HIV-infected adults (>15 years of age), who sought treatment at a government health center in the township of Ndirande, Blantyre, Malawi, were enrolled in a prospective, community-based study to determine the incidence of infections that were preventable by trimethoprim-sulfamethoxazole prophylaxis (8). Clinical evaluations were performed monthly and at sick visits occurring between the scheduled monthly evaluations. CD4 counts were determined every 4 months. Standardized diagnostic and treatment guidelines and case definitions were used. At the time of the study, in Malawi, antiretroviral therapy (ART) was rarely used, and trimethoprim-sulfamethoxazole prophylaxis was not recommended.

Cases of suspected PcP were identified by patients’ clinical signs and symptoms, chest x-ray results, oxygen desaturation exercise test results (9), CD4 count, and failure to improve with antimicrobial treatment without activity against P. jirovecii. Patients’ sputum production was induced by an ultrasonic nebulizer with hypertonic saline, followed by IF staining for P. jirovecii cysts. A case was classified as clinical PcP when the IF staining for P. jirovecii cysts was positive or the participant had strong clinical evidence of PcP and negative IF. Clinical follow-up data were collected after the episode of suspected PcP.

After the study, real-time PCR for the P. jirovecii dihydropteroate synthase and human RNAase P (control DNA) was performed on DNA extracted from the stored induced sputum slides (10). Clinicians were not aware of the PCR results during the study, and laboratory staff performing the PCR was blinded to clinical information and IF results. A final diagnosis of confirmed PcP was made for any episode with a positive IF result, positive PCR result, or both, unless recovery (defined as resolution of respiratory symptoms present at the start of the episode) without PcP treatment was observed with a minimum of 4 weeks of follow-up. If the PCR results were positive but the patient recovered without active treatment against PcP, the result was interpreted as Pneumocystis colonization. A negative PCR result ruled out PcP diagnosis in patients who had received PcP treatment on the basis of clinical evidence alone.

Incidence rates of respiratory diagnoses per 100 person-years of follow-up were calculated with 95% confidence intervals (CIs) based on Poisson distribution. First and subsequent episodes in the same person were counted separately, except for PcP, because patients with PcP received secondary prophylaxis and exited the study. The CD4 count at the time of the episode or within the previous 6 months was used for analysis.

We used χ², Mann-Whitney, and Student t tests for analysis of age, sex, and CD4 counts among diagnoses, respectively, using SPSS version 12 software (SPSS Inc., Chicago, IL, USA). The study was approved by the Institutional Review Boards of the University of Malawi College of Medicine, the University of Maryland, and Michigan State University.

Beginning in September 2002, 660 adults were enrolled in the study and followed up through August 2004. Baseline CD4 and World Health Organization stage data are shown in Table 1. Mean age was 31.7 years (range 16–66); 437 (66%) were female. Mean duration of follow up was 10.7 months (95% CI 10.4–11.5) per person. Eighty-six (13%) participants died, and 37 (6%) were withdrawn from the study because they started lifelong trimethoprim-sulfamethoxazole prophylaxis. Sixty-three participants (9.5%) left the area, 20 (3%) withdrew consent, and 119 (17%) were lost to follow up. A smaller proportion of patients from the low CD4 stratum exited the study than from the high CD4 content group.

Ninety-five episodes of suspected PcP occurred in 75 persons. Outcomes of these episodes are given in Table 2. A final diagnosis of confirmed PcP was made in 6 episodes, and 9 episodes of Pneumocystis colonization were recorded, with a mean follow up of 26 weeks (range 4–48 weeks). Table 3 shows the incidence rates of PcP and other respiratory conditions in the cohort.

With full diagnostic workup including posttreatment follow up as the gold standard for the diagnosis of PcP, the sensitivity of PCR alone was 100%, the specificity 88%, and the positive predictive value 31%. Among episodes in which PcP was suspected, the mean CD4 count in patients with confirmed PcP cases (42.5 cells/mm³, range 1–103) was not significantly lower than in those with Pneumocystis colonization (89.1 cells/mm³, range 7–194; p = 0.28), but was significantly lower than in those with other diagnoses (97.0 cells/mm³, range 1–311; p = 0.03). Mean age and sex distribution of confirmed PcP, Pneumocystis colonization, and other diagnoses were not significantly different. The case-fatality rate of confirmed PcP was 50%.

This is the first community-based prospective study of PcP in a developing country. We found an incidence of PcP in Malawian HIV-infected adults of 1.0/100 person-years, similar to the rates observed in studies that used less comprehensive diagnostic approaches in South African miners (0.5/100 person-years) (12) and the placebo arms of trials of trimethoprim-sulfamethoxazole prophylaxis in Côte d’Ivoire (13,14). The incidence in persons with CD4 counts <200/mm³ (2.5/100 person-years) was clearly lower than in AIDS patients in the United States before the introduction of routine trimethoprim-sulfamethoxazole prophylaxis and highly active ART (10/100 person-years [11]). In the lowest CD4 count range (<100/mm³), PcP was common, although the incidence was low compared with that of bacterial pneumonia and pulmonary tuberculosis.

We believe it is unlikely that we missed many PcP cases among other diagnoses or losses to follow-up because of the intensive active and passive follow-up and because our facility provided expeditious, high-quality care free of charge. Allowing for reduced sensitivity of induced sputum compared to BAL (7) and considering cases with diagnostic uncertainty as PcP cases would still leave the PcP incidence low in the HIV-infected population in general.

We found that Pneumocystis colonization and confirmed PcP were equally common among patients with suspected PcP. More sensitive molecular detection methods would possibly have detected higher rates of colonization. It remains uncertain why certain HIV-infected persons clear Pneumocystis colonization while others develop PcP. The level of immune suppression as indicated by the CD4 count is a possible explanation, although our data do not support this. Genetic differences between P. jirovecii strains may be relevant (15). Variation in worldwide distribution of strains, as well as differences in host genetics and shorter survival in patients with a low range of CD4 counts, are possible causes of the lower PcP incidence in Africa than in developed countries.

The incidence of PcP in HIV-infected Malawian adults, diagnosed clinically and confirmed with molecular analysis, was low compared with the incidence of bacterial pneumonia and pulmonary tuberculosis at all levels of immunosuppression. PcP rarely occurred with CD4 cell counts >100 mm³. Among the most immunocompromised patients, PcP is an important diagnostic consideration.