Common variable immunodeficiency

In the setting of human immunodeficiency virus (HIV) infection, the clinical implications of American Thoracic Society (ATS) diagnostic criteria and the significance of a single positive respiratory culture for Mycobacterium kansasii are unknown. We retrospectively studied HIV-infected patients with pulmonary M. kansasii isolated between 1989 and 2002 at one institution. Of 127 patients, 33% fulfilled ATS disease criteria. Twenty-nine percent received at least three active drugs for at least 3 months, and 53% died. In survival analysis, a lower CD4 count (hazard ratio [HR], 1.6; 95% confidence interval [CI], 1.1-2.3) and positive smear microscopy (HR, 2.8; 95% CI, 1.3-6.1) were associated with mortality, whereas antiretroviral therapy (HR, 0.3; 95% CI, 0.1-0.8) and M. kansasii treatment (HR, 0.4; 95% CI, 0.2-0.9) were associated with survival. ATS criteria did not predict mortality (HR, 0.9; 95% CI, 0.4-1.9). Fifteen patients (12%) apparently had indolent infection, not requiring immediate therapy. They had fewer positive cultures and lower rates of positive smear microscopy and ATS-defined disease. In HIV-infected patients with pulmonary M. kansasii infection, predictors of survival include higher CD4 counts, antiretroviral therapy, negative smear microscopy, and adequate treatment for M. kansasii infection, but not ATS diagnostic criteria. Withholding treatment in HIV-infected patients with respiratory M. kansasii isolates should only be considered with negative smear microscopy, few positive cultures, and mild immunosuppression.

Keywords: guidelines; mycobacterium infections, atypical; survival analysis

Mycobacterium kansasii is a common cause of serious pulmonary infections in patients with human immunodeficiency virus (HIV) infection. The estimated annual infection rate is 532/100,000 (1), and death is a frequent outcome (2-5). M. kansasii is one of the most pathogenic nontuberculous mycobacteria, with approximately 50% of respiratory cultures representing clinically important disease (6). In HIV-infected patients, some authorities state that a respiratory M. kansasii isolate always represents disease (4,7-10). However, the American Thoracic Society (ATS) guidelines apply similar criteria to M. kansasii and other nontuberculous mycobacteria, requiring multiple positive cultures to diagnose disease, regardless of HIV infection (11). Classifying a patient's syndrome as "indolent infection" (not requiring immediate treatment, a term we favor over "colonization") versus disease is important because treatment requires prolonged therapy with multiple drugs. Of added concern with HIV infection, there is a high frequency of drug toxicities and interactions between antimycobacterial and antirelroviral agents (12, 13).

Characteristics associated with survival and factors that distinguish disease from indolent infection have been incompletely studied in HIV-associated M. kansasii. In a systematic review, 8% of HIV-infected patients with pulmonary M. kansasii isolates had indolent infection, but definitions varied between studies (1). ATS guidelines require clinical, radiographic, and microbiologic criteria to diagnose disease (Table 1). In the review, patients who fulfilled ATS criteria likely had clinically important disease, but there were few data regarding outcomes in patients who did not fulfill ATS criteria. In addition, there were inadequate data to assess the prognostic importance of the ATS diagnostic criteria. Treatment with antimycobacterial therapy was associated with longer average survival (2, 3, 7, 9, 14-16), but other prognostic features were not assessed. A recent retrospective analysis of 55 patients with HIV-associated pulmonary M. kansasii found highly active antiretroviral therapy (HAART) and the absence of disseminated infection were associated with improved survival (5). This analysis was limited by the paucity of patients who did not meet ATS criteria (11 patients) and few deaths.

In the context of limited data regarding the prognostic factors and the uncertainty surrounding the identification of patients with indolent infection, we studied consecutive HIV-infected patients with a pulmonary M. kansasii isolate for over 13 years at one institution. We sought to investigate the prognostic importance of factors including ATS criteria and antimycobacterial treatment and to identify the frequency and characteristics of patients with indolent infection. Some of our results were previously reported in an abstract (17).

METHODS

Patients and Definitions

We retrospectively studied all HIV-infected patients with M. kansasii in a respiratory specimen at the San Francisco General Hospital between December 1989 and July 2002, identified by microbiology records. The committee on human research at the University of California, San Francisco approved this study.

ATS diagnostic criteria were determined retrospectively from microbiologie, clinical, and radiographic records using the 1997 guidelines in all cases. Drugs considered active against M. kansasii were isoniazid, rifamycins, ethambutol, newer macrolidcs, and fluoroquinolones. We defined treatment for M. kansasii infection in two ways. First, "any treatment" (at least two drugs for at least 3 months) was judged to be the minimum required to alter clinical outcomes. The absence of "any treatment" was required for the designation of "long-term untreated." second, "minimal significant treatment" (at least three drugs for at least 3 months), used for the survival analysis, was judged to be required to observe a significant clinical impact. Our treatment definitions were based on recommendations that treatment should include three drugs for 18 months (11). Indolent infection was defined by untreated long-term survival (at least 2 years without treatment with at least two drugs for at least 3 months). Characteristics of "untreated long-term survivors" were compared with "non-long-term survivors" (patients who survived less than 2 years regardless of treatment). Additional details regarding patients and definitions are provided in an online supplement.

Statistical Analysis

Continuous variables are presented as mean (SD) or median (interquartile range) and compared with t tests or Mann-Whitney U. Categoric variables were compared with Yates-corrected chi-square or Fisher's exact tests.

In survival analysis, we assessed ATS diagnostic criteria, smear microscopy, number of positive cultures, minimal significant antimycobacterial treatment, CD4 count, and HAART. Although smear microscopy and number of positive cultures are components of ATS diagnostic criteria, they were all retained because collinearity between them was lower than the predefined threshold of 0.6. An interaction between HAART (ever received during observation period) and time was modeled because the hazard ratio for HAART varied over time. Variables with missing values in greater than 15% of patients were excluded from analysis. Sex and age were not included in the primary analysis because of inadequate variability (90% men, very narrow age range). The era of diagnosis was classified as early (1989-1995) or recent (1996-2002), reflecting introduction of HAART, and studied in secondary analyses. The primary analysis was a multivariable extended Cox model containing all variables. secondary analyses included bivariate Cox models for all variables, Kaplan-Meier survival curve comparisons (Wilcoxon tests) for selected variables, models using backward selection, testing for interactions between treatment and baseline predictors, and models including sex, age, and radiographic abnormality. To optimize power, we maintained at least 10 events per variable (18).

To minimize bias favoring survival in treated patients, assessment of treatment was limited to patients surviving for at least 3 months, regardless of therapy. Analyses were performed using SAS^sup R^ version 8.0 (SAS Institute, Cary, NC). Additional details regarding statistical analysis are provided in an online supplement.

RESULTS

Patients, Treatment, and Outcomes

Over the 12-year study period, 156 patients with M. kanaasii infection were identified. One hundred forty-nine patients had available records. One hundred thirty-four patients were known to be HIV infected, and 127 patients had at least one pulmonary isolate. Two patients had documentation of previous M. kansasii isolation. Isolation of M. kansasli was made in the recent era (1996 or later) in 73 of 127 patients (57%). Patients in the two eras were generally similar, but CD4 counts were higher in recent-era patients (median [interquartile range], 41 [17-136] vs. 32 [9-59], p = 0.047). Patient characteristics are presented in Table 2. CD4 lymphocyte count, at the time of first isolate, was unavailable and was therefore imputed in 11 patients. Viral load was unavailable in 45% of patients and was therefore excluded from survival analysis. The minority of patients (33%) fulfilled ATS diagnostic criteria for disease. Chest radiographic criteria were fulfilled by 103 of 117 (88%) patients with adequate data (data incomplete in 8%). Clinical criteria were fulfilled by 81 of 116 (70%) patients with adequate data (data incomplete in 9%). Microbiologie criteria were fulfilled by 56 of 127 (44%) patients. Thirty three of 35 patients (94%) failed to fulfill the clinical component of ATS criteria because of a failure to reasonably exclude another illness as the cause of symptoms. Other diagnoses in this group included Pneumocystis pneumonia in 14 patients, community-acquired pneumonia in 10 patients, pulmonary vascular congestion in four patients, lymphoma in two patients, and tuberculosis, uremia, and multiple other problems in one patient each. Reclassifying the 33 patients in whom we failed to exclude another cause of their symptoms as patients fulfilling ATS diagnostic criteria did not significantly affect the results of subsequent analyses.

Data regarding treatment of M. kansasii are presented in Table 2. Patients who fulfilled ATS diagnostic criteria were more likely to receive minimal adequate anti-M. kansasii treatment (48% vs. 19%, p = 0.01). A total of 79 (62%) patients had therapy against M. kansasii initiated. Rates of treatment were similar in early and recent eras. Therapeutic regimens included a median of three (2-4) drugs and lasted a median of 9 (1-14) months. Of patients started on therapy, 18 of 79 (23%) were treated for at least 12 months. Regimens included ethambutol in 79 patients (100%), a rifamycin (usually rifampin) in 65 patients (82%), isoniazid in 58 patients (73%), a macrolide (clarithromycin or azithromycin) in 31 patients (39%), and a quinolone in 5 patients (6%).

The median (interquartile range) follow-up duration was 305 (115-831) days overall, 300 (138-721) days in patients who fulfilled ATS disease criteria, and 391 (116-1,181) days in patients who did not fulfill ATS disease criteria (p = 0.72 for the difference by ATS criteria). Mortality during follow-up was 53% overall, 56% in patients who fulfilled ATS disease criteria, and 50% in patients who did not fulfill ATS disease criteria (p = 0.64 for difference by ATS criteria). At 2 years, 29% of all patients, 23% of patients who fulfilled ATS disease criteria, and 31% of patients who did not fulfill ATS disease criteria had died (p = 0.50 for difference by ATS criteria). Of the 67 deaths, 5 (7%) were definitely attributable to M. kansasii infection, 7 (10%) to PCP, 19 (28%) to other infection or undiagnosed infection, 10 (15%) to undiagnosed respiratory failure, 9 (13%) to neoplasia, 8 (12%) to other causes, and 9 (13%) were unknown. It is likely that some of the deaths in the undiagnosed infection and respiratory failure groups were due to M. kansasii infection but could not be proven from the available data.

Survival Analysis

The primary analysis was a full model (all explanatory variables shown in Table 3 forced into the model). As noted in METHODS, sex and age were excluded because of inadequate variability. Hazard ratios from the primary analysis are presented in Table 3. Lower CD4 count, an absence of HAART, a lack of minimal adequate M. kansasii treatment, and positive smear microscopy were all statistically significantly associated with mortality. Differences between Kaplan-Meier survival curves for both positive smear microscopy and minimal adequate M. kansasii therapy were significant by Wilcoxon tests (Figure 1). Figure 1C suggests that a beneficial effect of treatment was most apparent in smearpositive patients. Of the variables entered into the model, only the number of positive respiratory cultures and ATS-defined disease status were not statistically significant.

In secondary analyses, age, sex, and type of chest radiographic abnormality were not statistically significant and did not significantly alter other results. In bivariate survival analysis, recentera patients had lower mortality; however, this effect was lost in multivariable analyses, and forcing this variable into multivariable models did not significantly affect the results of these analyses. Additional secondary analyses, including bivariate Cox models (assessing the variables individually) and the primary multivariable analysis performed with backward automated model selection, yielded results consistent with the primary analysis. In addition, studying the effect of treatment in the entire cohort, rather than exclusively the patients who survived for at least 3 months, showed a very strong association with survival. Whether patients fulfilled clinical, radiographie, or microbiologie components of the ATS diagnostic criteria was also assessed individually, and none was significantly associated with survival. We also repeated the primary analysis without imputed data for CD4 counts and with patients who fulfilled all ATS criteria except clinical criteria (caused by failure to exclude another cause for their symptoms) reclassified as fulfilling ATS disease criteria. Results of these analyses were consistent with the primary analysis. Finally, in testing for interactions with treatment, we found that the association between microscopy smear positivity and mortality was stronger in the absence of treatment (HR, 4.2; confidence interval, 1.8-9.9; vs. HR, 2.8; confidence interval, 1.3-6.1, in treated patients).

Long-term Untreated Survivors

Patients who survived for at least 2 years in the absence of treatment with at least two drugs active against M. kansasii for at least 3 months were classified as "long-term untreated survivors." For the purpose of our analysis, we assumed long-term untreated survivors had indolent infection with M. kansasii. Table 4 contains characteristics of the 15 long-term untreated survivors compared with characteristics of patients who died within 2 years, regardless of treatment ("non-long-term survivors"). Long-term untreated survivors had fewer positive cultures (pulmonary and extrapulmonary) and lower rates of positive smear microscopy and ATS-defined disease. The lower rate of ATS-defined disease was secondary to fewer total positive cultures. Long-term untreated survivors were also more likely to have received HAART and had higher CD4 counts and lower viral loads. We changed the definition of "long-term untreated survivor" to include only patients who received a maximum of one active agent for less than 1 month. Applying this more stringent definition decreased the number of patients with indolent infection to 12, with generally similar characteristics.

DISCUSSION

In this study, we found that factors that predicted mortality in HIV-infected patients with pulmonary M. kansasii were lower CD4 cell counts, lack of HAART, positive sputum smear microscopy, and lack of adequate mycobacterial treatment. Interestingly, ATS criteria were not associated with mortality, but patients who did not meet ATS criteria were more likely to have indolent infection than those who did. In addition, long-term untreated survivorship (colonization or indolent infection) was associated with a higher CD4 cell count, low HIV viral levels, use of HAART, and a lower frequency of positive sputum smear microscopy, number of positive cultures, and presence of extrapulmonary disease.

Taken together, the previously mentioned findings suggest that smear microscopy is a far more powerful predictor of mortality than ATS criteria. In addition, the benefit of treatment seemed to be most apparent in patients with positive sputum smear microscopy. As expected, we found that adequate therapy for M. kansasii, HAART, and higher CD4 counts were all associated with increased survival, suggesting that our sample and model provided valid results. Both HAART (4) and therapy against M. kansasii (2, 3, 7, 9, 14-16, 19) have been previously shown to be associated with survival.

Compared with many previous studies (2, 3, 5, 10, 15, 20, 21), ours has the advantage of including patients with a broad spectrum of seventy of pulmonary M. kansasii infection. We included all patients with at least one pulmonary M. kansasii isolate, regardless of ATS disease status. Our study is also the largest survival analysis in this disease and includes more untreated patients and more patients with indolent infection than previous reports. In addition, we limited our primary analysis to patients who survived for at least the duration of treatment we were testing, which rigorously addressed immortality bias. Immortality bias is introduced if patients fulfill criteria for entering a particular group (treatment or control) only if they survive for a given period of time (immortal period). In our study, treated patients must have survived for at least 3 months (minimal adequate treatment was defined as at least three active drugs for at least 3 months). An unadjusted analysis would not recognize this inherent survival advantage over untreated patients and would therefore bias the analysis toward finding a survival benefit in treated patients. We therefore limited the assessment of treatment exclusively to patients who survived for at least 3 months (see online supplement to METHODS). Our finding of a strong association in this context illustrates the robust effect of treatment on survival. Similar to previous reports, we observed that M. kansasii is primarily a pulmonary disease, even in HIV-infected patients (2-4, 7-10, 14-16, 19-21). Of 14 patients identified with extrapulmonary M. kansasii isolates, 7 had concomitant pulmonary isolates. Unfortunately, because of a limitation in our data collection and availability, we were unable to identify risk factors for the development of extrapulmonary M. kansasii.

Using our assumption that patients who survived for at least 2 years in the absence of treatment had indolent infection, the rate of "indolent infection" in our study was 12% (15 of 127). Although the retrospective definition of indolent infection is imperfect, the similarity to the median 8% rate of indolent infection from a recent review supports both our definition and the notion that a significant minority of HIV-infected patients with pulmonary M. kansasii isolates may not require immediate treatment (1). Based on our data, it may be appropriate to consider a conservative approach of observation in the setting of negative smear microscopy, only one or two cultures positive for M. kansasii and the lack of extrapulmonary disease in an HIV-infected patient. Additional factors supporting a conservative approach might include relatively high CD4 count, low viral load, and the presence of HAART. We must stress, however, that given the association between mortality and the lack of adequate treatment and the finding that the majority of our patients were treated, our results support treating the majority of HIV-infected patients with pulmonary M. kansasii. isolates. In addition, we were unable to address whether there had been progression of disease in the untreated long-term survivors, and the relatively large number of patients lost to follow-up makes the retrospective identification of patients with indolent infection even more difficult.

ATS criteria were not developed to aid in mortality prediction in HIV, rather to assist in the decision regarding initiating therapy. In this regard, it is desirable to test their ability to distinguish between active disease and indolent infection. This assessment is hindered by the lack of a gold standard for the presence of disease or indolent infection and the difficulty in retrospectively applying a surrogate empiric gold standard. We think that the ATS criteria have a high positive predictive value regarding the presence of disease, as they require the repeated isolation of an organism in the setting of radiographic abnormalities and symptoms. However, whether the absence of adequate criteria is an accurate definition of indolent infection is less certain and may vary depending on the species of nontuberculous mycobacteria and host factors. In our cohort, ATS criteria did not seem to be a critical factor in the decision to initiate therapy. Although significantly more patients who fulfilled ATS criteria received at least three drugs for at least 3 months, a significant fraction of patients who did not fulfill ATS criteria were treated for M. kansasii. In patients who did not meet ATS criteria, 33% received at least two drugs for at least 3 months, and 19% received at least three drugs for at least 3 months. Using the clinically perceived need for therapy as the definition of "disease," our observations suggest that a significant number of patients who do not fulfill ATS criteria actually have disease. Using ATS criteria to define indolent infection would therefore likely misclassify some patients with clinically important disease as having indolent infection. Although our study was not designed to address these issues specifically, we think that the threshold for initiating treatment should be lower in HIV-infected patients (especially with advanced immune suppression) than HIV-negative patients and lower for M. kansasii isolates than for isolates of less pathogenic nontuberculous mycobacteria species.

Similar to previous studies in this area, our study was limited by its retrospective design. Two key limitations resulting from a retrospective design include an increased risk of confounding and missing data. Confounding may have occurred if clinicians were more likely to treat certain patients with cither antiretrovirals or antimycobacterials depending on clinical variables not included in our analyses. Our secondary analyses, including interactions between several key variables and antimycobacterial treatment, found results consistent with the primary analysis. This suggests that confounding did not play a major role in our results. Missing data result in fewer patients included in multi-variable models, generally increasing the risk for both type one and type two errors. We limited this problem by eliminating some incomplete variables from the analysis and selectively imputing data for others. Furthermore, choosing a full model for the primary analysis (without automated selection) greatly decreases the number of statistical tests applied and lowers the risk of spurious results (22). Although we were able to address potential statistical pitfalls associated with missing data, the possibility remains that more complete data would have permitted the development of more informative models. Wc were unable to explore variables such as clinical stage of HIV infection, duration of exposure to HAART, and treatment effects on viral load and CD4 lymphocyte count in our analyses due to data limitations. In addition, constraining the number of variables in the models precluded assessment of other prognostic variables such as hemoglobin and serum albumin. Missing data also contribute to the difficully in determining the cause of death. We could attribute only a small proportion of deaths with certainty to M. kansasii. Although this may signify that M. kansasii infection does not increase mortality in HIV, we speculate that many of the deaths that were not directly attributed to M. kansasii may have been, at least in part, due to this infection. We attributed deaths to causes other than M. kansasii infection if they were present at the time of death, which would tend to underestimate the impact of M. kansasii infection on survival (see online supplement to METHODS). For this reason, we analyzed our data using all-cause mortality as the primary outcome. Our results are also most generalizable to patients with relatively low CD4 lymphocyte counts and high viral loads, given these characteristics in our patients. Despite the weaknesses of retrospective studies, which generally limit results to hypothesis generation, the absence of prospective and controlled data makes the results of our study and others important to consider in patient management.

In summary, we observed that in HIV-infected patients with pulmonary M. kansasii infection, a high CD4 lymphocyte count, negative sputum smear microscopy, and adequate treatment of M. kansasii were all associated with survival. ATS diagnostic criteria were not useful in predicting mortality. Although we did identify patients who apparently had indolent infection, this was a small minority, comprising a subgroup of patients who did not fulfill ATS criteria. Withholding treatment in HIV-infected patients with respiratory M. kansasii isolates should only be considered in the setting of negative smear microscopy, few positive cultures, relatively mild immune suppression, and the presence of HAART.

References

1. Marras TK, Daley CL. Systematic review of the clinical significance of pulmonary Mycobacterium kansasii isolates in human immunodeficiency virus infection. J Acquir Immune Defic Symlr (In press)

2. Campo RE, Campo CE. Mycobacterium kansasii disease in patients infected with human immunodeficiency virus. Clin Infect Dis 1997;24:1233-1238.

3. Carpenter JL, Parks JM. Mycobacterium kansasii infections in patients positive for human immunodeficiency virus. Rev Infect Dis 1991;13: 789-796.

4. Rooney G, Nelson MR, Gazzard B. Mycobacterium kansasii: its presentation, treatment and outcome in HIV infected patients. J Clin Pathol 1996;49:821-823.

5. Santin M, Alcaide F. Mycobacterium kansasii disease among patients infected with human immunodeficiency virus type 1: improved prognosis in the era of highly active antiretroviral therapy. Int J Tuberc Lung Dis 2003;7:673-677.

6. Marras TK, Daley CL. Epidemiology of human pulmonary infection with nontuberculous mycobacteria. Clin Chest Med 2002;23:553-567.

7. Levine B, Chaisson RE. Mycobacterium kansasii: a cause of treatable pulmonary disease associated with advanced human immunodeficiency virus (HIV) infection. Ann Intern Med 1991;114:861-868.

8. Valainis GT, Cardona LM, Greer DL. The spectrum of Mycobacterium kansasii disease associated with HIV-I infected patients. J Acquir Immune Defic Syndr 1991;4:516-520.

9. Witzig RS, Fazal BA, Mera RM, Mushatt DM, Dejace PMJT, Greer DL, Hyslop NE Jr. Clinical manifestations and implications of coinfection with Mycobacterium kansasii and human immunodeficiency virus type 1. Clin Infect Dis 1995;21:77-85.

10. Corbett EL, Blumberg L, Churchyard GJ, Moloi M, Mallory K, Clayton T, Williams BG, Chaisson RE, Hayes RJ, DeCock KM. Nontuberculous mycobacteria: defining disease in a prospective cohort of South African miners. Am J Respir Crit Care Med 1999;160:15-21.

11. American Thoracic Society. Diagnosis and treatment of disease caused by nontuberculous mycobacteria. Am J Respir Crit Care Med 1997;156: S1-S25.

12. Spradling P, Drociuk D, McLaughlin S, Lee LM, Peloquin CA, Gallicano K, Pozsik C, Onorato I, Castro KG, Ridzon R. Drug-drug interactions in inmates treated for human immunodeficiency virus and Mycobacterium tuberculosis infection or disease: an institutional tuberculosis outbreak. Clin Infect Dis 2002;35:1106-1112.

13. Burman WJ, Gallicano K, Peloquin C. Therapeutic implications of drug interactions in the treatment of human immunodeficiency virus-related tuberculosis. Clin Infect Dis 1999;28:419-429.

14. Klein JL, Corbett EL, Slade PM, Miller RF, Coker RJ. Mycobacterium kansasii and human immunodeficiency virus co-infection in London. J Infect 1998;37:252-259.

15. Lortholary O, Deniel F, Boudon P, LePennec MP, Mathieu M, Soilleux M, LePendeven C, Loiseau P, Vincent V, Valeyre D. Mycobacterium kansasii infection in a Paris suburb: comparison of disease presentation and outcome according to human immunodeficiency virus status. Int J Tuberc Lung Dis 1999;3:68-73.

16. Pintado V, Gomez-Mampaso E, Martin-Davila P, Cobo J, Navas E, Quereda C, Fortun J, Guerrero A. Mycobacterium kansasii infection in patients infected with the human immunodeficiency virus. Eur J Clin Microbiol Infect Dis 1999;18:582-586.

17. Marras TK, Gonzalez LC, Daley CL. Pulmonary Mycobacterium kansasii infection in human immunodeficiency virus infected individuals: diagnostic criteria and treatment in predicting survival [abstract]. Am J Respir Crit Care Med 2003;167:A708.

18. Pcduzzi P, Concato J, Feinstein AR, Holford TR. Importance of events per independent variable in proportional hazards regression analysis: II: accuracy and precision of regression estimates. J Clin Epidemiol 1995;48:1503-1510.

19. Bamberger DM, Driks MR, Gupta MR, O'Connor MC, Jost PM, Neihart RE, McKinsey DS, Moore LA. Mycobacterium kansasii among patients infected with human immunodeficiency virus in Kansas City. Clin Infect Dis 1994;18:395-400.

20. Corbett EL, Churchyard GJ, Hay M, Herselman P, Clayton T, Williams B, Hayes R, Mulder D, De Cock KM. The impact of HIV infection on Mycobacterium kansasii disease in South African gold miners. Am J Respir Crit Care Med 1999;160:10-14.

21. Corbett EL, Churchyard GJ, clayton TC, Williams BG, Mulder D, Hayes RJ, De Cock KM. HIV infection and silicosis: the impact of two potent risk factors on the incidence of mycobacterial disease in South African miners. AIDS 2000;14:2759-2768.

22. Steyerberg EW, Eijkemans MJ. Harrell FE Jr, Habbema JD. Prognostic modeling with logistic regression analysis: a comparison of selection and estimation methods in small data sets. Slat Med 2000;19:1059-1079.

Theodore K. Marras, Alison Morris, Leah C. Gonzalez, and Charles L Daley

Department of Medicine (Respirology), University of Toronto, Toronto, Ontario, Canada; Department of Medicine (Pulmonary and Critical Care), University of Southern California, Los Angeles; and Department of Medicine (Pulmonary and Critical Care), University of California, San Francisco, San Francisco, California

(Received in original form February 5, 2004; accepted in final form June 22, 2004)

Supported by National Institutes of Health grant number M01RR00083-41.

Correspondence and requests for reprints should be addressed to Theodore K. Marras, M.D., Toronto Western Hospital, EC4-022, 399 Bathurst Street, Toronto, ON, MST 2S8 Canada. E-mail: ted.marras@utoronto.ca

This article has an online supplement, which is accessible from this issue's table of contents online at www.atsjournals.org

Am J Respir Crit Care Med Vol 170. pp 793-798, 2004

Originally Published in Press as DOI: 10.1164/rccm.200402-162OC on June 23, 2004

Internet address: www.atsjournals.org

Conflict of Interest Statement: T.K.M. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; A.M. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; L.C.G. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; C.L.D. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript.

Copyright American Thoracic Society Oct 1, 2004
Provided by ProQuest Information and Learning Company. All rights Reserved