As standard clinical practice, HIV-positive patients who came to San Francisco General Hospital with respiratory signs and symptoms compatible with pneumocystosis were evaluated by a uniform protocol that has been described previously (21). This protocol included obtaining specimens by induced sputum and, if necessary, bronchoscopy with bronchoalveolar lavage. Microscopic examination and cultures were used to establish a specific etiologic diagnosis. Consecutive patients undergoing sputum induction or bronchoscopy to diagnose PCP were enrolled in this study and provided written, informed consent to allow their medical records to be abstracted with a standardized data form. Study investigators classified patients as either PCP positive or PCP negative (controls), according to predetermined definitions that were blinded to serologic results. Pneumocystis patients were those patients with a microscopically confirmed diagnosis of P. jirovecii; these patients were treated with standard anti-Pneumocystis drugs as part of their regular medical care. Control patients were those whose microscopic examinations were negative for P. jirovecii, had Pneumocystis treatment discontinued, and recovered from acute pneumonia.

The study was conducted during a 4.5-year period (May 2000 through September 2004). During the first half of the study (2000–2002), an acute-phase serum specimen was drawn at the time of hospital admission for pneumonia, and a single convalescent-phase specimen was drawn at different intervals 5–12 weeks later. Preliminary analysis suggested that the Pneumocystis patients experienced a rise in mean serum antibody levels, whereas controls did not. Thus, during the later part of the study (2003–2004), additional serial convalescent-phase serum specimens were drawn every 1–2 weeks for 6 weeks from patients with pneumocystosis to measure early changes in antibody levels. Serum specimens were stored at -70°C and shipped to the University of Cincinnati for analysis. University of California San Francisco and University of Cincinnati institutional review boards approved the protocol.

Serum antibody levels to MsgA, MsgB, and MsgC were measured in a blinded manner by an ELISA as previously described (14,17,18). All serum specimens and the standard reference serum were diluted 1:100 and tested in duplicate wells of a 96-well plate against the following reagents: recombinant Msg fragments, Escherichia coli extract expressing the pET vector without insert (vector control), tetanus toxoid (TT) (positive control), and phosphate-buffered saline (PBS) without antigen (negative control). As an additional negative control, PBS was substituted for the serum specimen. Plates were washed, horseradish peroxidase (HRP)–labeled goat anti-human immunoglobulin G was added, plates were washed again, and tetramethylbenzidine substrate was added. The reaction was stopped by adding 0.18 mol/L H₂SO₄, and the plates were read at a wavelength of 450 nm. The reference serum specimen, which was obtained from a single person and had known reactivity to Msg, was run on each day as another control. HRP-labeled S-protein was used as a positive control and to correct for antigen loading. During the early part of the study, patient and reference serum specimens were tested at 1:100, 1:500 and 1:2,500 dilutions. The best results were obtained with the 1:100 dilution, so this dilution was used for the remainder of the study. The reactivity of each serum specimen to Msg was expressed as the ratio of reactivity to the pET vector: (mean optical density [OD] Msg_{test serum} – mean OD PBS_{test serum}) / (mean OD pET_{test serum} ^__ mean OD PBS_{test serum}).

Sex and racial distributions of Pneumocystis patients and controls were compared by using χ² tests of equality of proportions. Means and standard deviations were calculated to compare the center and spread of age and continuous parameters measured at enrollment (baseline). Mean values for Pneumocystis patients and controls were compared and tested for equality by using unpaired t tests with adjustment for unequal variances, either on the original or logarithmic scale. Geometric means and geometric standard deviations were calculated when data were approximately lognormally distributed, as judged by Shapiro-Wilks tests and visual assessment. Quantile-quantile plots of serum antibody levels showed that their distributions were approximately lognormal, except for larger numbers of values at the lower quantiles. Pearson product moment correlations among clinical parameters and Msg fragments were obtained. Analyses of trends in mean values of each outcome were performed for all patients enrolled at baseline during either study period and followed up at least once at weeks 1–2, 3–4, or 5–6. Sparse data at later follow-up weeks precluded their inclusion in the regression analyses because of instability of parameter estimates. In addition, the numbers of patients observed at baseline and each follow-up time were too small to provide meaningful inferences from analysis.

Two stratifications of Pneumocystis patients were investigated with respect to trends and mean levels of Msg fragments: patients with CD4+ counts above and below the median value (≈50 cells/μL) and patients with and without a history of pneumocystosis. The analysis provided comparisons between means for patient categories at each follow-up week, as well as differences in patterns of change over time between patient groups. We obtained p values that compared mean values between periods and levels of CD4+ count or history of Pneumocystis pneumonia by using t tests with Tukey's adjustment for multiple comparisons. Patient-specific deviations from group means were included in the analysis as a random effect, which allowed the correlation between repeated measurements on the same patient over time to be included in between- and within-patient variance estimates. A p value <0.05 (2-tailed) was considered significant, unless stated otherwise. Analyses were performed by using the SAS procedure PROC MIXED (SAS for Windows version 9.2, SAS Institute Inc., Cary, NC, USA)

No significant differences were seen between the 80 Pneumocystis patients and the 41 control patients with other causes of pneumonia in terms of sex, race, or age (Table 1). Baseline serum albumin levels, arterial blood gas measurements, and the proportion of patients who required mechanical ventilation were also similar, which indicates that the overall general health of patients and the severity of pneumonia in these 2 groups were also comparable. In addition, the proportion of patients with prior pneumocystosis was similar in both groups. However, the Pneumocystis patients had more advanced or more poorly controlled HIV infection than did the controls. The P. jirovecii patients had a significantly lower geometric mean CD4+ count and a significantly higher mean plasma HIV RNA level than did control patients. Patients with pneumocystosis also exhibited a significantly higher mean serum lactate dehydrogenase (LDH) level than did controls; this finding is consistent with the observation that elevated serum LDH is a nonspecific indicator of PCP.

At the time of hospital admission for pneumonia, geometric mean serum antibody levels to MsgC in Pneumocystis patients and controls were similar (Table 1). Forty-one of 80 patients with PCP had >1 convalescent-phase serum specimen drawn in the first 6 weeks after hospital admission. The total number of patient visits was 62. Patients observed at weeks 3–4 (n = 19) had a higher mean serum antibody level to MsgC than the average level of all patients at baseline (n = 80) and patients observed at weeks 1–2 (n = 25) and weeks 5–6 (n = 18). Differences were significant, as determined by t statistics comparing group means, adjusted for paired comparisons (p<0.01 to p<0.001) (Table 2). Mean antibody levels at subsequent time points were 3.7 at 5–6 weeks, 4.0 at 7–8 weeks, and 3.0 at 9–12 weeks (data not shown).

Analysis of serum antibody levels in individual patients showed different patterns of reactivity (Figure). Eleven (58%) of the 19 patients studied at 3–4 weeks had an increase in their antibody levels, ranging from 1.4- to 22-fold above baseline levels. To determine if the rise in serum antibodies in the Pneumocystis patients at 3–4 weeks was specific for P. jirovecii or part of a broader increase in antibody reactivity, we examined the changes in antibody levels to TT in these 19 patients. The geometric mean antibody levels of 80 U at baseline and 117 U at 3–4 weeks were not significantly different.

We examined clinical characteristics associated with an increase in antibody levels. No relationship was found between antibody level and sex, race, age, HIV viral load, serum albumin level, oxygenation, LDH level, or use of mechanical ventilation. CD4+ cell count (analyzed as CD4+ count <50 cells/μL or >50 cells/μL) was significantly associated with the ability to generate an increase in antibody levels (Table 2). Patients with CD4+ counts <50 cells/μL exhibited no significant changes in antibody levels to MsgC at any time point. In contrast, patients with CD4+ counts >50 cells/μL had a rise in mean antibody level, which peaked at 30.4 at 3–4 weeks and which was significantly higher than antibody levels at baseline (p<0.001) and at 5–6 weeks (p<0.001). The mean antibody level at 3–4 weeks in these patients was also significantly higher than the corresponding level in patients with <50 CD4+ cells (30.4 vs. 6.3, p<0.03).

The lack of a history of pneumocystosis also seemed to influence antibody response but not baseline antibody level. The mean antibody level to MsgC in the 59 patients with first episode of PCP was 4.1; this value was not significantly different from the mean value of 4.2 in the 19 patients with recurrent pneumocystosis (Table 3). Sequential antibody responses were compared in PCP patients whose episode of pneumonia was their first experience with the disease versus those whose pneumonia was a recurrent bout of the disease. Antibody levels in naive patients increased and reached a peak mean value of 17.9 at 3–4 weeks compared to 4.1 at baseline (p<0.001), whereas patients with a history of P. jirovecii pneumonia experienced no increase. Because the number of patients was low, we could not analyze the combined effects of PCP history and CD4+ cell count.

The baseline geometric mean serum antibody level to MsgA of 5.6 in PCP patients was not significantly different from the mean level of 3.7 in the controls (Table 1). Patients with pneumocystosis exhibited a different pattern of antibody response to MsgA than to MsgC in that the mean peak antibody levels to MsgA at 1–2 weeks and 3–4 weeks were similar (Table 4). However, none of the differences in antibody levels at different time points reached significance. In addition, no significant differences were found in the antibody levels at different time points in patients with and without a history of pneumocystosis or in patients with CD4+ cell counts <50 cells/μL or patients with CD4+ counts >50 cells/μL (data not shown).

No significant difference in baseline geometric mean serum antibody levels to MsgB was seen in patients with PCP and controls (Table 1). No significant differences were seen in antibody levels related to different time points, CD4+ counts, or history of pneumocystosis (data not shown).