This article is mentioned in:

Introduction

Acquired immunodeficiency syndrome (AIDS) is among the most complex diseases in the field of medicine. The incidence of Pneumocystis jiroveci pneumonia (PCP) has increased significantly and has become the most common opportunistic infection among patients with human immunodeficiency virus (HIV) (1). PCP is the most common cause of AIDS-related mortality, and the mortality rate of patients with AIDS that contract PCP in the early stages of the disease increases to 10–20% as the necessity for mechanical ventilation is significantly increased (2). Based on the results of five randomized controlled trials (RCTs), the use of corticosteroids was recommended for the treatment of patients co-infected with HIV and PCP by an expert panel in 1990 (3). Adjunctive corticosteroid treatment refers to the administration of corticosteroids in combination with sulfamethoxazole-trimethoprim (SMZ-TMP) or pentamidine (3,4). Corticosteroids may be categorized as long-acting, middle-acting or short-acting agents. The primary therapeutic agents used to treat PCP in patients with HIV include prednisone and prednisolone (5). On the basis of the results of previous clinical trials, a systematic review was conducted by Briel et al (5) in 2006, which demonstrated the feasibility of adjunctive corticosteroid treatment for the treatment PCP in patients co-infected with HIV. Adjunctive corticosteroid therapy was considered as an alternative to SMZ-TMP alone, and effectively improved survival in moderate to severe cases, reducing complications such as pneumothorax and respiratory failure (6). Therefore, adjunctive corticosteroid therapy has been recommended by the American CDC Guidelines to treat PCP associated with HIV-1 infection (6). However, corticosteroid therapy may increase the occurrence of opportunistic infections, by causing deterioration of cell-mediated immunity (7,8). Thus, the present meta-analysis aimed to evaluate the effects of adjunctive corticosteroid treatment for PCP in patients co-infected with HIV and to provide suggestions for clinical practice.

Materials and methods

Search methods

In order to analyze relevant RCTs, a search of the literature from the earliest available date to March 2014 was performed using various websites, including PubMed (http://www.ncbi.nlm.nih.gov/pmc/), Embase (https://www.elsevier.com/solutions/embase-biomedical-research) and Ovid (http://gateway.ovid.com/), using the following keywords: Corticosteroids; glucocorticoide; cortisol; corticosterone; HIV/AIDS; P. jiroveci pneumonia; and PCP.

Inclusion criteria

The following inclusion criteria were used in the present meta-analysis: i) Only RCT; ii) object of study, PCP in patients with HIV; iii) intervening measure, adjunctive corticosteroids treatment for patients with HIV in addition to the standardized treatment in the experimental group, and the use of placebo based on the standardized treatment, or standardized treatment alone administered to the control group; iv) patient reported outcomes, the mortality rate of patients following ~1 month (28–35 days); and v) language, English.

Exclusion criteria

Studies that met the inclusion criteria were excluded according to the following criteria: Experimental design was not strict (i.e. lacked a control group); sample data was unclear or missing and experimental data could not be extracted; the information was not accurate and the data could not be used; or the content of the literature was repetitive.

Assessment of study quality

Standard Jadad scoring was used to assess the methodological quality of the included RCTs, based on the adequacy of randomization, blinding and follow-up, with a maximum score of 5 points. A score of 0–2 indicated low quality, whereas a score of 3–4 indicated higher quality and a score of 5 denoted high quality.

Data synthesis and meta-analysis

In order to calculate the relative risk (RR) with 95% confidence interval (CI), data were analyzed using Stata 11.0 software (StataCorp LP, College Station, TX, USA). The χ2 test was used to analyze the heterogeneity of the RCTs. P>0.05 indicated that there was no statistically significant heterogeneity, therefore a fixed-effects model was applied using the Mantel-Haenzel (M-H) method; whereas P<0.05 indicated statistically significant heterogeneity, therefore a random-effects model was applied using the method proposed by DerSimonian and Laird (9).

Evaluating bias

In the present meta-analysis, potential publication bias was examined using a funnel plot. The symmetrical characteristics of the funnel plot were evaluated using the test proposed by Begg and Mazumdar (10) where a symmetrical result demonstrated no publication bias, and asymmetry indicated potential publication bias.

Results

Characteristics of included RCTs

A total of 550 studies were initially included; however, 527 were subsequently excluded due to duplicate content (n=385), an irrelevant study topic, abstract or other dependency (n=140) or because they were not published in English (n=2) (Fig. 1). Following detailed examination of the remaining 23 studies and the exclusion of citations that lacked sufficient detail or were not RCTs, six RCTs that met the inclusion criteria (11–16) were finally identified and the meta-analysis was performed. Table I outlines the RCTs included in the present meta-analysis.

Figure 1. Flow chart summarizing the literature search.

Table I. Characteristics of the RCTs included in the present meta-analysis.

Table I.

Characteristics of the RCTs included in the present meta-analysis.

			Mortality rate (%)		Intervention
RCT (ref.)	Year	Duration (days)	Experimental group	Control group	Experimental group	Control group
Terblanche et al (11)	2008	35	38.3 (18/47)	47.2 (25/53)	SMZ-TMP + corticosteroids	SMZ-TMP + placebo
Gagnon et al (12)	1990	28	25 (3/12)	81.8 (9/11)	SMZ-TMP + corticosteroids	SMZ-TMP + placebo
Bozzette et al (13)	1990	31	10.6 (13/123)	21.9 (28/128)	SMZ-TMP + corticosteroids	SMZ-TMP
Nielsen et al (14)	1992	34	6.7 (2/30)	31.0 (9/29)	SMZ-TMP + corticosteroids	SMZ-TMP
Montaner et al (15)	1990	30	5.6 (1/18)	0 (0/19)	SMZ-TMP or	SMZ-TMP or
					pentamidline + corticosteroids	pentamidline + placebo
Walmsley et al (16)	1995	35	10 (4/40)	15.8 (6/38)	SMZ-TMP or	SMZ-TMP or
					pentamidline + corticosteroids	pentamidline + placebo

[i] RCT, randomized controlled trial; SMZ-TMP, sulfamethoxazole and trimethoprim.

Quality assessment of the included trials

The six included studies were all RCTs, four of which were double-blinded (11,12,15,16) and two were not (13,14). The respective control groups received placebo (11,12,15,16) or null (13,14) treatment (Table II). No significant differences were determined between the characteristics from the six trials. No patient withdrawal or dropout occurred in the six included trials. Jadad scores were as follows: 3 (n=3), 4 (n=2) and 5 (n=1).

Table II. Quality assessment of the included trials.

Table II.

Quality assessment of the included trials.

RCT (ref.)	Randomization	Blinding	Jadad
Terblanche et al (11)	Adequate	Double-blind	5
Gagnon et al (12)	Unclear	Unclear	3
Bozzette et al (13)	Adequate	Non-blind	3
Nielsen et al (14)	Adequate	Non-blind	3
Montaner et al (15)	Adequate	Unclear	4
Walmsley et al (16)	Adequate	Unclear	4

[i] No patient dropout or withdrawal occurred in any of the studies. RCT, randomized controlled trial.

Meta-analysis

Six RCTs were enrolled in the present meta-analysis, which included a total of 548 patients with HIV suffering from PCP. The duration of observation was ~1 month (28–37 days), defined in terms of the patients' mortality rate. A total of were enrolled in The experimental groups, which contained 270 patients, demonstrated a mortality rate of 15.2% (n=41), as compared with 27.7% (n=77) in the control groups, (n=278). A heterogeneity test demonstrated that χ2=6.97 (P=0.223; I2=28.3%), therefore no significant heterogeneity was detected and the M-H method was employed. Subsequent meta-analysis demonstrated that the experimental group had a significantly decreased risk of mortality (0.55 times; P<0.05), as compared with the control group (RR, 0.55; 95% CI, 0.35–0.85) (Fig. 2).

Figure 2. Mantel-Haenzel model for meta-analysis.RR, relative ratio; CI, confidence interval.

Publication bias

Funnel plot analysis of publication bias in the six RCTs demonstrated that all six trials were present in the funnel plot and the majority were in the middle or top sections (Fig. 3). Funnel chart linear regression showed the bias coefficient was t=0.38 [Pr >|t|=0.707 (continuity corrected)], therefore the result was not statistically significant. The 95% CI was −3.52 to 2.34. The funnel plot was considered symmetrical, indicating no publication bias.

Figure 3. Begg's funnel plot analysis of the six included trials to investigate potential publication bias, with 95% confidence limits.

Sensitivity analysis

Sensitivity analysis of the included RCTs was performed by altering the inclusion criteria, excluding low quality RCTs and employing different statistical methods for analysis, using Stata 11.0 software. The results demonstrated that there were minor differences between the studies and the exclusion of any one study had minimal effects on the results (Fig. 4). When the random-effects model was used, the RR was 0.55 (95%CI, 0.35–0.85), approximated to the fixed-effects model, indicating a small but statistically insignificant heterogeneity among the included studies.

Figure 4. Sensitivity analysis of the included studies.

Discussion

The effective treatment of AIDS and its complications remains a complex worldwide problem. The incidence of PCP has increased significantly and it has become the most common opportunistic infection in patients with HIV (17,18). Adjunctive corticosteroids were initially suggested as a treatment for PCP in patients with HIV in 1990, and current treatment guidelines remain in agreement with this consensus statement (19).

Six RCTs with 548 cases of PCP in patients with HIV were enrolled in the present meta-analysis. The smallest sample size was 23 patients and the largest was 253 patients. The intervening measure was the use of adjunctive corticosteroids for the treatment of PCP in patients with HIV, with SMZ-TMP or pentamidline used as the standard treatment. Various types of control were used: Bozzette et al (13) and Nielsen et al (14) used blank controls, whereas the remaining RCTs (11,12,15,16) used placebo treatments. The results of one study differed from the other five RCTs. In the study conducted by Montaner et al (15) the RR was 3.16, suggesting that patients in the experimental group had a risk of mortality that was 3.16 times higher compared with the control group. Notably, the remaining 5 RCTs all indicated that the use of adjunctive corticosteroids for the treatment of PCP in patients with HIV may reduce patient mortality. The results of the present meta-analysis demonstrated that the risk of mortality in the experimental group was 0.55 times lower compared with the control group, indicating that the use of adjunctive corticosteroids for the treatment of PCP in patients with HIV may reduce mortality in the early phase of the disease.

As bias may have an effect on the credibility of the present meta-analysis, potential biases were analyzed. Publication bias is the most common bias to consider when conducting meta-analyses as positive results are more frequently published. In the present meta-analysis, potential publication bias was examined using a funnel plot. The funnel plot analysis of the 6 RCTs indicated that all the trials were represented in the plot and the majority were in the middle or top sections. Furthermore, the results of the linear regression indicated that the funnel plot was symmetrical, therefore no publications bias was detected in the present meta-analysis. Two of the studies (15,16) enrolled in the present meta-analysis used SMZ-TMP or pentamidline as standard treatment, whereas the others (11–14) used SMZ-TMP. The results of the heterogeneity test demonstrated that there was little heterogeneity between them (P=0.223; I2=28.3%). Furthermore, the results of the sensitivity analyses demonstrated that this did not have any influence on the results of the meta-analysis even following the exclusion of any one study. These results demonstrated that the difference between using SMZ-TMP/pentamidline and SMZ-TMP monotherapy was within the acceptable limits, therefore the results of the six RCTs could be combined for analysis. The results did not change when a random-effects model was used, demonstrating that there was minimal publication bias, and that any potential bias would have no substantial influence on the results of the present meta-analysis.

The results of the present meta-analysis indicated that the use of adjunctive corticosteroids for the treatment of PCP in patients with HIV may reduce mortality in the early stages of the disease. Therefore, adjunctive corticosteroids may be considered for the treatment of PCP in patients with HIV. However, there were a number of limitations to the present meta-analysis: An insufficient number of RCTs were analyzed; the sample sizes of certain studies were too low; and there were confounding factors, such as the origin of literature and the quantity of information, as not all six of the RCTs were double-blinded studies. Therefore, the conclusions of the present meta-analysis have certain limitations. Future studies will require RCTs with larger sample sizes in order to ensure the analyses of the effects of adjunctive corticosteroids on PCP in patients with HIV are more comprehensive, leading to a more reliable conclusion.

Acknowledgements

The present study was supported by the National Natural Science Foundation of China (grant no. 81360259), the Extracurricular Scientific Research Project of Guangxi Medical University (grant no. WLXSZX1411; Guangxi, China) and the Guangxi University Science and Technology Research Project (grant no. KY2015ZD026).

References