A convenience sample of 115 San Francisco General Hospital (San Francisco, CA, USA) health care workers was enrolled in the longitudinal study from January 2007 through February 2009. HIV/AIDS Division and Division of Pulmonary and Critical Care Medicine staff were sought preferentially because they worked most consistently with patients who were infected with HIV and/or PCP, the presumed reservoirs of P. jirovecii. Recruitment was conducted primarily by word of mouth; emails were sent to departmental listservs; and announcements were made at staff meetings, medical conferences, and orientations for medical students and residents. At study entry, no respondents reported PCP infection. Participants who provided serial (2–5) serum samples were included in this analysis; these persons are a subset of the participants included in our prior cross-sectional study (23). The University of California, San Francisco, and the University of Cincinnati Institutional Review Boards approved this study. All participants provided written, informed consent.

Participants completed an initial baseline questionnaire at enrollment and a follow-up questionnaire every 4–8 weeks. Questions about demographic characteristics and medical history (i.e., cigarette smoking, chronic lung disease, and immunocompromising condition) were asked on the initial questionnaire. The questions regarding chronic lung disease and immunocompromising condition were purposefully generalized, and specification of any condition was made optional to protect confidentiality regarding a participant’s medical history. On both the initial and follow-up questionnaires, participants were asked whether they had ever been exposed to a patient with PCP and, if so, when their most recent exposure occurred. Participants were asked to choose from the following possible answers: within 1 hour, within 24 hours, within 7 days, within 1 month, and >1 month ago.

Participants were placed in the following 2 categories: health care workers with patient contact and health care workers without patient contact, as described (23). Health care workers with patient contact (clinical staff) were persons who worked directly with patients in a clinical or research setting. Clinical staff included direct care providers, such as attending physicians, fellows, interns/residents, medical students, nurse practitioners, and nurses, as well as ancillary clinic staff, such as medical assistants, social workers, pharmacists, and clinical research assistants. Health care workers without patient contact (nonclinical staff) were persons who worked at the hospital but who did not work directly with patients in either a clinical or a research setting. Nonclinical staff included administrative and laboratory staff.

Serum specimens were collected at the time of enrollment and quarterly (equal to a median of 3.2 months after the previous visit [interquartile range 3.0–3.5 months]) for up to 1 year, or for as long as each health care worker participated in the study. From each participant, 2–5 serum specimens were collected. All serum specimens were stored at −80°C and subsequently sent to the University of Cincinnati (Cincinnati, OH, USA) for analysis. An ELISA that included recombinant fragments derived from a single P. jirovecii Msg isoform was used to measure IgG levels (22). IgG responses against MsgA, MsgB, MsgC1, MsgC3, MsgC8, and MsgC9 were measured. All serum specimens from the same participant and a standard reference serum specimen were placed in duplicate wells in a 96-well plate and were tested against all 6 Msg fragments. Phosphate-buffered saline without antigen was used as a negative control. Reactivity was corrected by subtracting the reactivity of the serum to the phosphate-buffered saline from the reactivity of the serum to the antigen, and results were quantified by the method of Bishop and Kovacs (24). A standard reference serum specimen with specificity for each Msg construct was prepared by mixing serum from 4–6 specimens with high reactivity for the specific construct. These specimens were selected from testing banks of serum specimens from HIV-infected patients and healthy blood donors. From this, a standard curve was generated for each Msg construct on each day the assay was used. This curve was used to calculate the units of reactivity to the Msg construct. For each standard serum pool, we assigned a value of 100 units of reactivity to its target Msg construct in 100 μL of a 1:100 dilution. Test serum samples were diluted at 1:100–1:200 to fit the linear portion of the standard curves. Taking into account the dilution, we then calculated units of reactivity. The lowest possible value of 1 U was assigned to all specimens with values below the standard curve.

Demographic, health, and professional characteristics were compared by occupation (clinical vs. nonclinical). Simple categorical variables (sex, ethnicity, health conditions, exposure to PCP-infected patients) were compared by using the χ² test. Fisher exact test was used when expected counts were <5. Multicategorical variables (race, work department) were compared by using logistic regression. Continuous variables (age) were compared by using the Student t test. Antibody levels were normalized by using a log transformation; results were exponentiated and presented as estimated geometric means (EGMs) with 95% CIs. Tobit mixed model regression for censored data was used to estimate the difference between antibody response in clinical staff and that in nonclinical staff. For a subset of workers who self-identified as having been exposed to a PCP-infected patient within 1 month before or after having a study serum specimen drawn, the changes in antibody levels from the time of exposure to 3 months and 6 months afterward were calculated and compared with changes from baseline to subsequent serum antibody levels in workers with no known P. jirovecii exposure. We compared antibody changes within each group using paired t tests and compared differences between the groups using a general linear model with 3-month or 6-month change as the dependent variable. Statistical significance was defined as p<0.05. All calculations were performed with SAS software 9.2 (SAS Institute Inc., Cary, NC, USA).

We enrolled 115 staff members, and each staff member provided at least 2 serum specimens. Participants ranged from 22 to 80 years of age (mean 39.5 years), and 66 (57.4%) were female (Table 1). Seventy (60.9%) participants were White/Caucasian, 30 (26.1%) were Asian, and 3 (2.6%) were Black/African American. Seventeen (14.8%) were ethnically Hispanic/Latino. Thirty-nine (33.9%) participants had smoked at least 100 cigarettes in their lifetime; 19 (16.5%) had an underlying lung condition; and 8 (7.0%) had an immunocompromising condition. Fifty-two (45.2%) participants were part of the HIV/AIDS Division, 30 (26.1%) were part of the Division of Pulmonary and Critical Care Medicine (CCM), 27 (23.5%) were part of the Department of Medicine, and 6 (5.2%) were members of other departments (Obstetrics and Gynecology, Psychiatry, and Radiology). Of the 115 participants, 79 (68.7%) had a known exposure to a PCP-infected patient before the study period.

Ninety-nine (86.1%) participants were categorized as clinical staff, and 16 (13.9%) were categorized as nonclinical staff (Table 1). No significant differences were found between clinical and nonclinical staff in age, sex, race, ethnicity, smoking habits, underlying lung condition, or department/division. However, a significantly greater proportion of nonclinical staff than clinical staff reported having an immunocompromising condition (25.0% vs. 4.0%; p = 0.01). As expected, a significantly greater proportion of clinical staff than nonclinical staff reported exposure to a PCP-infected patient (77.8% vs. 12.5%; p<0.001).

All staff had detectable antibodies against all Msg fragments: MsgA, MsgB, MsgC1, MsgC3, MsgC8, and MsgC9 (Table 2). For this study, the 99 clinical staff members provided a total of 396 serum specimens (mean 4 specimens), and the 16 nonclinical staff members provided 80 serum specimens (each provided 5 specimens). Overall, when levels of clinical and nonclinical staff were compared, clinical staff had significantly higher EGM antibody levels against MsgC1 (EGM 38.4 vs. 19.8; adjusted p = 0.004) and MsgC8 (EGM 46.0 vs. 27.6; adjusted p = 0.02), even when we controlled for age and immunocompromising condition. EGM antibody levels to MsgA, MsgB, MsgC3, and MsgC9 were not significantly different between clinical and nonclinical staff.

To evaluate the relationship between PCP exposure and antibody response over time, we identified participants who reported being exposed to a PCP-infected patient within 1 month before any serum collection, irrespective of their occupation (exposed, n = 37), and compared their antibody levels at the time of exposure with their antibody levels 3 and 6 months later. As a control comparison, baseline serum antibody levels of never PCP-exposed persons (never exposed, n = 20), were compared with their antibody levels 3 and 6 months later. Thirty-three of the 37 (89%) exposed participants had been exposed to PCP-infected patients repeatedly over the study period. Members of the PCP-exposed group showed no significant difference in EGM antibody levels after PCP exposure for any Msg variant over time (Figure 1, panels A–F). In contrast to the findings in the PCP-exposed group, EGM antibody levels against MsgC1 in the never PCP-exposed group declined significantly after 3 months (mean change –2.87, 95% CI –5.74 to –0.01; p = 0.049) (Figure 1, panel C). Declines in EGM antibody levels against MsgC3 and MsgC8 over 3 months were of borderline significance (mean change –7.26, 95% CI –14.7 to 0.18; p = 0.06; and mean change –4.30, 95% CI –8.73 to 0.13; p = 0.06; respectively) (Figure 1, panels D, E). However, no significant differences were found in MsgA (Figure 1, panel A), MsgB (Figure 1, panel B), or MsgC9 (Figure 1, panel F) during this time.

Mean changes in EGM antibody levels within the PCP-exposed group were then compared with mean changes in the never PCP-exposed group (Figure 2, panels A–F). No difference was found in EGM antibody levels at baseline between exposed (antibodies measured at the time of exposure) and never-exposed (antibodies measured at the time of baseline enrollment) participants. In contrast, the difference in mean change was significant after 3 months for MsgC1 (mean change 1.67 vs. –2.87; p = 0.04) (Figure 2, panel C), after 3 and 6 months for MsgC3 (mean change 4.09 vs. –7.26, p = 0.02 and 5.10 vs. –8.24, p = 0.03, respectively) (Figure 2, panel D), after 3 and 6 months for MsgC8 (mean change 2.29 vs. –4.30, p = 0.02 and 1.71 vs. –3.30, p = 0.048, respectively) (Figure 2, panel E), and after 6 months for MsgC9 (mean change 1.67 vs. –3.11, p = 0.03) (Figure 2, panel F). After we adjusted for age and an immunocompromising condition, mean change after 6 months in MsgC1 became significant (mean change 1.31 vs. −3.43, p = 0.02). However, mean changes in MsgC3 and MsgC8 lost statistical significance. In contrast, no significant differences were found between the exposed and never-exposed groups in mean change for MsgA (mean change after 3 months 1.50 vs. –1.46, p = 0.09; mean change after 6 months 0.04 vs. –2.51, p = 0.30, respectively) (Figure 2, panel A), MsgB (mean change after 3 months –0.10 vs. –0.10, p = 1.00; mean change after 6 months 0.02 vs. 0.10, p = 0.93; respectively) (Figure 2, panel B), or MsgC9 after 3 months (mean change 1.14 vs. –0.35, p = 0.43) (Figure 2, panel F).