Adjunctive corticosteroids for Pneumocystis jiroveci pneumonia in patients with HIV infection
Abstract
Background
Pneumocystis jiroveci pneumonia (PCP) remains the most common opportunistic infection in patients infected with the human immunodeficiency virus (HIV). Among patients with HIV infection and PCP the mortality rate is 10% to 20% during the initial infection and this increases substantially with the need for mechanical ventilation. It has been suggested that corticosteroids adjunctive to standard treatment for PCP could prevent the need for mechanical ventilation and decrease mortality in these patients.
Objectives
To assess the effects of adjunctive corticosteroids on overall mortality and the need for mechanical ventilation in HIV‐infected patients with PCP and substantial hypoxaemia (arterial oxygen partial pressure < 70 mmHg or alveolar‐arterial gradient > 35 mmHg on room air).
Search methods
For the original review we searched The Cochrane Library (2004, Issue 4), MEDLINE (January 1980 to December 2004) and EMBASE (January 1985 to December 2004) without language restrictions. We further reviewed the reference lists from previously published overviews, searched UptoDate version 2005 and Clinical Evidence Concise (Issue 12, 2004), contacted experts in the field and searched the reference lists of identified publications for citations of additional relevant articles.
In this update of our review, we searched the above‐mentioned databases in September 2010 and April 2014 for trials published since our original review. We also searched for ongoing trials in ClinicalTrials.gov and the World Health Organization International Clinical Trial Registry Platform (ICTRP). We searched for conference abstracts via AEGIS.
Selection criteria
Randomised controlled trials that compared corticosteroids to placebo or usual care in HIV‐infected patients with PCP in addition to baseline treatment with trimethoprim‐sulfamethoxazole, pentamidine or dapsone‐trimethoprim, and reported mortality data. We excluded trials in patients with no or mild hypoxaemia (arterial oxygen partial pressure > 70 mmHg or an alveolar‐arterial gradient < 35 mmHg on room air) and trials with a follow‐up of less than 30 days.
Data collection and analysis
Two teams of review authors independently evaluated the methodology and extracted data from each primary study. We pooled treatment effects across studies and calculated a weighted average risk ratio of overall mortality in the treatment and control groups using a random‐effects model.
In this update of our review, we used the GRADE methodology to assess evidence quality.
Main results
Of 2029 screened records, we included seven studies in the review and six in the meta‐analysis. Risk of bias varied: the randomisation and allocation process was often not clearly described, five of seven studies were double‐blind and there was almost no missing data. The quality of the evidence for mortality was high. Risk ratios for overall mortality for adjunctive corticosteroids were 0.56 (95% confidence interval (CI) 0.32 to 0.98) at one month and 0.59 (95% CI 0.41 to 0.85) at three to four months of follow‐up. In adults, to prevent one death, numbers needed to treat are nine patients in a setting without highly active antiretroviral therapy (HAART) available, and 23 patients with HAART available. The three largest trials provided moderate quality data on the need for mechanical ventilation, with a risk ratio of 0.38 (95% CI 0.20 to 0.73) in favour of adjunctive corticosteroids. One study was conducted in infants, suggesting a risk ratio for death in hospital of 0.81 (95% CI 0.51 to 1.29; moderate quality evidence).
Authors' conclusions
The number and size of trials investigating adjunctive corticosteroids for HIV‐infected patients with PCP is small, but the evidence from this review suggests a beneficial effect for adult patients with substantial hypoxaemia. There is insufficient evidence on the effect of adjunctive corticosteroids on survival in infants.
PICOs
Plain language summary
Adjunctive corticosteroids for Pneumocystis jiroveci pneumonia in patients with HIV infection
Pneumocystis jiroveci pneumonia (PCP), formerly called Pneumocystis carinii pneumonia, is the most common opportunistic infection among patients infected with HIV. In 1990, based on evidence from five randomised controlled trials, an expert panel recommended the use of corticosteroids for HIV‐infected patients with PCP and substantial hypoxaemia (low levels of oxygen in the blood).
The objective of this systematic review was to assess the effects of adjunctive (additional) corticosteroids on mortality and the need for mechanical ventilation in patients co‐infected with HIV and PCP. We searched for eligible studies up to April 2014. We included seven studies in this review and six in the meta‐analysis (combining of study data).
The number and size of the trials investigating adjunctive corticosteroids for HIV‐infected patients co‐infected with PCP is small (the six trials included in the meta‐analysis comprised 242 individuals in the intervention groups and 247 individuals in the control groups; the trial on infants comprised 47 individuals in the intervention group and 53 in the control group). Follow‐up ranged from three to 14 months. The evidence from this review was of high quality for mortality and of moderate quality for need for mechanical ventilation and suggests a beneficial effect for adult patients with substantial hypoxaemia. For infants (18 months or younger) with HIV and suspected PCP there is insufficient evidence on whether the effect of adjunctive corticosteroids could improve survival (the confidence interval for the estimate of effect is wide, includes both clinically relevant benefit and harm and is of moderate quality).
Authors' conclusions
Summary of findings
| Adjunctive corticosteroids versus no such treatment for Pneumocystis jiroveci pneumonia in patients with HIV infection | ||||||
| Patient or population: patients with Pneumocystis jiroveci pneumonia and HIV infection Comparison: no adjunctive corticosteroids | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| Without adjunctive corticosteroids | With adjunctive corticosteroids | |||||
| Death at 1 month (adults) | Study population1 | RR 0.56 | 489 | ⊕⊕⊕⊕ | ||
| 247 per 1000 | 138 per 1000 | |||||
| Low1 | ||||||
| 100 per 1000 | 56 per 1000 | |||||
| High1 | ||||||
| 250 per 1000 | 140 per 1000 | |||||
| Death at 3 to 4 months (adults) | Study population1 | RR 0.59 | 448 | ⊕⊕⊕⊕ | ||
| 258 per 1000 | 152 per 1000 | |||||
| Low1 | ||||||
| 100 per 1000 | 59 per 1000 | |||||
| High1 | ||||||
| 250 per 1000 | 147 per 1000 | |||||
| Death in hospital; children | 472 per 1000 | 382 per 1000 | RR 0.81 | 100 | ⊕⊕⊕⊝ | |
| Need for mechanical ventilation at 1 month | 164 per 1000 | 62 per 1000 | RR 0.38 | 388 | ⊕⊕⊕⊝ | |
| *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| 1In Western countries where HAART is widely available the mortality rate is 10%; in developing countries where HAART is often not available the mortality rate is 25%. 6Quality of evidence downgraded from high to moderate because lack of blinding could influence the decision of a caregiver to initiate mechanical ventilation. | ||||||
Background
Description of the condition
With the introduction of highly active antiretroviral therapy (HAART) more than two decades ago, the incidence of Pneumocystis jiroveci pneumonia (Stringer 2002) has decreased significantly in the Western hemisphere. However, PCP (the acronym stands for pneumocystis pneumonia) still remains one of the most common opportunistic infections in patients infected with the human immunodeficiency virus (HIV) (Kaplan 2000). Among patients with HIV infection and PCP the mortality rate is 10% to 20% during the initial infection and this increases substantially with the need for mechanical ventilation (Randall 2000).
Description of the intervention
Two to three days after starting anti‐PCP therapy, the respiratory situation of PCP patients often worsens because of increased inflammation in the lungs as a reaction to pneumocystis particles from killed organisms. Corticosteroids given in conjunction with anti‐PCP therapy may help to better control the inflammatory process. Therefore the corticosteroid treatment should be started as early as possible but within 72 hours after starting the PCP‐specific therapy. So far, there is no evidence about an optimal dose or duration of adjunctive corticosteroids. The following 21‐day oral regimen with prednisone has been recommended: 40 mg orally twice daily for days one to five, 40 mg once daily for days six to 10, and 20 mg once daily for days 11 to 21 (Benson 2004; EACS 2013). If parenteral administration is necessary, it is recommended to use methylprednisolone at 75% of the respective prednisone dose (CDC 2013). In children with severe PCP, it is recommended to start the corticosteroid treatment as early as possible but within 72 hours after diagnosis. The recommended dose of prednisone for children is 1 mg/kg of body weight twice daily for days one to five, 0.5 mg/kg once daily for days six to 10, and 0.5 mg/kg once daily for days 11 to 21 (NIH 2013). The alternative regimen with methylprednisolone (intravenous) is 1 mg/kg/dose every six hours for days one to seven, 1 mg/kg/dose once daily for days eight to nine, 0.5 mg/kg/dose twice daily for days 10 to 11, and 1 mg/kg/dose once daily for days 12 to 16 (NIH 2013).
How the intervention might work
Within two to three days of the initiation of anti‐PCP therapy, the health status of patients often worsens. It is presumed that the patient's alveolar‐arterial oxygen gradient and the inflammatory processes in the lungs increase as organisms are killed (Sax 2012). Adjunctive corticosteroids administered with initiation of anti‐PCP therapy may reduce the inflammatory process and prevent this clinical worsening.
Why it is important to do this review
In 1990, an expert panel recommended the use of corticosteroids for HIV‐infected patients with PCP and substantial hypoxaemia (initial arterial oxygen partial pressure of < 70 mmHg or alveolar‐arterial gradient > 35 mmHg on room air) based on the evidence from five randomised controlled trials (Consensus 1990). The studies used for the consensus statement still represent the basis for the current guidelines for the treatment of adults (CDC 2013) and children (NIH 2013). However, at the time of the first consensus statement, one trial was not yet completed (Nielsen 1992), two trials had been stopped prematurely (Gagnon 1990; Montaner 1990), and one trial was not published in full (Clement 1989). In 1992, a systematic review qualitatively summarised the same incomplete data (Sistek 1992).
Objectives
To assess the effects of adjunctive corticosteroids on overall mortality and the need for mechanical ventilation in HIV‐infected patients with PCP and substantial hypoxaemia (arterial oxygen partial pressure < 70 mmHg or alveolar‐arterial gradient > 35 mmHg on room air).
Methods
Criteria for considering studies for this review
Types of studies
We considered trials eligible for this review if they used random allocation of participants into parallel groups and reported mortality data. We excluded trials with a follow‐up of less than 30 days.
Types of participants
All HIV‐infected patients with moderate‐severe PCP, defined as PCP with substantial hypoxaemia. We excluded patients with no or mild hypoxaemia (arterial oxygen partial pressure > 70 mmHg or an alveolar‐arterial gradient < 35 mmHg on room air).
Types of interventions
We included studies if they compared corticosteroids to placebo or usual care in HIV‐infected patients with PCP in addition to baseline treatment with trimethoprim‐sulfamethoxazole, pentamidine or dapsone‐trimethoprim.
Types of outcome measures
The main outcome measure of interest was overall mortality at one and three to four months of follow‐up.
A secondary outcome measure was the need for mechanical ventilation.
Search methods for identification of studies
Electronic searches
We searched The Cochrane Library (2004, Issue 4), MEDLINE (January 1985 to December 2004) and EMBASE (January 1985 to December 2004) without language restrictions to identify randomised controlled trials that compared adjunctive corticosteroids to control in HIV‐infected patients with PCP. We used the terms steroid*, corticosteroid*, glucocorticoid*, Pneumocystis, PCP, *carinii and *jiroveci as text words and Glucocorticoids, Adrenal Cortex Hormones, Steroids, Pneumocystis Infections, Pneumocystis jiroveci and Pneumonia, Pneumocystis as Medical Subject Headings. We restricted the search to articles indexed as randomised controlled trials (publication type) or drug therapy (subject heading) or those that included the words random* or placebo in their titles or abstracts. In this review update, we searched the above‐mentioned databases in September 2010 and April 2014 for trials published since our original review. A detailed search strategy is presented in Appendix 1.
Searching other resources
We further reviewed the reference lists from previously published overviews (Consensus 1990; Sistek 1992), searched UptoDate version 2005 and Clinical Evidence Concise (Issue 12, 2004), contacted experts in the field and searched the reference lists of identified publications for citations of additional relevant articles. In this update of our review, we searched for ongoing trials in ClinicalTrials.gov and the World Health Organization International Clinical Trial Registry Platform (ICTRP). We searched for conference abstracts via AEGIS.
Data collection and analysis
Selection of studies
Two teams of investigators (MB/HCB and RB/HF) independently assessed study eligibility and quality and resolved any disagreement by consensus. We abstracted data from eligible trials in duplicate. HR and MB assessed trials identified during the update in 2010; HE and MB assessed potentially eligible trials identified in 2014.
Assessment of risk of bias in included studies
We assessed the risk of bias of the included trials with respect to random sequence generation; concealment of treatment allocation; blinding of patients, caregivers or assessors of clinical outcomes; completeness of follow‐up; performance of a sample size calculation; and whether the trial was stopped early for benefit (Juni 1999; Montori 2005). HR and MB assessed trials identified during the update in 2010; HE and MB assessed potentially eligible trials identified in 2014.
Assessment of heterogeneity
We tested for heterogeneity with the Cochrane Q test and measured inconsistency of treatment effects across studies using the I2 statistic (the percentage of total variance across studies that is due to heterogeneity rather than chance) (Higgins 2002; Higgins 2003).
Assessment of reporting biases
We investigated the presence of publication bias by means of funnel plots (Sterne 2001).
Data synthesis
We performed all analyses according to the intention‐to‐treat principle, i.e. we analysed patients in the groups to which they were randomised. In our main analysis we assumed that patients lost to follow‐up did not have an event. In a sensitivity analysis we assumed that patients lost to follow‐up experienced an event. We pooled treatment effects across studies and calculated a weighted average risk ratio of overall mortality in the treatment and control groups using a random‐effects model. We performed statistical analyses using Review Manager 5.3 (RevMan 2014). We calculated numbers needed to treat to prevent one death by multiplying the mean relative risk reduction with an initial mean baseline risk (Marx 2003).
Sensitivity analysis
We carried out sensitivity analyses to examine treatment effects according to quality components of trials, whether the publication was a peer‐reviewed article or just in abstract form, and whether trials were stopped early for benefit versus not.
'Summary of findings' table
In this update of our review, we used the GRADE methodology to assess the quality of the evidence (Guyatt 2008). We have presented the quality of the evidence for the following outcomes in a 'Summary of findings' table:
-
death at one month (adults);
-
death at three to four months (adults);
-
death in hospital (children).
-
need for mechanical ventilation
Results
Description of studies
In our original search, we screened 1591 titles and abstracts. We excluded 1583 records and assessed eight in full text. Of these, two trials did not meet our inclusion criteria and we excluded them. One trial investigated only patients with mild hypoxaemia and had a short follow‐up of only three days (Jeantils 1993), and another trial, Montaner 1993, was a subgroup analysis of a larger included trial (Montaner 1990). The remaining six trials investigated adults with HIV and moderate‐severe PCP; one was only published in abstract form (Clement 1989), two were stopped prematurely due to apparent benefits in the treatment groups with adjunctive glucocorticoid therapy (Gagnon 1990; Montaner 1990), and one was stopped early due to published evidence from other studies in favour of adjunctive corticosteroids (Nielsen 1992). Only three studies were completed and published in full versions (Bozzette 1990; Nielsen 1992; Walmsley 1995).
The update search in 2010 yielded 335 records. We excluded 334 on the basis of the title and abstract. We assessed one study in full text and subsequently included it (Terblanche 2008). This study investigated adjunctive corticosteroid treatment in HIV‐exposed infants less than 18 months old.
For the update search in 2014, we screened 103 records as titles and abstracts, which did not yield any further eligible studies. Overall, we screened 2029 titles and abstracts, excluded 2020, screened nine in full text and included seven of these in our systematic review. Details of the included and excluded trials are provided in the Characteristics of included studies and Characteristics of excluded studies tables, respectively.
Risk of bias in included studies
The quality assessment for each study is shown in the 'Risk of bias' tables in Characteristics of included studies.
Overall, the risk of bias of the six studies included in the meta‐analysis varied. Details of random sequence generation were reported in three of seven studies. Patient follow‐up was almost complete for all included studies due to short follow‐up periods (the total number of patients lost to follow‐up was four). In the six studies conducted in adults, concealed allocation of participants was reported in three studies (Bozzette 1990; Montaner 1990; Walmsley 1995). Four trials reported double‐blinding without specifying who was blinded (Clement 1989; Gagnon 1990; Montaner 1990; Walmsley 1995). Three trials were single‐centre (Clement 1989; Gagnon 1990: Montaner 1990), and three were multi‐centre (Bozzette 1990; Nielsen 1992; Walmsley 1995). Four trials reported the performance of a sample size calculation (Bozzette 1990; Gagnon 1990; Montaner 1990; Walmsley 1995). Two trials were stopped prematurely due to apparent treatment benefits of adjunctive corticosteroids (Gagnon 1990; Montaner 1990), and one was stopped due to external evidence of the benefits of corticosteroids (Nielsen 1992). The stopping early of clinical trials for benefit may lead to an overestimation of treatment effects due to catching the apparent benefit of treatment at a "random high" (Montori 2005). The overall risk of bias of the study conducted in infants was low (Terblanche 2008). Figure 1 gives an overview of the 'Risk of bias' assessment.
'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.
Effects of interventions
We included seven trials in this systematic review. The six trials included in the meta‐analysis comprised a total of 242 individuals in the intervention groups and 247 individuals in the control groups, with a follow‐up ranging between three and 14 months. Figure 2 (funnel plot) indicates no evidence of publication bias. The trial in infants comprised 47 individuals in the intervention group and 53 in the control group, who were followed up until hospital discharge. summary of findings Table for the main comparison gives an overview of the intervention effects, with a GRADE analysis of evidence quality for each outcome.
Funnel plot to evaluate the presence of publication bias in trials investigating adjunctive corticosteroids for Pneumocystis jiroveci pneumonia in HIV‐infected patients. The funnel graph plots the log of the treatment odds ratio against the standard error (SE) of the log odds ratio (an indicator of sample size). Open circles represent trials included in the meta‐analysis. The line in the centre indicates the summary log odds ratio. In the absence of publication bias, the log odds ratio estimates from smaller trials are expected to be scattered above and below the summary estimate, producing a symmetric triangular or funnel shape. When smaller trials with larger log odds ratios are missing, the funnel plot appears asymmetric and may indicate the presence of publication bias. In our systematic review the funnel plot looks symmetric. The Egger test for publication bias was not statistically significant (P value = 0.91).
Adjunctive corticosteroids versus placebo or usual care
Overall mortality
Risk ratios for overall mortality were significantly reduced for adjunctive corticosteroids in adult patients with HIV at one month (RR 0.56; 95% CI 0.32 to 0.98) (Analysis 1.1; Figure 3) and at three to four months of follow‐up (RR 0.59; 95% CI 0.41 to 0.85) (Analysis 1.2; Figure 4). We found some evidence of heterogeneity among trials at one month (test of heterogeneity P value = 0.12; I2 = 43%), whereas at three to four months treatment effects looked more homogenous (P value = 0.78; I2 = 0%). We judged the quality of the evidence for these outcomes to be high as the influence of blinding is not of significant importance for the outcome mortality.
Forest plot of comparison: 1 Adjunctive corticosteroids versus no such treatment, outcome: 1.1 Death at 1 month; adults.
Forest plot of comparison: 1 Adjunctive corticosteroids versus no such treatment, outcome: 1.2 Death at 3 to 4 months; adults.
We explored inconsistencies among studies in sensitivity analyses. Heterogeneity was considerably reduced when the analysis was limited to trials reporting early adjunctive corticosteroids (within three days) that were published in full, i.e. excluding Clement 1989 (summary risk ratio for mortality at one month 0.44; 95% CI 0.28 to 0.69, heterogeneity P value = 0.49; I2 = 0%) (Analysis 1.6). We carried out further sensitivity analyses for the mortality endpoint at one month. In trials that reported concealed allocation (Bozzette 1990; Montaner 1990; Walmsley 1995), the summary risk ratio was 0.54 (95% CI 0.32 to 0.92, heterogeneity P value = 0.49; I2 = 0%) (Analysis 1.7). In trials reporting blinding of patients and caregivers (Clement 1989; Gagnon 1990; Montaner 1990; Walmsley 1995), the summary risk ratio was 0.71 (95% CI 0.33 to 1.55, heterogeneity P value = 0.14; I2 = 44%) (Analysis 1.8). In trials not prematurely halted (Bozzette 1990; Clement 1989; Walmsley 1995), the summary risk ratio was 0.71 (95% CI 0.39 to 1.31, heterogeneity P value = 0.16; I2 = 45%) (Analysis 1.9). The assumption that patients lost to follow‐up at one month died gives a summary risk ratio of 0.56 (95% CI 0.32 to 0.97, heterogeneity P value = 0.11; I2 = 44%) (Analysis 1.5).
The risk ratio for in‐hospital mortality in infants receiving adjunctive corticosteroids was 0.81 (95% CI 0.51 to 1.29) (Terblanche 2008) (Analysis 1.3; Figure 5). We downgraded the quality of the evidence for this outcome from high to moderate as there was only one trial available, with a wide confidence interval including clinically relevant benefit and harm.
Forest plot of comparison: 1 Adjunctive corticosteroids versus no such treatment, outcome: 1.3 Death in hospital; children.
Need for mechanical ventilation
Reliable data on the need for mechanical ventilation were only available for the three largest trials (Bozzette 1990; Nielsen 1992; Walmsley 1995). Again, the risk ratio for this endpoint was largely reduced in the group with early adjunctive corticosteroids (0.38; 95% CI 0.20 to 0.73; P value = 0.40; I2 = 0%) (Analysis 1.4; Figure 6). We downgraded the quality of the evidence for this outcome from high to moderate because lack of blinding could influence the decision of a caregiver to initiate mechanical ventilation.
Forest plot of comparison: 1 Adjunctive corticosteroids versus no such treatment, outcome: 1.4 Need for mechanical ventilation at 1 month; adults.
Discussion
The meta‐analysis of six randomised controlled trials in HIV‐infected adults with Pneumocystis jiroveci pneumonia (PCP) and substantial hypoxaemia found a significant reduction in the relative risk of death with adjunctive corticosteroids of 44% at one month and 41% at three to four months. The average weighted mean mortality in the control groups of these trials at one month was 25%. This initial mortality rate of 25% can be assumed in settings where highly active antiretroviral therapy (HAART) is not available, which is still the case for most developing countries (Fisk 2003). In this situation we estimated that nine (95% confidence interval (CI) 6 to 200) HIV‐infected adults with PCP have to be treated early with adjunctive corticosteroids to prevent one death during the first month after PCP diagnosis. In Western countries, where HAART is widely available, we estimated the respective number needed to treat to be 23 patients (95% CI 15 to 500), assuming an initial mortality rate of 10% (Sepkowitz 2002). With regard to the need for mechanical ventilation, the risk reduction with adjunctive corticosteroids was even greater in the investigated patient population, but the number of trials was small (n = 3).
Only one trial reported overall mortality during the hospitalisation of infants under 18 months of age diagnosed with PCP. There was insufficient evidence to be certain that adjunctive corticosteroids had an effect on mortality (confidence intervals were wide, including both clinically relevant benefit and harm). Due to limited resources, the authors of that trial fixed the number of participating infants at 100 (Terblanche 2008).
This review has several strengths and limitations. We conducted an extensive literature search and updates to retrieve all eligible trials. However, formal testing for publication bias was not powerful because of the small number of included trials. Even with a symmetric funnel plot (Figure 2), such bias cannot be ruled out. Moreover, with a small number of included trials the uncertainty interval for the inconsistency among trials may not be very informative (Higgins 2003). We focused mainly on mortality data, which may be less prone to ascertainment bias, and we analysed the data according to the intention‐to‐treat principle to get more conservative estimates. Finally, the trials included in this meta‐analysis used different corticosteroid regimens. So far, neither the dosing nor the length and tapering schedule of corticosteroids has been adequately addressed in randomised trials. In former and current recommendations for adults (Benson 2004; CDC 2013), the corticosteroid schedule of the largest trial was adapted (Bozzette 1990). The former (Frieden 2009) and current guidelines (NIH 2013) for children have taken the trials in adults into account as well as three comparative observational studies in children (Bye 1994; McLaughlin 1995; Sleasman 1993).
There has been some concern among physicians treating patients with AIDS that further immunosuppression due to corticosteroid therapy could accelerate the onset of other HIV‐related opportunistic complications (Lambertus 1990; Nelson 1993). However, with the exception of an increase in muco‐cutaneous herpes simplex infection episodes (Bozzette 1990), adjunctive corticosteroids were not associated with an increase in opportunistic complications in any of the included trials. A large cohort study that used a standard 21‐day tapering course of adjunctive corticosteroids found no difference in the risk of AIDS‐related complications apart from an increase in oesophageal candidiasis (Gallant 1998).
It is possible that adjunctive corticosteroids are also beneficial for HIV‐infected patients with mild hypoxaemia due to PCP (Jeantils 1993). However, in this situation the short‐term mortality is low and possible unfavourable effects of corticosteroids might outweigh the benefits. Evidence from randomised controlled trials for non‐HIV‐infected patients with severe PCP is still lacking. Retrospective observational studies have shown conflicting results: while Pareja 1998 found that corticosteroids might be beneficial, Delclaux 1999 and Moon 2011 did not find an improvement of outcomes in non‐HIV‐infected patients with severe PCP.
'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.
Funnel plot to evaluate the presence of publication bias in trials investigating adjunctive corticosteroids for Pneumocystis jiroveci pneumonia in HIV‐infected patients. The funnel graph plots the log of the treatment odds ratio against the standard error (SE) of the log odds ratio (an indicator of sample size). Open circles represent trials included in the meta‐analysis. The line in the centre indicates the summary log odds ratio. In the absence of publication bias, the log odds ratio estimates from smaller trials are expected to be scattered above and below the summary estimate, producing a symmetric triangular or funnel shape. When smaller trials with larger log odds ratios are missing, the funnel plot appears asymmetric and may indicate the presence of publication bias. In our systematic review the funnel plot looks symmetric. The Egger test for publication bias was not statistically significant (P value = 0.91).
Forest plot of comparison: 1 Adjunctive corticosteroids versus no such treatment, outcome: 1.1 Death at 1 month; adults.
Forest plot of comparison: 1 Adjunctive corticosteroids versus no such treatment, outcome: 1.2 Death at 3 to 4 months; adults.
Forest plot of comparison: 1 Adjunctive corticosteroids versus no such treatment, outcome: 1.3 Death in hospital; children.
Forest plot of comparison: 1 Adjunctive corticosteroids versus no such treatment, outcome: 1.4 Need for mechanical ventilation at 1 month; adults.
Comparison 1 Adjunctive corticosteroids versus no such treatment, Outcome 1 Death at 1 month; adults.
Comparison 1 Adjunctive corticosteroids versus no such treatment, Outcome 2 Death at 3 to 4 months; adults.
Comparison 1 Adjunctive corticosteroids versus no such treatment, Outcome 3 Death in hospital; children.
Comparison 1 Adjunctive corticosteroids versus no such treatment, Outcome 4 Need for mechanical ventilation at 1 month; adults.
Comparison 1 Adjunctive corticosteroids versus no such treatment, Outcome 5 Sensitivity analysis: Death at 1 month; adults; loss to follow‐up = dead.
Comparison 1 Adjunctive corticosteroids versus no such treatment, Outcome 6 Sensitivity analysis: Death at 1 month; adults; without Clement 1989.
Comparison 1 Adjunctive corticosteroids versus no such treatment, Outcome 7 Sensitivity analysis: Death at 1 month; adults; concealed allocation only.
Comparison 1 Adjunctive corticosteroids versus no such treatment, Outcome 8 Sensitivity analysis: Death at 1 month; adults; blinding of patients and caregivers.
Comparison 1 Adjunctive corticosteroids versus no such treatment, Outcome 9 Sensitivity analysis: Death at 1 month; adults; trials not prematurely halted.
| Adjunctive corticosteroids versus no such treatment for Pneumocystis jiroveci pneumonia in patients with HIV infection | ||||||
| Patient or population: patients with Pneumocystis jiroveci pneumonia and HIV infection Comparison: no adjunctive corticosteroids | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| Without adjunctive corticosteroids | With adjunctive corticosteroids | |||||
| Death at 1 month (adults) | Study population1 | RR 0.56 | 489 | ⊕⊕⊕⊕ | ||
| 247 per 1000 | 138 per 1000 | |||||
| Low1 | ||||||
| 100 per 1000 | 56 per 1000 | |||||
| High1 | ||||||
| 250 per 1000 | 140 per 1000 | |||||
| Death at 3 to 4 months (adults) | Study population1 | RR 0.59 | 448 | ⊕⊕⊕⊕ | ||
| 258 per 1000 | 152 per 1000 | |||||
| Low1 | ||||||
| 100 per 1000 | 59 per 1000 | |||||
| High1 | ||||||
| 250 per 1000 | 147 per 1000 | |||||
| Death in hospital; children | 472 per 1000 | 382 per 1000 | RR 0.81 | 100 | ⊕⊕⊕⊝ | |
| Need for mechanical ventilation at 1 month | 164 per 1000 | 62 per 1000 | RR 0.38 | 388 | ⊕⊕⊕⊝ | |
| *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| 1In Western countries where HAART is widely available the mortality rate is 10%; in developing countries where HAART is often not available the mortality rate is 25%. 6Quality of evidence downgraded from high to moderate because lack of blinding could influence the decision of a caregiver to initiate mechanical ventilation. | ||||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Death at 1 month; adults Show forest plot | 6 | 489 | Risk Ratio (M‐H, Random, 95% CI) | 0.56 [0.32, 0.98] |
| 2 Death at 3 to 4 months; adults Show forest plot | 5 | 448 | Risk Ratio (M‐H, Random, 95% CI) | 0.59 [0.41, 0.85] |
| 3 Death in hospital; children Show forest plot | 1 | 100 | Risk Ratio (M‐H, Random, 95% CI) | 0.81 [0.51, 1.29] |
| 4 Need for mechanical ventilation at 1 month; adults Show forest plot | 3 | 388 | Risk Ratio (M‐H, Random, 95% CI) | 0.38 [0.20, 0.73] |
| 5 Sensitivity analysis: Death at 1 month; adults; loss to follow‐up = dead Show forest plot | 6 | 489 | Risk Ratio (M‐H, Random, 95% CI) | 0.56 [0.32, 0.97] |
| 6 Sensitivity analysis: Death at 1 month; adults; without Clement 1989 Show forest plot | 5 | 448 | Risk Ratio (M‐H, Random, 95% CI) | 0.44 [0.28, 0.69] |
| 7 Sensitivity analysis: Death at 1 month; adults; concealed allocation only Show forest plot | 3 | 366 | Risk Ratio (M‐H, Random, 95% CI) | 0.54 [0.32, 0.92] |
| 8 Sensitivity analysis: Death at 1 month; adults; blinding of patients and caregivers Show forest plot | 4 | 179 | Risk Ratio (M‐H, Random, 95% CI) | 0.71 [0.33, 1.55] |
| 9 Sensitivity analysis: Death at 1 month; adults; trials not prematurely halted Show forest plot | 3 | 370 | Risk Ratio (M‐H, Random, 95% CI) | 0.71 [0.39, 1.31] |
| Treatment group | PCP PCR in sputum | PCP IF | Nasopharyngeal aspirate for respiratory viruses |
| Prednisone group | 5/9 | 3/39 | RSV: 1/41 CMV: 1/41 |
| Placebo group | 3/6 | 5/46 | RSV: 7/47 Adenovirus: 1/47 Influenza A: 2/47 Parainfluenza 3: 1/47 |
| PCP: Pneumocystis jiroveci pneumonia | |||
| Treatment group | PCP PCR in sputum | PCP IF | Nasopharyngeal aspirate for respiratory viruses |
| Prednisone group | 5/9 | 3/39 | RSV: 1/41 CMV: 1/41 |
| Placebo group | 3/6 | 5/46 | RSV: 7/47 Adenovirus: 1/47 Influenza A: 2/47 Parainfluenza 3: 1/47 |
| PCP: Pneumocystis jiroveci pneumonia | |||
