, UO1-AI-31834, UO1-AI-34994, UO1-AI-34989, UO1-AI-34993, and UO1-AI-42590). Funding is also provided by the National Institute of Child Health and Human Development (UO1-HD-32632) and the National Center for Research Resources (MO1-RR-00071, MO1-RR-00079, MO1-RR-00083). Dr. H. López-Gatell was supported by a scholarship credit from the National Council for Science and Technology (CONACyT), México.

Data were collected by the WIHS Collaborative Study Group at the following study centers: New York City/Bronx Consortium (Principal Investigator (PI), Dr. Kathryn Anastos); Brooklyn, New York (PI, Dr. Howard Minkoff); the Connie Wofsy Study Consortium of Northern California (PI, Dr. Ruth Greenblatt); the Los Angeles County/Southern California Consortium (PI, Dr. Alexandra Levine); the Chicago Consortium (PI, Dr. Mardge Cohen); and the Data Coordinating Center (PI, Dr. Stephen Gange).

The authors thank Drs. Richard Chaisson, Miguel A. Hernán, and Alvaro Muñoz for constructive suggestions.

An earlier version of this work was awarded the Student Paper Prize at the 2006 North American Congress of Epidemiology.

Conflict of interest: none declared.


   References
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 References
 

  1. Corbett EL, Watt CJ, Walker N, et al. The growing burden of tuberculosis: global trends and interactions with the HIV epidemic. Arch Intern Med (2003) 163:1009–21.[Abstract/Free Full Text]
  2. Joint United Nations Programme on HIV/AIDS (UNAIDS) and World Health Organization (WHO). AIDS epidemic update: December 2005. (Publication no. UNAIDS/05.19E). Geneva, Switzerland: UNAIDS/WHO, 2005. (http://data.unaids.org/publications/IRC-pub06/epi_update2005_en.pdf).
  3. World Health Organization. Global tuberculosis control—surveillance, planning, financing: WHO report 2006. (Publication no. WHO/HTM/TB/2006.362). (2006) Geneva, Switzerland: World Health Organization. (http://www.who.int/tb/publications/global_report/2006/en/).
  4. Hertoghe T, Wajja A, Ntambi L, et al. T cell activation, apoptosis and cytokine dysregulation in the (co)pathogenesis of HIV and pulmonary tuberculosis (TB). Clin Exp Immunol (2000) 122:350–7.[CrossRef][Web of Science][Medline]
  5. Toossi Z. Virological and immunological impact of tuberculosis on human immunodeficiency virus type 1 disease. J Infect Dis (2003) 188:1146–55.[CrossRef][Web of Science][Medline]
  6. Vanham G, Edmonds K, Qing L, et al. Generalized immune activation in pulmonary tuberculosis: co-activation with HIV infection. Clin Exp Immunol (1996) 103:30–4.[CrossRef][Web of Science][Medline]
  7. Del Amo J, Malin AS, Pozniak A, et al. Does tuberculosis accelerate the progression of HIV disease? Evidence from basic science and epidemiology. AIDS (1999) 13:1151–8.[CrossRef][Web of Science][Medline]
  8. Whalen C, Horsburgh CR, Hom D, et al. Accelerated course of human immunodeficiency virus infection after tuberculosis. Am J Respir Crit Care Med (1995) 151:129–35.[Abstract]
  9. Perneger TV, Sudre P, Lundgren JD, et al. Does the onset of tuberculosis in AIDS predict shorter survival? Results of a cohort study in 17 European countries over 13 years. AIDS in Europe Study Group. BMJ (1995) 311:1468–71.[Abstract/Free Full Text]
  10. Leroy V, Salmi LR, Dupon M, et al. Progression of human immunodeficiency virus infection in patients with tuberculosis disease. A cohort study in Bordeaux, France, 1988 –1994. The Groupe d'Epidemiologie Clinique du Sida en Aquitaine (GECSA). Am J Epidemiol (1997) 145:293–300.[Abstract/Free Full Text]
  11. Munsiff SS, Alpert PL, Gourevitch MN, et al. A prospective study of tuberculosis and HIV disease progression. J Acquir Immune Defic Syndr Hum Retrovirol (1998) 19:361–6.[Medline]
  12. Whalen CC, Nsubuga P, Okwera A, et al. Impact of pulmonary tuberculosis on survival of HIV-infected adults: a prospective epidemiologic study in Uganda. AIDS (2000) 14:1219–28.[CrossRef][Web of Science][Medline]
  13. Badri M, Ehrlich R, Wood R, et al. Association between tuberculosis and HIV disease progression in a high tuberculosis prevalence area. Int J Tuberc Lung Dis (2001) 5:225–32.[Web of Science][Medline]
  14. Mocroft AJ, Lundgren JD, d'Armino Monforte A, et al. Survival of AIDS patients according to type of AIDS-defining event. The AIDS in Europe Study Group. Int J Epidemiol (1997) 26:400–7.[Abstract/Free Full Text]
  15. Del Amo J, Petruckevitch A, Phillips A, et al. Disease progression and survival in HIV-1-infected Africans in London. AIDS (1998) 12:1203–9.[CrossRef][Web of Science][Medline]
  16. Petruckevitch A, Del Amo J, Phillips AN, et al. Disease progression and survival following specific AIDS-defining conditions: a retrospective cohort study of 2048 HIV-infected persons in London. AIDS (1998) 12:1007–13.[CrossRef][Web of Science][Medline]
  17. Del Amo J, Perez-Hoyos S, Hernandez Aguado I, et al. Impact of tuberculosis on HIV disease progression in persons with well-documented time of HIV seroconversion. J Acquir Immune Defic Syndr (2003) 33:184–90.[CrossRef][Web of Science][Medline]
  18. Barkan SE, Melnick SL, Preston-Martin S, et al. The Women's Interagency HIV Study. WIHS Collaborative Study Group. Epidemiology (1998) 9:117–25.[CrossRef][Web of Science][Medline]
  19. revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR Recomm Rep (1992) 41:1–19.[Medline]
  20. Cohen MH, French AL, Benning L, et al. Causes of death among women with human immunodeficiency virus infection in the era of combination antiretroviral therapy. Am J Med (2002) 113:91–8.[CrossRef][Web of Science][Medline]
  21. Hessol NA, Schwarcz S, Ameli N, et al. Accuracy of self-reports of acquired immunodeficiency syndrome and acquired immunodeficiency syndrome-related conditions in women. Am J Epidemiol (2001) 153:1128–33.[Abstract/Free Full Text]
  22. US Department of Health and Human Services and Henry J. Kaiser Family Foundation Panel on Clinical Practices for the Treatment of HIV Infection. Guidelines for the use of antiretroviral agents in HIV-infected adults and adolescents. (2004) Bethesda, MD: National Institutes of Health. (http://aidsinfo.nih.gov).
  23. Cole SR, Hernan MA, Robins JM, et al. Effect of highly active antiretroviral therapy on time to acquired immunodeficiency syndrome or death using marginal structural models. Am J Epidemiol (2003) 158:687–94.[Abstract/Free Full Text]
  24. Calvelli T, Denny TN, Paxton H, et al. Guideline for flow cytometric immunophenotyping: a report from the National Institute of Allergy and Infectious Diseases, Division of AIDS. Cytometry (1993) 14:702–15.[CrossRef][Web of Science][Medline]
  25. Feldman JG, Gange SJ, Bacchetti P, et al. Serum albumin is a powerful predictor of survival among HIV-1-infected women. J Acquir Immune Defic Syndr (2003) 33:66–73.[Web of Science][Medline]
  26. Hernan MA, Brumback B, Robins JM. Marginal structural models to estimate the joint causal effect of nonrandomized treatments. J Am Stat Assoc (2001) 96:440–8.[CrossRef][Web of Science]
  27. D'Agostino RB, Lee ML, Belanger AJ, et al. Relation of pooled logistic regression to time dependent Cox regression analysis: The Framingham Heart Study. Stat Med (1990) 9:1501–15.[Web of Science][Medline]
  28. Robins JM, Hernan MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology (2000) 11:550–60.[CrossRef][Web of Science][Medline]
  29. Hernan MA, Brumback B, Robins JM. Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men. Epidemiology (2000) 11:561–70.[CrossRef][Web of Science][Medline]
  30. Carpenter J, Bithell J. Bootstrap confidence intervals: when, which, what? A practical guide for medical statisticians. Stat Med (2000) 19:1141–64.[CrossRef][Web of Science][Medline]
  31. Cole SR, Hernan MA. Fallibility in estimating direct effects. Int J Epidemiol (2002) 31:163–5.[Abstract/Free Full Text]
  32. Hammer SM, Squires KE, Hughes MD, et al. A controlled trial of two nucleoside analogues plus indinavir in persons with human immunodeficiency virus infection and CD4 cell counts of 200 per cubic millimeter or less. AIDS Clinical Trials Group 320 Study Team. N Engl J Med (1997) 337:725–33.[Abstract/Free Full Text]
  33. Detels R, Muñoz A, McFarlane G, et al. Effectiveness of potent antiretroviral therapy on time to AIDS and death in men with known HIV infection duration. Multicenter AIDS Cohort Study Investigators. JAMA (1998) 280:1497–503.[Abstract/Free Full Text]
  34. Gange SJ, Barron Y, Greenblatt RM, et al. Effectiveness of highly active antiretroviral therapy among HIV-1 infected women. J Epidemiol Community Health (2002) 56:153–9.[Abstract/Free Full Text]
  35. Peloquin CA. Therapeutic drug monitoring in the treatment of tuberculosis. Drugs (2002) 62:2169–83.[CrossRef][Web of Science][Medline]
  36. Gange SJ, Schneider MF, Grant RM, et al. Genotypic resistance and immunologic outcomes among HIV-1-infected women with viral failure. J Acquir Immune Defic Syndr (2006) 41:68–74.[CrossRef][Web of Science][Medline]
  37. Lawn SD, Bekker LG, Miller RF. Immune reconstitution disease associated with mycobacterial infections in HIV-infected individuals receiving antiretrovirals. Lancet Infect Dis (2005) 5:361–73.[CrossRef][Web of Science][Medline]
  38. Tarwater PM, Mellors J, Gore ME, et al. Methods to assess population effectiveness of therapies in human immunodeficiency virus incident and prevalent cohorts. Am J Epidemiol (2001) 154:675–81.[Abstract/Free Full Text]

Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?


This article has been cited by other articles:


Home page
 Am J EpidemiolHome page
S. R. Cole and M. A. Hernan
Constructing Inverse Probability Weights for Marginal Structural Models
Am. J. Epidemiol., September 15, 2008; 168(6): 656 - 664.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow All Versions of this Article:
165/10/1134    most recent
kwk116v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrow Search for citing articles in:
ISI Web of Science (1)
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by López-Gatell, H
Right arrow Articles by Anastos, K
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by López-Gatell, H
Right arrow Articles by Anastos, K
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?