Since the initial descriptions of Pneumocystis carinii pneu- cases per 100 person-years of follow time (incidence density

Transcription

1 1126 Risk Factors for Primary Pneumocystis carinii Pneumonia in Human Immunodeficiency Virus Infected Adolescents and Adults in the United States: Reassessment of Indications for Chemoprophylaxis Jonathan E. Kaplan, Debra L. Hanson, Thomas R. Navin, and Jeffrey L. Jones Division of HIV/AIDS Prevention Surveillance and Epidemiology, National Center for AIDS, STD, and TB Prevention; Division of Parasitic Diseases, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia Risk factors for the development of a first episode of Pneumocystis carinii pneumonia (PCP) were investigated in the Adult and Adolescent Spectrum of Disease Project, a medical record review study involving longitudinal follow-up of human immunodeficiency virus infected adults in 9 US cities. Risk factors included decreasing CD4 lymphocyte count and history of AIDS-defining illness, non P. carinii pneumonia, oral thrush, or unexplained fever for 2 days; PCP prophylaxis was protective. PCP incidence/100 person-years of observation among persons not prescribed PCP prophylaxis was higher in those with CD4 lymphocyte counts õ250 cells/ml or CD4 cell percent õ14% (8.3; 95% confidence interval [CI], ) than in persons with CD4 cell counts õ200 or history of thrush or fever, which constitute current criteria for prophylaxis against PCP (5.9; 95% CI, ). Because of increased efficiency in capturing persons at highest risk, CD4 cell count õ250 or CD4 cell percent õ14% should be considered as criteria for prophylaxis against first episodes of PCP. Since the initial descriptions of Pneumocystis carinii pneu- cases per 100 person-years of follow time (incidence density monia (PCP) in patients with AIDS in 1981 [1], PCP has [ID]) in persons who had various combinations of PCP risk remained the most common serious opportunistic infection in factors but who had not been offered PCP prophylaxis. adults and adolescents infected with human immunodeficiency virus (HIV) in the United States [2]. Chemoprophylaxis against PCP has been recommended since 1989 [3] and is currently Methods recommended for HIV-infected adolescents and adults with CD4 lymphocyte counts õ200 cells/ml, history of oral candidi- ASD project. The methods of the ASD Project have been deasis, history of unexplained fever for 2 weeks, or history of scribed previously [5]. Briefly, beginning in 1990, HIV-infected PCP [4]. The drug of choice is trimethoprim-sulfamethoxazole; persons 13 years of age were enrolled in 116 medical facilities in dapsone and aerosolized pentamidine are alternatives for those 11 US cities. Because of incomplete information on opportunistic who cannot tolerate trimethoprim-sulfamethoxazole. However, illnesses from 2 cities, data from 108 facilities in 9 cities are some cases of PCP occur in persons in medical care who have included in this analysis (Atlanta, New Orleans, Dallas, San Anto- not been offered PCP prophylaxis because they lack any of nio, Houston, Detroit, Denver, Los Angeles, and Seattle). When these risk factors. the patients are enrolled, medical records are reviewed for demo- graphic information, HIV exposure mode, and history of AIDS- We evaluated risk factors for the development of a first defining illnesses; information about other medical conditions, episode of PCP in HIV-infected adults and adolescents in care treatments, and laboratory tests are abstracted for the 12 months in the Adult and Adolescent Spectrum of Disease (ASD) Projbefore enrollment. These other conditions include oral candidiasis, ect, a longitudinal medical record review surveillance study of unexplained fever for 2 days, and pneumonia other than PCP. HIV-infected persons 13 years of age in ú90 medical care Laboratory tests include CD4 lymphocyte counts. Medical records facilities in 9 cities in the United States [5]. To determine are reviewed every 6 months thereafter until the patient s death or whether the current criteria for PCP prophylaxis could be imthe Centers for Disease Control and Prevention (CDC) through loss to follow-up. These analyses were done on data obtained at proved, we then computed and compared the number of PCP February 1997, at which time nearly 35,000 persons had been enrolled in the 9 cities. Definitions of persons in care, PCP prophylaxis, and diagnosis of PCP. Because subjects may be diagnosed with an AIDS- Received 12 January 1998; revised 15 May Presented in part: 4th Conference on Retroviruses and Opportunistic Infections, January 1997, Washington, DC (abstract 290). fore a CD4 lymphocyte count is obtained, we limited our analysis defining illness, including PCP, before enrollment in ASD or be- Reprints or correspondence: Dr. Jonathan E. Kaplan, Division of HIV/AIDS to subjects who had no history of PCP at the time of enrollment Prevention, Mailstop G-29, Centers for Disease Control and Prevention, Atand who were in care, as defined by enrollment in the study lanta, GA (jxk2@cdc.gov). and measurement of a CD4 lymphocyte count. Follow-up time The Journal of Infectious Diseases 1998;178: was calculated from the month in which both of these criteria were This article is in the public domain satisfied.

2 JID 1998;178 (October) Risk Factors for P. carinii Pneumonia 1127 For antimicrobial therapy, the reviewer recorded whether treat- injecting drug users or persons of unidentified HIV risk but ment was prescribed during each 6- to 12-month period under were similar with regard to race, sex, and age to those included review; neither the dates of initiation and termination of therapy in the analysis. PCP developed in 2625 (11.6%) of the 22,698 nor the reason for the prescription was recorded. Therefore, for HIV-infected persons with a CD4 lymphocyte count and 1 each interval of observation before the interval in which PCP month of follow-up who were included in the proportional occurred (and for each interval of observation for those in whom hazard analysis. Clinically related risk factors for PCP included the disease did not develop), PCP prophylaxis was defined as lower CD4 lymphocyte count (increasing risk ratios with deprescribed treatment with trimethoprim-sulfamethoxazole, dapcreasing CD4 cell count), and history of AIDS-defining illness, sone, or aerosolized pentamidine. Diagnoses of PCP were recorded as confirmed or presumptive, pneumonia other than PCP, oral candidiasis, and unexplained according to the 1993 CDC AIDS surveillance case definition fever for 2 days (table 1). The risk ratios for history of AIDS- [6]. Both confirmed and presumptive cases were included in the defining illness, non P. carinii pneumonia, and unexplained analyses; differences in risk factors associated with both combined fever were among persons with CD4 cell counts 200. and separate (confirmed vs. presumptive) diagnoses were evaluated. PCP prophylaxis was protective (table 1). The effect of antiret- roviral therapy could not be assessed because of a strong associ- Evaluation of potential risk factors for first episode of PCP. ation between antiretroviral therapy and PCP prophylaxis (data Potential risk factors for the occurrence of PCP included CD4 not shown). lymphocyte count (the lowest count recorded in each interval), Of the 2625 cases of PCP, 803 (30.6%) occurred in persons PCP prophylaxis, and antiretroviral therapy (prescribed in an inter- while not prescribed prophylaxis; 1822 (69.4%) occurred in val before the diagnosis of PCP); history (any time) of oral candidipersons who had been prescribed prophylaxis. Risk factors for asis, unexplained fever for 2 days, AIDS-defining illness, or PCP were the same for persons prescribed PCP prophylaxis as non P. carinii pneumonia; age; sex; race; and HIV exposure for those who were not (data not shown). mode. Risk factors were evaluated by using the counting process formulation of the Cox multivariate proportional hazard regression Further analyses of the risk for PCP associated with specific model [7]. All risk factors other than demographic characteristics causes or categories of AIDS illnesses (bacterial, fungal, viral, and HIV exposure mode were modeled as time-dependent covarigroup of pathogens or conditions responsible for the association protozoal, neoplastic) did not reveal a specific pathogen or ates. Assumptions of proportional hazards were investigated by using between history of AIDS illnesses and PCP. A similar assessplots of weighted Schoenfeld residuals versus time [8]. Relationships ment of the association between history of non P. carinii pneu- between antecedent diagnosis of AIDS opportunistic infec- monia and PCP could not be made, because there were too tion, non P. carinii pneumonia, oral candidiasis, and unexplained few persons with a history of non P. carinii pneumonia for fever and different levels of CD4 lymphocyte counts as risk factors which the cause was specified. for PCP were studied by model stratification according to CD4 ID of PCP among persons not prescribed prophylaxis. This cell risk categories. Interaction effects between PCP prophylaxis, analysis included 16,655 persons who were observed without antiretroviral therapy, and HIV risk transmission category were PCP prophylaxis (total of 25,370 person-years of follow time evaluated similarly. while not prescribed prophylaxis). PCP developed in 803 per- Determination of ID of PCP in persons with PCP-associated sons (ID, 3.2 cases/100 person-years; 95% confidence interval risk factors who had not received PCP prophylaxis. The ID of PCP (i.e, cases of PCP per 100 person-years of follow time) was [CI], ). The ID of PCP was 11.5 cases/100 persondetermined in persons who had PCP-associated risk factors, as years (95% CI, ) in persons with CD4 cell counts specified in the multivariate regression analysis, among those who õ200; this rate was higher than among subjects with any other had not received PCP prophylaxis. Since CD4 lymphocyte count single risk factor evaluated (table 2). Among those with CD4 was the strongest predictor of PCP (see Results), we estimated the cell counts 200, the ID was highest (4.2; 95% CI, ) ID of PCP at several levels of absolute counts and percentage of in those with a CD4 cell percent õ14%; in those with CD4 CD4 cells. The ID of PCP was estimated stepwise for various cell count 200 and CD4 cell percent 14%, the IDs were combinations of risk factors: Combinations of two risk factors equally high in those with CD4 cell count õ250 (i.e., between included the risk factor with the highest ID, combinations of three 200 and 250: ID, 3.4; 95% CI, ) and those with history risk factors included the combination of two with the highest ID, of an AIDS-defining illness (ID also 3.4; 95% CI, ; and so on until the ID was less than that of the overall population. table 2). Among persons with CD4 cell counts õ250 or CD4 cell percent õ14%, the ID was (577/6936) Å 8.3 (95% Results CI, ); among those with any of these risks or history of AIDS illness or non P. carinii pneumonia, the ID was (635/ Risk factors for PCP. As of February 1997, 24,189 persons 8927) Å 7.1 (95% CI, ). History of oral thrush, without a history of PCP had been enrolled in ASD and had history of unexplained fever, CD4 cell count õ300, and CD4 1 month of follow-up; 22,698 of these had 1 month of cell percent õ18% were consistently associated with lower IDs follow-up after a CD4 cell count was available and were in- in these analyses (table 2). cluded in the analysis. The 1491 excluded on the basis of Because the current criteria for prophylaxis against the first insufficient CD4 cell data were nearly twice as likely to be episode of PCP include CD4 cell count õ200 or history of

3 1128 Kaplan et al. JID 1998;178 (October) Table 1. Risk factors for development of PCP, Cox proportional hazard analysis, Adult and Adolescent Spectrum of Disease Project, Risk ratio (95% Risk factor No. (%) confidence interval) P CD4 T lymphocyte count, cells/ml (31.0) 40.7 ( ) õ (17.5) 23.0 ( ) õ (26.3) 11.1 ( ) õ (38.2) 3.75 ( ) õ (31.8) Reference Antecedent illnesses AIDS-defining illness CD (61.9) 1.02 ( ).71 CD (29.0) 1.34 ( ) õ.001 CD (9.2) 1.92 ( ) õ.001 Pneumonia CD (18.6) 1.24 ( ).002 CD (12.5) 1.41 ( ) õ.001 CD (8.2) 1.94 ( ) õ.001 Oral thrush CD (68.0) 1.04 ( ).53 CD (49.2) 1.34 ( ) õ.001 CD (22.6) 1.41 ( ).01 Unexplained fever CD (49.2) 1.11 ( ).05 CD (35.3) 1.48 ( ) õ.001 CD (23.3) 1.83 ( ) õ.001 HIV exposure mode Male-male sex 12,055 (53.1) Reference Injecting drug use 3331 (14.7) 1.12 ( ).10 Male-male sex & injecting drug use 2593 (11.4) 1.16 ( ).02 Heterosexual contact 2021 (8.9) 0.83 ( ).06 Blood or blood products 465 (2.0) 0.79 ( ).11 Unidentified risk 2233 (9.8) 0.86 ( ).09 Race/ethnicity White 10,490 (46.2) Reference Black 9173 (40.4) 1.04 ( ).40 Hispanic 2744 (12.1) 1.04 ( ).50 Asian 171 (0.8) 0.67 ( ).17 Native American 119 (0.5) 0.87 ( ).64 Sex Male 18,598 (81.9) Reference Female 4109 (18.1) 1.07 ( ).35 Age at enrollment, 10-year increase* 1.00 ( ).91 PCP prophylaxis 14,509 (63.9) 0.82 ( ) õ.001 NOTE. Analysis includes all subjects without prior history of PCP who were in care. Both confirmed and presumptive cases of PCP are included. Percentages in analyses that include CD4 (which refers to CD4 cell count) may not sum to 100. CD4 cell count is time-varying covariate; as such, no. represents persons who at some observation time were within that CD4 cell range. In CD4 cell count stratified analysis of persons with antecedent illness, no. represents persons who had antecedent illness and were observed in that CD4 cell range; % is proportion of all persons observed in that CD4 cell range. * Compared with persons 10 years younger at enrollment. oral candidiasis or unexplained fever, we also determined the ID of PCP in persons who had one or more of these risk factors but who had not been prescribed PCP prophylaxis. This combination of risk factors (i.e., criteria in current guidelines) was associated with a number of PCP cases similar to that associated with the combination of CD4 cell count õ250, CD4 cell percent õ14%, history of AIDS illness, or history of non P. carinii pneumonia (638 cases vs. 635 cases) but with an ID of 5.9 (vs. 7.1; 95% CI, and , respectively); that is, persons in this category were at lower risk for PCP (table 3). Since our analyses included cases of PCP diagnosed both definitively and presumptively, and since 1350 of the 2625 cases (51.4%) were diagnosed presumptively, we repeated our

4 JID 1998;178 (October) Risk Factors for P. carinii Pneumonia 1129 Table 2. PCP diagnoses, person-years of observation, and incidence density (cases/100 person-years) for patients not prescribed PCP prophylaxis. Incidence density PCP Person-years of (95% confidence Risk factor cases observation interval) Overall , ( ) Any risk factor , ( ) No risk factor ( ) CD4 õ200 cells/ml ( ) Hx of AIDS illness ( ) CD4% õ14% ( ) CD4 õ250 cells/ml ( ) CD4 õ300 cells/ml ( ) CD4% õ18% ( ) Hx of oral thrush ( ) Hx of non P. carinii pneumonia ( ) Hx of unexplained fever ( ) CD4 200 cells/ml but: CD4% õ14% ( ) CD4 õ250 cells/ml ( ) Hx of AIDS illness ( ) Hx of oral thrush ( ) CD4% õ18% ( ) CD4 õ300 cells/ml ( ) Hx of non P. carinii pneumonia ( ) Hx of unexplained fever ( ) CD4 200 cells/ml and CD4% 14% but: CD4 õ250 cells/ml ( ) Hx of AIDS illness ( ) Hx of non P. carinii pneumonia ( ) Hx of oral thrush ( ) CD4 õ300 cells/ml ( ) Hx of unexplained fever ( ) CD4% õ18% ( ) CD4 250 cells/ml and CD4% 14% but: Hx of AIDS illness ( ) Hx of non P. carinii pneumonia ( ) Hx of oral thrush ( ) Hx of unexplained fever ( ) CD4% õ18% ( ) CD4 õ300 cells/ml ( ) CD4 250 cells/ml, CD4% 14%, no hx of AIDS illness but: Hx of non P. carinii pneumonia ( ) Hx of unexplained fever ( ) Hx of oral thrush ( ) CD4% õ18% ( ) CD4 õ300 cells/ml ( ) CD4 250 cells/ml, CD4% 14%, no hx of AIDS illness or non P. carinii pneumonia but: Hx of unexplained fever ( ) Hx of oral thrush ( ) CD4 õ300 cells/ml ( ) CD4% õ18% ( ) NOTE. Hx Å history. PCP cases include both confirmed and presumptive cases. Patients observation time is computed from most recent CD4 (which refers to CD4 cell count) or diagnosis of risk factor. Data are ordered within groups by incidence density; risk factor with highest incidence density is listed first.

5 1130 Kaplan et al. JID 1998;178 (October) Table 3. PCP diagnoses, person-years of observation, and incidence risk factors for PCP and indications for prophylaxis in the large density (cases/100 person-years) for patients not prescribed PCP pro- database provided by the ASD Project. phylaxis in relation to current guideline for prophylaxis. In our analysis, low CD4 cell count was the strongest risk Incidence density factor for a first episode of PCP, a finding that is consistent PCP Person-years of (95% confidence with that in earlier studies. History of oral candidiasis and Risk factor cases observation interval) history of unexplained fever were also significant risk factors. History of (non-pcp) AIDS illness and non P. carinii pneu- Overall , ( ) monia conferred independent risks for the development of PCP; Any risk factor , ( ) No risk factor ( ) to our knowledge, these associations have not been described previously. History of an AIDS illness, despite survival after Current guideline* , ( ) the acute episode, seems to be associated with an accelerated Not current guideline, but: course of HIV disease [12, 13]; this observation may be ex- CD4% õ14% ( ) Hx of non P. carinii plained by enhanced HIV replication as a result of the infection pneumonia ( ) or by a premorbid condition (such as elevation of HIV plasma CD4 õ250 cells/ml ( ) RNA) that was responsible for both the earlier illness and Hx of AIDS illness ( ) the accelerated progression of HIV disease. The elevated risk CD4% õ18% ( ) conferred by non P. carinii pneumonia might be explained CD4 õ300 cells/ml ( ) by residual pulmonary pathology that in some unknown way NOTE. Hx Å history. PCP cases include both confirmed and presumptive increases the risk that an infection with P. carinii will become cases. Patients observation time is computed from most recent CD4 (which established (or predisposing pathology that increases the risk refers to CD4 cell count) or diagnosis of risk factor. * Cases diagnosed after CD4 cell count õ200 cells/ml or diagnosis of oral both for non P. carinii pneumonia and PCP). Some of these thrush or unexplained fever. cases of non P. carinii pneumonia might have been undiagnosed PCP. Our analysis of cases of PCP per 100 person-years of observation analyses with attention to the definitively diagnosed (conindicated (ID) during periods without prescribed PCP prophylaxis firmed) cases only. In the risk factor analysis, presumptive that for total and confirmed cases only, persons with cases were more likely than confirmed cases to be diagnosed a CD4 cell count õ250 or a CD4 cell percent õ14% had the at higher CD4 cell levels and to be diagnosed in blacks or in highest IDs for the development of PCP; the ID in this group those with heterosexual or unknown HIV risk; the beneficial exceeded those for other groups examined, including persons association with prophylaxis was stronger for confirmed than with a history of oral candidiasis or unexplained fever, which for presumptive cases, but the association with unexplained are recommended criteria for prophylaxis in current guidelines. fever was weaker (data not shown). In the ID analysis, the IDs In fact, in the overall analysis, among persons with CD4 cell for confirmed cases were roughly one-half of those shown in counts 250 and CD4 cell percent 14%, none of the IDs tables 2 and 3 (because of the exclusion of presumptive cases). for PCP exceeded that in the overall population (3.2%; 95% The ordering of the PCP risks was identical to that shown for CI, ); indeed, none exceeded 3.0%. These observations the combined analysis with regard to CD4 cell count õ200, suggest that this combination of risk factors may constitute the CD4 cell percent õ14%, and CD4 cell count õ250, but sub- best criteria for prophylaxis against the first episode of PCP. jects with CD4 cell counts õ300 and CD4 cell percent õ18% Our analysis also indicated that the combination of CD4 cell had higher IDs than did subjects with history of AIDS illness count õ250, CD4 cell percent õ14%, history of AIDS illness, or history of non P. carinii pneumonia (data not shown). or history of non P. carinii pneumonia preceded 635 of the PCP cases in the analysis, a number similar to that associated with the criteria in the current guidelines (638), but with a Discussion higher ID (7.1 vs. 5.9). Therefore, this combination could also be considered for prophylaxis, although in the analysis of confirmed Current indications for chemoprophylaxis against the first cases only, histories of AIDS illness and non P. carinii episode of PCP in HIV-infected adults and adolescents include pneumonia were not associated with the same conclusion. CD4 cell count õ200 or history of oral candidiasis or unexplained One difficulty in interpreting our data is the possibility that fever of 2 weeks duration [4]. Although low CD4 persons not prescribed PCP prophylaxis may be followed up cell count is a well-documented risk factor for PCP, the additional less frequently, thus having less frequent CD4 cell counts, and indications for PCP prophylaxis emanate from only two a greater likelihood that PCP will develop after a CD4 cell studies, to our knowledge [9, 10]. Providers continue to debate count 200 is observed. In our ID analysis, the median interval whether prophylaxis should be offered to other patients, such between measurement of CD4 cell count and PCP diagnosis as those with CD4 cell counts ú200 but õ250 or CD4 cell was 6 months (25% quartile, 2 months; 75% quartile, 14 percent õ14% [11]. Therefore, it seemed reasonable to reassess months). Therefore, it is possible that more timely measurement

6 JID 1998;178 (October) Risk Factors for P. carinii Pneumonia 1131 of CD4 cells would have decreased the risks we observed in we repeated our analyses limiting PCP cases to the 1275 (of persons with CD4 cell counts ) that were diagnosed definitively and found no significant The rate of decline of the CD4 cell count could conceivably changes in our results, with the exception of the ordering of contribute to assessment of risk in persons with CD4 cell counts the IDs for history of AIDS illness and history of non P We found that a decline of 15 CD4 cells/month was carinii pneumonia. Despite these limitations, ASD offers the an independent risk factor for PCP when this variable was opportunity to examine a large database that is diverse with entered into our multivariate analysis (risk ratio, 1.21; 95% CI, regard to age, sex, race, HIV exposure mode, and geography; ). However, the number of subjects with CD4 cell such a database cannot be found elsewhere in the United States. count 200 but õ250 who developed PCP and who had such In conclusion, our data suggest that (1) a CD4 cell count data available was too small to generate meaningful information õ250 or CD4 cell percent õ14% or (2) any one or more of regarding this variable in the ID analysis (data not shown). CD4 cell count õ250, CD4 cell percent õ14%, history of Measurement of HIV plasma RNA (virus load) may also AIDS illness, or history of non P. carinii pneumonia should prove useful in predicting risk of PCP; elevated levels of HIV be considered as the criteria for prophylaxis against the first RNA have been shown to constitute a risk of PCP and other episode of PCP. However, because of the limitations of this opportunistic infections independent of CD4 cell count [14]. study, such a policy change should await confirmation of our However, collection of HIV RNA data has been instituted only findings in other populations, as well as thorough cost-effecrecently in the ASD project and could not be assessed in this tiveness research to determine the threshold ID at which PCP analysis. prophylaxis should be offered. Another issue in interpreting our data is determining the Finally, there are many reasons that HIV-infected persons optimal ID threshold at which PCP prophylaxis should be offered. in the United States continue to develop PCP. As many as one- Providers would be unwilling to offer prophylaxis to half of all cases of PCP occur in persons who are unaware of all HIV-infected persons (a strategy that would prevent the their HIV infection or have not otherwise received medical maximum number of cases of PCP) because of the risk of drug care [18 20]. Others occur because of the provider s failure toxicity. Trimethoprim-sulfamethoxazole, the drug of choice to offer prophylaxis or poor patient adherence to prophylaxis. for PCP prophylaxis, causes significant toxicity in up to 40% Therefore, in addition to the appropriate use of PCP chemoprophylaxis, of HIV-infected patients [15], although recent data suggest the prevention of this disease will require counseling that a gradual introduction of trimethoprim-sulfamethoxazole, and testing for HIV infection, referral to medical care, and compared with introduction at the full dose of one doublestrength education of both patients and providers. tablet per day, may improve tolerance to the drug [16]. In addition to the risk of PCP, a true assessment of the threshold risk at which HIV-infected persons should be offered Acknowledgments prophylaxis would need to take into account the cost We gratefully acknowledge the Adult and Adolescent Spectrum of prophylaxis, the rate and consequences (including cost of of HIV Disease Investigators: Susan Buskin and Sharon G. Hopmanagement) of drug toxicity, and other factors, including the kins, King County Department of Public Health, Seattle; Frank additional benefits (such as protection against toxoplasmosis Sorvillo and Dorothy Masters, Los Angeles County Department and bacterial respiratory diseases) and the liabilities (such as of Health Services, Los Angeles; Susan Burkham and Sharon Melthe development of drug-resistant infections) of prophylaxis. ville, Texas Department of Health, Austin; Wes McNeely and Such studies are beyond the scope of this analysis but are Kaye Reynolds, Houston Department of Health and Human Services, underway elsewhere. Houston; Susan Troxler and Anne Morse, Louisiana Office Several limitations of this study are imposed by the ASD of Public Health, New Orleans; Melanie Thompson, AIDS Re- Project. ASD is an observational surveillance system for HIV- search Consortium of Atlanta, Atlanta; Judy Sackoff and Jeffrey related conditions and was not designed to address the efficacy McFarland, City of New York Department of Health, New York; Linda Wotring and Eve D. Mokotoff, Detroit Health Department, of chemoprophylaxis. Neither the exact dates, the reasons for Detroit; Arthur Davidson, David L. Cohn, and Cornelius Rietthe prescription of antibiotics or other drugs, nor adherence to meijer, Denver Department of Health and Hospitals, Denver; José filling prescriptions or taking medication are recorded. Hence, Otero, Robert Hunter, and Maria de los Angeles Gomez, University the risk ratio for PCP prophylaxis observed in the risk factor Central del Caribe, Bayamón, Puerto Rico; and Sandra Miranda, analysis (0.82) must be interpreted cautiously and cannot be Puerto Rico Department of Health, San Juan, Puerto Rico. We used to imply a lesser degree of efficacy of prophylaxis than also thank John Karon and R. J. Simonds for review of the manuhas been demonstrated in clinical trials [17]. All of the risk script and Sharon Cox for secretarial assistance. factors we investigated are determined by medical record review and are not routinely confirmed by other methods; unexplained fever is recorded for 2 days, and it is possible that References some undiagnosed (non P. carinii) pneumonias were cases of 1. Centers for Disease Control. Pneumocystis pneumonia Los Angeles. PCP. Presumed cases of PCP may be misdiagnosed, although MMWR Morb Mortal Wkly Rep 1981;30:250 2.

7 1132 Kaplan et al. JID 1998;178 (October) 2. Kaplan JE, Masur H, Holmes KK, et al. USPHS/IDSA guidelines for the 13. Chaisson RE, Keruly JC. Impact of opportunistic diseases on survival in prevention of opportunistic infections in persons infected with human HIV disease [abstract 366]. In: 4th Conference on Retroviruses and immunodeficiency virus: introduction. Clin Infect Dis 1995;21(suppl Opportunistic Infections: program and abstracts (Washington, DC). Al- 1):S1 11. exandria, VA: Infectious Diseases Society of America, 1997: Centers for Disease Control. Guidelines for prophylaxis against Pneumoand 14. Swindells S, Currier JS, Williams P. DACS 071: correlation of viral load cystis carinii pneumonia for persons infected with human immunodefion risk for opportunistic infection [abstract 359]. In: 4th Conference ciency virus. MMWR Morb Mortal Wkly Rep 1989;38(S-5):1 9. Retroviruses and Opportunistic Infections: program and abstracts 4. USPHS/IDSA Prevention of Opportunistic Infections Working Group. (Washington, DC). Alexandria, VA: Infectious Diseases Society of 1997 USPHS/IDSA guidelines for the prevention of opportunistic infec- America, 1997:130. tions in persons infected with human immunodeficiency virus. MMWR 15. Simonds RJ, Hughes WT, Feinberg J, Navin TR. Preventing Pneumocystis Morb Mortal Wkly Rep 1997;46(RR-12):1 46. carinii pneumonia in persons infected with human immunodeficiency 5. Farizo KM, Buehler JW, Chamberland ME, et al. Spectrum of disease in virus. Clin Infect Dis 1995;21(suppl 1):S44 8. persons with human immunodeficiency virus infection in the United 16. Para MF, Dohn M, Frame P, et al. ACTG 268 trial gradual initiation of States. JAMA 1992;267: trimethoprim/sulfamethoxazole (T/S) as primary prophylaxis for Pneu- 6. Centers for Disease Control and Prevention revised classification mocystis carinii pneumonia (PCP) [abstract 2]. In: 4th Conference on system for HIV infection and expanded surveillance case definition for Retroviruses and Opportunistic Infections: program and abstracts AIDS among adolescents and adults. MMWR Morb Mortal Wkly Rep (Washington, DC). Alexandria, VA: Infectious Diseases Society of America, 1997: ;41: Ionniadis JP, Cappelleri JC, Skolnik PR, Lau J, Sacks HS. A meta-analysis 7. Anderson PK, Borgan O, Gill RD, Keiding N. Statistical models based of the relative efficacy and toxicity of Pneumocystis carinii pneumonia on counting processes. New York: Springer-Verlag, prophylactic regimens. Arch Intern Med 1996;156: Grambsch PM, Therneau TM. Proportional hazards tests and diagnostics 18. Schwarcz SK, Katz MH, Hirozawa A, Gurley J, Lemp GF. Prevention of based on weighted residuals. Biometrics 1994;81: Pneumocystis carinii pneumonia: who are we missing? AIDS 1997;11: 9. Phair J, Munoz A, Detels R, et al. The risk of Pneumocystis carinii pneu monia among men infected with human immunodeficiency virus type 19. Shapiro J, Simon P. Late HIV diagnosis and failure to receive chemopro- 1. N Engl J Med 1990;322: phylaxis against Pneumocystis carinii pneumonia [abstract I93]. In: 10. Goedert JJ, Biggar RJ, Melbye M, et al. Effect of T4 count and cofactors Program and abstracts of the 36th Interscience Conference on Antimion the incidence of AIDS in homosexual men infected with human crobial Agents and Chemotherapy (New Orleans). Washington, DC: immunodeficiency virus. JAMA 1987;257: American Society for Microbiology, 1996: Chu SY, Hanson DL, Ciesielski C, Ward JW. Prophylaxis against Pneumo- 20. Duchin JS, Sohlberg B, Buskin S, Hopkins S, Simon P. Risk factors for cystis carinii pneumonia at higher CD4 / T-cell counts. JAMA 1995; Pneumocystis carinii pneumonia: delayed diagnosis of HIV infection 273:848. and failure to receive prophylactic therapy [abstract Tu.B.114]. In: Pro- 12. Whalen C, Horsburgh CR, Hom D, Lahart C, Simberkoff M, Ellner J. gram and abstracts: XI International Conference on AIDS (Vancouver, Accelerated course of human immunodeficiency virus infection after Canada). Vancouver: XI International Conference on AIDS Society, tuberculosis. Am J Respir Crit Care Med 1995;151: :230.