Clinical Factors Associated With Pneumocystis Pneumonia Despite Its Primary Prophylaxis: When to Stop Prophylaxis?

Abstract

Objective

Although previous studies show that primary prophylaxis against Pneumocystis jirovecii pneumonia (PJP) is effective in patients with rheumatic diseases receiving immunosuppressive treatment, there is limited evidence regarding the optimal timing for prophylaxis withdrawal. This study aimed to identify the risk factors for PJP despite prophylaxis and provide evidence for an optimal prophylaxis schedule.

Methods

This case‐control study included 1,294 prophylactic episodes in 1,148 patients with rheumatic disease who received immunosuppressants and prophylactic trimethoprim‐sulfamethoxazole (TMP‐SMX). The primary outcome was a one‐year incidence of PJP. A Cox proportional hazards model with least absolute shrinkage and selection operator was used to evaluate clinical factors associated with outcomes.

Results

During 1,174 person‐years of observation, 10 cases of PJP were identified, with an incidence rate of 0.85 per 100 person‐years. The mean ± SD duration of TMP‐SMX prophylaxis was 181.9 ± 128.7 days. Except in one case, PJP occurred after discontinuation of TMP‐SMX, with a median (interquartile range [IQR]) interval of 117.0 (86.0–161.0) days. The dose of glucocorticoids at the time of TMP‐SMX discontinuation was significantly higher in the PJP group relative to the control group (median [IQR]: 22 [20–40] vs 10 [5–15] mg). Discontinuing TMP‐SMX while on a glucocorticoid dose >12.5 mg/day of prednisone equivalent significantly increased the risk of PJP (adjusted hazard ratio: 13.84; 95% confidence interval, 1.71–111.80). There were 63 cases of adverse events during the observation period, and 10 (15.9%) were attributed to TMP‐SMX with probable causality.

Conclusion

Tapering glucocorticoids with 12.5 mg/day of prednisone equivalent could be a reasonable timepoint to initiate the withdrawal of PJP prophylaxis in patients with rheumatic diseases.

INTRODUCTION

Pneumocystis jirovecii pneumonia (PJP) is a potentially life‐threatening opportunistic infection that occurs mainly in immunocompromised patients. ¹ Although first identified during the onset of the HIV epidemic in the early 1980s, ² PJP infection has become more common in non–HIV‐immunocompromised patients suffering from various conditions, such as malignancies, solid organ transplants, and autoimmune or inflammatory disease (AIID). ³ , ⁴ Furthermore, PJP in patients without HIV typically exhibits rapid progression with a higher mortality rate than in HIV patients. ³ , ⁵ , ⁶

Trimethoprim‐sulfamethoxazole (TMP‐SMX) is the gold standard for PJP prophylaxis, and its efficacy and safety have been demonstrated in randomized controlled trials (RCTs) in both HIV‐infected and non–HIV‐infected immunocompromised patients. ⁷ , ⁸ , ⁹ Although no RCTs have specifically addressed this issue in patients with AIID, many observational studies have shown that prophylactic TMP‐SMX treatment significantly decreases the incidence of PJP in these patients, particularly when they receive prolonged, high‐dose glucocorticoid (GC) treatment. ¹⁰ , ¹¹ , ¹² , ¹³ Based on these results, despite the weak level of evidence and the variation in risk depending on specific diseases and clinical settings, a recent recommendation from EULAR states that PJP prophylaxis should be taken into consideration when patients with AIID receive daily doses of prednisone exceeding 15 to 30 mg for longer than two to four weeks. ¹⁴

Despite the established benefits of primary PJP prophylaxis in patients with AIID, the optimal prophylaxis schedule remains uncertain. One of the most important uncertainties is the timing of discontinuation of PJP prophylaxis. TMP‐SMX carries some risk of adverse events (AEs), some of which may be severe. ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ Therefore, minimizing exposure to TMP‐SMX while preserving its optimal prophylactic efficacy is a logical next step in building evidence for the universal recommendation for primary PJP prophylaxis. In patients with HIV, the decision to discontinue primary PJP prophylaxis is guided by CD4 counts and response to antiretroviral therapy (ART). Guidelines recommend the cessation of prophylaxis when CD4 counts have risen to more than 200 cells/mm ³ and are maintained for at least three months in response to ART. ¹⁹ However, the correlation between CD4 counts and PJP risk is less clear in patients without HIV, including those with AIID, ²⁰ , ²¹ which renders CD4 counts a less reliable surrogate measure to direct prophylaxis. Although previous studies have suggested that PJP mainly occurs when a patient receives >15 mg/day of prednisone or its equivalent, it is uncertain whether this could be a relevant indication to consider the cessation of PJP prophylaxis. ²² , ²³ This study aimed to investigate the risk factors associated with PJP infection despite prophylaxis and to offer insights into establishing an ideal primary prophylaxis regimen for patients with AIID.

PATIENTS AND METHODS

Patients

This study included 1,148 patients who received immunosuppressants and concomitant prophylactic TMP‐SMX for the treatment of underlying AIID at a nationwide tertiary referral center in South Korea between January 2010 and December 2022. The patients received TMP‐SMX at either a dose of one single‐strength tablet daily or a double‐strength tablet three times per week. In cases with decreased renal function, TMP‐SMX dosing was adjusted according to the estimated glomerular filtration rate (eGFR). Patients were excluded if they had HIV infection, underwent solid organ transplantation, were younger than 18 years of age, or had a follow‐up duration of less than 28 days. In this study, none of the patients received prophylactic agents other than TMP‐SMX, such as dapsone or pentamidine (Supplementary Figure 1).

This study was conducted in accordance with the Declaration of Helsinki and was approved by the Seoul National University Hospital Institutional Review Board (no. 2009‐065‐1156). The ethics committee waived the requirement for written informed consent due to the retrospective nature of the study.

Study design and prophylaxis episodes

To identify potential risk factors associated with PJP occurrence despite primary prophylaxis, a case‐control study design was selected given the low expected incidence of PJP in the study population. As some patients may have undergone multiple courses of prophylactic TMP‐SMX along with immunosuppressive therapy, we examined each episode of prophylaxis (referred to as a prophylaxis episode) within the study population. Episodes in which prophylactic TMP‐SMX was administered for less than 14 days were excluded from the analysis of PJP incidence. However, AEs that occurred during these episodes were collected for safety assessment.

The observation scheme for the prophylactic episodes is presented in Supplementary Figure 2. The baseline date was defined as the date of prophylactic TMP‐SMX initiation. The index period was designated as four weeks before and after the baseline. The demographics and immunosuppressive agents prescribed during this period were defined as the baseline features of the study population. The observation began from the baseline date and continued until the date of loss to follow‐up, occurrence of PJP, or 52 weeks after the baseline date, whichever occurred first. Each prophylaxis was considered a distinct episode if the patient discontinued TMP‐SMX and restarted it at least 52 weeks after the baseline date. If a patient had more than one episode, time‐varying features such as eGFR, lymphocyte count, and immunosuppressive treatment use at the baseline of each episode were collected.

Data collection

The clinical data of the study population were obtained from the electronic medical records database of Seoul National University Hospital. Collected variables included patient demographics, underlying AIID, comorbidities (malignancy diagnosed within five years from the baseline, diabetes mellitus, chronic obstructive pulmonary disease, chronic liver disease, and heart failure), duration of TMP‐SMX prescription, and potential risk factors for PJP, such as baseline lymphopenia (lymphocyte <800/μL), azotemia (eGFR <60 mL/min), and interstitial lung disease. In addition, data on the prescribed immunosuppressive therapies (GC, rituximab, cyclophosphamide, mycophenolate mofetil, tacrolimus, azathioprine, JAK inhibitors, and tocilizumab) were collected. Concomitant medications were defined as medications prescribed during the index period. GC pulse therapy was defined as intravenous methylprednisolone administration at a dose of 250 to 1,000 mg/day for a duration of one to five days. To collect data on the underlying AIID, we first collected data on all registered International Classification of Disease, Tenth Revision (ICD‐10) codes for each study population. If a patient had more than one ICD‐10 code for rheumatic disease, one investigator (JYK) reviewed the patient's medical records to confirm the final diagnosis. We also reviewed the medical records and gathered information on AEs that occurred with the use of prophylactic TMP‐SMX. AEs were graded according to the Common Terminology Criteria for Adverse Events version 5.0. ²⁴ Severe AEs were defined as AEs of grade 3 or higher. Two independent investigators (JYK and JWP) assessed the severity and cause of TMP‐SMX in each case. Cases with probable or definite causality were defined as having drug reactions related to TMP‐SMX. ²⁵

PJP detection

The detection of PJP is based on a combination of clinical, radiologic, and laboratory criteria, which are confirmed based on (1) the presence of clinical and radiographic features suggestive of PJP, and (2) results of the microbiologic test for identification. A two‐step algorithm was designed to capture all PJP cases without misclassification. ¹² At the outset, we collected all cases that exhibited positive results from direct immunofluorescence staining and/or polymerase chain reaction assays of respiratory specimens, including sputum, induced sputum, and bronchoalveolar lavage, over the observation period. Next, all the medical records of the selected cases were reviewed by two expert investigators (JYK and JWP) to confirm the diagnosis of PJP. Finally, PJP cases were confirmed when both assessors reached a consensus on the diagnosis.

Statistical analysis

All clinical variables used for the analysis had <5% missing values; therefore, no imputation was performed in the main analysis. For continuous variables, Student's t‐test was used when the normality assumption was satisfied according to the Shapiro test, whereas the Wilcoxon rank‐sum test was applied when it was not. Fisher's exact test was used for categorical variables. A Cox proportional hazards with least absolute shrinkage and selection operator (LASSO) model was used to identify the risk factors associated with PJP. The LASSO method was employed to prevent overfitting and to identify the most relevant predictive variables among independent clinical factors that exhibited a significant association (P < 0.1) in the univariable analysis. Due to multicollinearity among the variables related to decreased renal function (eGFR, azotemia, and hemodialysis status), azotemia was selected for the main multivariable analysis. The chosen variables were used to fit a multivariable Cox proportional hazards regression model, adjusted for intracluster correlation to account for patients with multiple prophylaxis episodes. Firth penalized regression model was used in case of complete separation. Finally, clinical factors that showed a statistically significant association in the multivariable Cox regression analysis with LASSO were identified as risk factors for PJP despite primary prophylaxis.

To investigate the optimal timing for withdrawing PJP prophylaxis, we used the maximally selected rank statistics method to identify the optimal cutoff dose of GC at the time of prophylaxis discontinuation. ²⁶ Subsequently, the association between the determined cutoff dose and cumulative incidence of PJP was evaluated using the Cox proportional hazards model, in which the hazard ratio (HR) was adjusted for the risk factors selected in the final multivariable analysis.

Several sensitivity analyses were conducted. First, the optimal cutoff dose of GC at the time of prophylaxis discontinuation was re‐evaluated by excluding patients with recent concomitant malignancies diagnosed within the five years before the baseline date (n = 193). Second, the categorical variable “azotemia” was replaced with the continuous variable eGFR, and the analysis was repeated. Third, multiple imputation was performed using the Multivariate Imputation by Chained Equations (MICE) package in R. Following imputation, a Cox proportional hazards regression model was fitted to the imputed data sets, using the variables selected in the final model in the main analysis. The results from the Cox regression models were pooled using Rubin's rules to obtain combined estimates. Finally, the random survival forest (RSF) model was employed to assess the importance of each clinical variable using the variable importance (VIMP) method. All statistical analyses were performed using R V.3.3.1 software, and P < 0.05 was considered statistically significant.

RESULTS

Baseline characteristics

A total of 1,294 prophylactic episodes were analyzed in this study (Supplementary Figure 1). The most common underlying AIID was systemic lupus erythematosus (n = 273, 21.1%), followed by antineutrophil cytoplasmic antibody–associated vasculitis (AAV) (n = 226, 17.5%) and idiopathic inflammatory myositis (n = 204, 15.8%). The mean ± SD duration of TMP‐SMX prophylaxis was 181.9 ± 128.7 days. In most episodes (n = 1,287, 99.5%), patients received TMP‐SMX concomitant with GC, with a mean daily dose of 41.5 ± 24.2 mg/day of prednisone or equivalent at the baseline. The most frequently used concomitant immunosuppressant was cyclophosphamide (n = 258, 19.9%), followed by rituximab (n = 168, 13.0%), mycophenolate mofetil (n = 120, 9.3%), azathioprine (n = 114, 8.8%), tacrolimus (n = 64, 4.9%), and cyclosporine A (n = 63, 4.9%). Rituximab was used solely for induction therapy. A combination of two or more non‐GC immunosuppressants was administered to 56 (4.3%) patients.

PJP incidence in the study population

During the 1,174 person‐years (PY) of observation, 10 cases of PJP occurred, with a one‐year incidence rate (per 100 PY) of 0.85 (95% confidence interval [CI]: 0.41–1.57). In all cases, patients were evaluated by infection specialists and confirmed PJP. The clinical features of the PJP cases are summarized in Supplementary Table 1. Among these cases, only one occurred as a breakthrough infection during TMP‐SMX prophylaxis, whereas the others occurred after the discontinuation of TMP‐SMX. In these cases, the median (interquartile range [IQR]) interval between stopping prophylaxis and PJP was 117.0 (86.0–161.0) days. The median (IQR) duration of prophylaxis was 43.0 (22.0–124.0) days, where prophylaxis was stopped due to either AEs (n = 3) or the treating physician's discretion (n = 6).

The baseline characteristics of the PJP and control groups are presented in Table 1. The PJP group tended to be older, had more lymphopenia and compromised renal function, and had a shorter prophylactic period for TMP‐SMX (median [IQR]: 148.0 [68.0–350.0] vs 51.0 [22.0–124.0] days; P = 0.007). PJP group had a greater proportion of patients who received renal replacement therapy. AAV constituted a significantly higher proportion of the PJP group than other underlying diseases. Regarding concomitant immunosuppressive treatment, the GC dose at baseline was comparable between the two groups. However, the GC dose at the time of discontinuation of TMP‐SMX was significantly higher in the PJP group (median [IQR]: 22.5 [20.0–40.0] vs 10.0 [5.0–15.0] mg of prednisone; P < 0.001) (Figure 1).

Table 1.

Clinical features of the study population^*

	Control group (n = 1,284)	PJP group (n = 10)	P value
Age, median (IQR), y	55.0 (41.0–66.0)	70.5 (65.0–76.0)	0.001
Female sex, n (%)	798 (62.1)	6 (60.0)	1.000
Smoking, n (%)			0.240
Never‐smoker	842 (78.2)	6 (60.0)
Ever‐smoker	235 (21.8)	4 (40.0)
Underlying rheumatic disease, n (%)
SLE	272 (21.2)	1 (10.0)	0.698
AAV	221 (17.2)	5 (50.0)	0.018
IIM	203 (15.8)	1 (10.0)	1.000
IgA vasculitis/nephropathy	185 (14.4)	0 (0.0)	0.374
IgG4‐related disease	64 (5.0)	0 (0.0)	1.000
AOSD	45 (3.5)	0 (0.0)	1.000
RA	41 (3.2)	1 (10.0)	0.282
Vasculitides, unclassified	40 (3.1)	0 (0.0)	1.000
Polyarteritis nodosa	38 (3.0)	0 (0.0)	1.000
Behcet's disease	37 (2.9)	1 (10.0)	0.259
Systemic sclerosis	34 (2.6)	0 (0.0)	1.000
Sjögren disease	21 (1.6)	0 (0.0)	1.000
Large vessel vasculitis	15 (1.2)	0 (0.0)	1.000
Relapsing polychondritis	15 (1.2)	0 (0.0)	1.000
Polymyalgia rheumatica	12 (0.9)	1 (10.0)	0.096
Cryoglobulinemic vasculitis	5 (0.4)	0 (0.0)	1.000
Cutaneous rheumatic disease	11 (0.9)	0 (0.0)	1.000
MCTD/UCTD	8 (0.6)	0 (0.0)	1.000
Antiphospholipid syndrome	5 (0.4)	0 (0.0)	1.000
Others	19 (1.5)	0 (0.0)	1.000
Comorbidities, n (%)
Malignancy	192 (15.0)	1 (10.0)	1.000
Diabetes mellitus	359 (28.0)	6 (60.0)	0.035
COPD	156 (12.1)	2 (20.0)	0.350
Interstitial lung disease	334 (26.0)	4 (40.0)	0.298
Heart failure	77 (6.0)	2 (20.0)	0.120
Chronic liver disease	130 (10.1)	0 (0.0)	0.611
Azotemia a	400 (31.2)	9 (90.0)	<0.001
Renal replacement therapy	118 (9.2)	4 (40.0)	0.010
Duration of TMP‐SMX, median (IQR), d	148.0 (68.0–350.0)	51.0 (22.0–124.0)	0.007
eGFR, median (IQR)	85.7 (48.6–108.7)	21.6 (11.0–46.5)	<0.001
Lymphocyte count, median (IQR)	1,280.0 (808.0–1,993.0)	571.0 (329.0–1,682.0)	0.037
Lymphopenia, b n (%)	294 (23.5)	7 (70.0)	0.003
Treatment at baseline
Glucocorticoid pulse, n (%)	136 (10.6)	3 (30.0)	0.083
Baseline GC dose, c median (IQR), mg/day	40.0 (25.0–60.0)	50.0 (25.0–62.5)	0.431
Rituximab, n (%)	163 (12.7)	5 (50.0)	0.005
Cyclophosphamide, n (%)	255 (19.9)	3 (30.0)	0.427
Mycophenolate mofetil, n (%)	119 (9.3)	1 (10.0)	1.000
Cyclosporine, n (%)	63 (4.9)	0 (0.0)	1.000
Tacrolimus, n (%)	64 (5.0)	0 (0.0)	1.000
Azathioprine, n (%)	113 (8.8)	1 (10.0)	0.604
JAK inhibitor, n (%)	4 (0.3)	0 (0.0)	1.000
Tocilizumab, n (%)	20 (1.6)	0 (0.0)	1.000
Combination non‐GC immunosuppressive treatment, n (%)	55 (4.3)	1 (10.0)	0.359
GC dose c at the end of prophylaxis, median (IQR), mg/day	10.0 (5.0–15.0)	22.5 (20.0–40.0)	<0.001

^*AAV, antineutrophil cytoplasmic antibody–associated vasculitis; AOSD, adult‐onset Still disease; COPD, chronic obstructive pulmonary disease; eGFR, estimated glomerular filtration rate; GC, glucocorticoid; IIM, idiopathic inflammatory myopathies; IQR, interquartile range; MCTD, mixed connective tissue disease; PJP, Pneumocystis jirovecii pneumonia; RA, rheumatoid arthritis; SLE, systemic lupus erythematosus; TMP‐SMX, trimethoprim‐sulfamethoxazole; UCTD, undifferentiated connective tissue disease.

^a Defined as a glomerular filtration rate <60 mL/minute.

^b Defined as <800 lymphocytes per microliter.

^c Based on the dose of prednisone.

Figure 1.

Density plot showing the dose of GC at the time of trimethoprim‐sulfamethoxazole discontinuation in the PJP and the control group. GC, glucocorticoid; PJP, Pneumocystis jirovecii pneumonia. Color figure can be viewed in the online issue, which is available at http://onlinelibrary.wiley.com/doi/10.1002/art.43167/abstract.

Risk factors for PJP

Univariable Cox regression analysis indicated that age, decreased renal function, lymphopenia, diabetes, underlying AAV, polymyalgia rheumatica, and the concomitant use of GC pulse therapy or rituximab were associated with the occurrence of PJP (Table 2). In contrast, baseline GC dose (HR, 1.01; 95% CI, 0.99–1.03) and combination of two or more non‐GC immunosuppressants (HR, 2.56; 95% CI, 0.32–20.21) did not increase the risk. Rituximab was administered as maintenance therapy in 95 episodes (7.3%), all within the control group. Rituximab maintenance was not associated with an increased incidence of PJP in this cohort (HR, 0.56; 95% CI, 0.00–4.33). Among the relevant factors discussed, age, impaired renal function, lymphopenia, and the use of rituximab emerged as significant predictors of increased PJP risk in the multivariable analysis conducted with LASSO.

Table 2.

Clinical factors associated with PJP despite primary prophylaxis^*

	Univariable		Multivariable a
	HR (95% CI)	P value	HR (95% CI)	P value
Age	1.08 (1.03–1.14)	<0.01	1.06 (1.00–1.13)	0.04
Sex, female	0.88 (0.25–3.11)	0.84
eGFR	0.96 (0.94–0.99)	<0.01
Azotemia b	20.01 (2.53–157.93)	<0.01	9.73 (1.17–80.74)	0.04
Hemodialysis	6.77 (1.91–23.99)	<0.01
Lymphopenia c	7.97 (2.06–30.82)	<0.01	5.79 (1.44–23.28)	0.01
Cancer	0.70 (0.09–5.50)	0.73
Diabetes mellitus	3.70 (1.05–13.12)	0.04	1.29 (0.32–5.11)	0.72
Heart failure	3.90 (0.83–18.35)	0.09	3.57 (0.68–18.73)	0.13
COPD	1.83 (0.39–8.62)	0.44
Interstitial lung disease	1.88 (0.53–6.68)	0.33
SLE	0.40 (0.05–3.19)	0.39
AAV	4.75 (1.37–16.41)	0.01
IIM	0.59 (0.07–4.64)	0.62
RA	3.75 (0.47–29.57)	0.21
Behçet's disease	3.54 (0.45–27.95)	0.23
Polymyalgia rheumatica	11.18 (1.42–88.31)	0.02	9.29 (0.89–97.35)	0.06
Rituximab	6.75 (1.96–23.33)	<0.01	5.53 (1.37–22.29)	0.02
Cyclophosphamide	1.74 (0.45–6.72)	0.42
Mycophenolate mofetil	1.08 (0.14–8.52)	0.94
Azathioprine	1.16 (0.15–9.12)	0.89
JAK inhibitor	15.99 (0.12–124.16)	0.18
Tocilizumab	2.81 (0.02–21.75)	0.54
Combination non‐GC immunosuppressive treatment	2.56 (0.32–20.21)	0.37
Baseline GC dose	1.01 (0.99–1.03)	0.44
Baseline GC pulse	3.71 (0.96–14.36)	0.06

^*AAV, antineutrophil cytoplasmic antibody–associated vasculitis; CI, confidence interval; COPD, chronic obstructive pulmonary disease; eGFR, estimated glomerular filtration rate; GC, glucocorticoid; HR, hazard ratio; IIM, idiopathic inflammatory myopathies; PJP, Pneumocystis jirovecii pneumonia; RA, rheumatoid arthritis; SLE, systemic lupus erythematosus.

^a Least absolute shrinkage and selection operator analysis was used to select clinical variables that constitute the optimal predictive model.

^b Defined as a glomerular filtration rate below 60 mL/minute.

^c Defined as <800 lymphocytes/μL per microliter.

Impact of GC dose at discontinuing TMP‐SMX on the incidence of PJP

Because the GC dose at TMP‐SMX discontinuation was significantly higher in the PJP group, we further evaluated its impact on PJP occurrence. Using maximally selected rank statistics, the optimal cutoff GC dose that best distinguished the PJP from the control group was identified as 12.5 mg prednisone equivalent (P < 0.001). In fact, 90% (9 of 10) of the patients who developed PJP in the prophylaxis group were taking >12.5 mg/day of prednisone or its equivalent dosage at the time of prophylaxis withdrawal, compared with only 35.1% (n = 451) in the control group. When the entire prophylactic episodes were divided into two groups based on the GC dose at the end of prophylaxis (>12.5 mg vs ≤12.5 mg of prednisone equivalent), the episodes with higher GC dose (n = 460, 35.5%) at the time of prophylaxis withdrawal showed a significantly higher risk of PJP (HR, 18.55; 95% CI, 2.35–146.42; P = 0.006) (Figure 2). The result was consistent in the multivariable Cox regression analysis in which relevant risk factors for PJP were adjusted (adjusted HR, 13.84; 95% CI, 1.71–111.80; P = 0.014) (Table 3).

Figure 2.

Kaplan–Meier curve indicating the Pneumocystis jirovecii pneumonia incidence stratified by the dose of GC at the time of discontinuing TMP‐SMX. GC, glucocorticoid; TMP‐SMX, trimethoprim‐sulfamethoxazole.

Table 3.

Effect of GC dose at the end of prophylaxis on 1‐year Pneumocystis jirovecii pneumonia incidence^*

	Univariable, HR (95% CI)	Multivariable, adjusted HR a (95% CI)
GC >12.5 mg	18.55 (2.35–146.42)	13.84 (1.71–111.80)
P value for HR	0.0056	0.0137

^*CI, confidence interval; eGFR, estimated glomerular filtration rate; GC, glucocorticoid; HR, hazard ratio.

^a Adjusted for clinical factors with relevant association (P < 0.05) in multivariable analysis (age, azotemia defined as eGFR below 60 mL/minute, lymphopenia, rituximab).

Safety of prophylactic TMP‐SMX

There were 63 AEs that occurred in 54 episodes (4.2%): 50 in the control group and 4 in the PJP group, with an incidence of 5.1 per 100 PY (95% CI, 3.9–6.5). The most frequent AEs were liver enzyme abnormalities (n = 13, 20.6%), rashes (n = 11, 17.5%), thrombocytopenia (n = 10, 15.9%), fever (n = 7, 11.1%), and renal dysfunction (n = 6, 9.5%) (Table 4). Most AEs were mild to moderate in severity and did not require additional interventions. The causality of each TMP‐SMX–related AE was assessed, with 10 cases (15.9%) rated as “probable,” 37 (58.7%) rated as “possible,” and 16 (25.4%) rated as “unlikely.” However, in most cases (n = 60, 95.2%), the TMP‐SMX treatment was discontinued after the occurrence of AEs.

Table 4.

Adverse events that occurred during the prophylactic TMP‐SMX administration^*

	Causality, n (%)			Total AEs
	Probable	Possible	Unlikely
Total AEs	10 (100.0)	37 (100.0)	16 (100.0)	63 (100.0)
Mild to moderate AEs	8 (80.0)	36 (97.3)	15 (93.8)	59 (93.6)
AST or ALT increased	1 (10.0)	10 (27.0)	2 (12.5)	13 (20.6)
Rash	3 (30.0)	5 (13.5)	3 (18.8)	11 (17.5)
Thrombocytopenia	0	6 (16.2)	4 (25.0)	10 (15.9)
Leukopenia	0	2 (5.4)	2 (12.5)	4 (6.3)
Fever	1 (10.0)	4 (10.8)	2 (12.5)	7 (11.1)
Renal dysfunction	2 (20.0)	3 (8.1)	1 (6.3)	6 (9.5)
Pneumonitis	1 (10.0)	0	0	1 (1.6)
Headache	0	2 (5.4)	0	2 (3.2)
Allergic reaction	0	1 (2.7)	0	1 (1.6)
Nausea	0	1 (2.7)	1 (6.3)	2 (3.2)
Tremor	0	1 (2.7)	0	1 (1.6)
Cognitive disturbance	0	1 (2.7)	0	1 (1.6)
Severe AE	2 (20.0)	1 (2.7)	1 (6.3)	4 (6.3)
Pancytopenia	0	1 (2.7)	1 (6.3)	2 (3.2)
Stevens‐Johnson syndrome	2 (20.0)	0	0	2 (3.2)

^*No AEs were evaluated as definite in causality assessment. AE, adverse event; ALT, alanine aminotransferase; AST, aspartate aminotransferase; TMP‐SMX, trimethoprim‐sulfamethoxazole.

During the observation period, there were four cases of severe AEs, including two cases of pancytopenia and two cases of Stevens‐Johnson syndrome (SJS). Two SJS cases were deemed to have probable causality, whereas the remaining cases had possible causality. All patients with SJS recovered after discontinuation of TMP‐SMX.

Sensitivity analysis

Multiple sensitivity analyses were performed. Initially, we replicated the analysis after excluding individuals who had been diagnosed with cancer within the five years before the baseline date (n = 193). This analysis led to the same result that a GC dose greater than 12.5 mg was significantly associated with the occurrence of PJP (adjusted HR, 15.10; 95% CI, 1.88–120.95) (Supplementary Table 2). Secondly, we replaced covariable “azotemia” with “eGFR” and conducted a repeated analysis, which also showed the consistent increased risk of PJP with the GC >12.5 mg at the end of prophylaxis (adjusted HR, 14.09; 95% CI, 1.75–113.56) (Supplementary Table 3). Third, multivariable Cox regression was fit on the imputed data sets obtained after multiple imputation. The pooled results indicated that the effect of GC >12.5 mg of prednisone equivalent at the end of prophylaxis was consistent with the findings from the original analysis (adjusted HR, 13.84; 95% CI, 1.70–112.69). Lastly, the RSF analysis was conducted to identify the variables of importance that were associated with PJP development according to the VIMP method. The importance of variables was ranked in the order of the GC dose at the end of prophylaxis, lymphopenia, rituximab, eGFR, and age (Supplementary Figure 3).

DISCUSSION

Although the effectiveness of PJP prophylaxis in patients with AIIDs is widely recognized, there is no clear consensus on the optimal regimen, particularly regarding the duration of maintenance. Although previous studies have demonstrated the excellent prophylactic efficacy of TMP‐SMX in reducing the incidence of PJP, its use carries a potential risk of various AEs. Therefore, establishing appropriate criteria for discontinuing prophylaxis could enhance patient outcomes while improving safety. To the best of our knowledge, this is the first large‐scale study to investigate the potential risk factors associated with PJP despite prophylaxis, which may provide crucial evidence for establishing an optimal prophylaxis protocol.

This study showed that the one‐year incidence of PJP among patients receiving prophylactic TMP‐SMX was low, reinforcing previous reports that TMP‐SMX has high prophylactic efficacy against PJP. ⁹ Furthermore, there was only one case of breakthrough PJP during TMP‐SMX prophylaxis. Given the high effectiveness of TMP‐SMX in PJP prophylaxis, further research is needed to address next steps, such as identifying specific risk groups and determining the optimal duration for maintaining prophylaxis, to ensure effective prevention.

As in line with previous studies, risk factors such as lymphopenia, impaired renal function, age, and rituximab use were significantly associated with PJP. ¹¹ , ¹³ , ¹⁸ , ²³ , ²⁷ , ²⁸ , ²⁹ Our earlier research demonstrated that TMP‐SMX prophylaxis reduces the incidence of PJP and related mortality in patients treated with rituximab. ³⁰ An intriguing finding in this study was that among the 16 patients who developed PJP despite prophylaxis, 15 experienced the onset of PJP merely a few months following the cessation of prophylactic TMP‐SMX.

In this context, it is noteworthy that the PJP group in our study had a shorter prophylaxis duration than the control group, leading to the discontinuation of prophylaxis while the patients were still receiving relatively higher doses of GCs. Previous studies on PJP in patients with AIID showed that most PJP cases occurred when patients received >15–20 mg/day of prednisone or its equivalents. ¹¹ , ¹² , ³¹ Expanding upon this finding, we demonstrated that the risk of PJP persists even with prophylaxis, unless it is continued until GC doses are tapered to ≤12.5 mg/day of prednisone or its equivalent. In fact, patients who stopped prophylactic TMP‐SMX at a dose of concomitant GC higher than the threshold showed an approximately 10‐times‐higher risk of PJP than other patients. This result suggests that tapering the GC dose by 12.5 mg/day or less can be a relevant threshold for considering the withdrawal of prophylactic TMP‐SMX.

However, approximately 40% of the study population discontinued TMP‐SMX before the threshold was reached. This result could be attributed to various factors, including safety concerns associated with TMP‐SMX. Despite the well‐established prophylactic efficacy of TMP‐SMX, there are concerns regarding adverse drug reactions (ADRs) ranging from severe conditions such as SJS and toxic epidermal necrolysis to minor reactions such as gastrointestinal symptoms, rash, increased serum creatinine, and elevated liver enzymes. ³² , ³³ , ³⁴ , ³⁵ The study by Nettleton et al calculated the risk of possible AE associated to PJP prophylaxis and reported that leukopenia (HR, 3.1), rash (HR, 1.9), and nephropathy (HR, 2.6) were higher among those who received prophylaxis as opposed those who did not. ³⁶ However, most of these studies did not thoroughly assess the causality of AEs, leaving the true incidence of ADRs directly related to prophylactic TMP‐SMX debatable. It should be noted that AEs commonly linked to TMP‐SMX may also stem from underlying diseases and concomitant treatment. This aligns with our finding that only 4.2% of prophylactic episodes were suspected to be AEs. Among these AEs, 10 (15.9%) were identified as true ADR, and only two cases showed serious AEs that required intervention other than stopping TMP‐SMX, in which all patients recovered after proper management. Notably, a previous cohort study that included patients treated with rituximab and prophylactic TMP‐SMX showed that, among 2,113 AEs, only 92 (4.4%) had definite or probable causality related to TMP‐SMX. Of these, only 10 severe ADRs required any intervention. ³⁰ Taken together, these findings suggest that the potential risks associated with the prolonged use of prophylactic TMP‐SMX may be less significant than those previously reported.

The study has several limitations that should be addressed. First, due to the small number of PJP cases in this study, including even a few covariates in the multivariable model could potentially result in model overfitting. Although variable selection using LASSO and RSF analysis helped reduce this potential bias, the results of this study should be validated in future larger‐scale studies. Second, the prophylactic effect of TMP‐SMX may vary among individual AIIDs. However, this study could not thoroughly address this variability due to the limited number of patients in each specific disease subgroup. Furthermore, the risk of PJP is not driven by a single factor alone but rather by a complex interplay of multiple interacting factors. Although this study suggests that GC dose could serve as a relevant and easily applicable threshold for considering prophylaxis withdrawal, the decision should be further guided by more patient‐specific factors, which need to be further investigated in future studies. Third, because this study retrospectively collected data from medical records, it is possible that some potential AEs related to TMP‐SMX were missed in our analysis. Finally, the results of this study may have been affected by unmeasured factors, an inherent limitation of observational studies. An RCT would provide a clear understanding of this issue. However, considering the extremely low incidence of PJP in the study population, the sample size required to test the hypothesis with an adequate statistical power would not be feasible. As an alternative, further comparative observational studies may offer a viable option for generating additional evidence.

In conclusion, discontinuation of TMP‐SMX during GC treatment at doses >12.5 mg prednisone equivalent significantly increased the risk of PJP. This suggests that tapering the GC to 12.5 mg/day may be a reasonable consideration when withdrawing PJP prophylaxis. Although these results need to be confirmed in future studies, they may assist physicians in determining the appropriate timing for discontinuing PJP prophylaxis in patients with AIIDs.

AUTHOR CONTRIBUTIONS

All authors contributed to at least one of the following manuscript preparation roles: conceptualization AND/OR methodology, software, investigation, formal analysis, data curation, visualization, and validation AND drafting or reviewing/editing the final draft. As corresponding author, Dr Park confirms that all authors have provided the final approval of the version to be published, and takes responsibility for the affirmations regarding article submission (eg, not under consideration by another journal), the integrity of the data presented, and the statements regarding compliance with institutional review board/Declaration of Helsinki requirements.

Supporting information

Disclosure form.

Appendix S1: Supplementary Information