Literature search
In total, 439 citations were identified by the electronic search, after excluding studies for relevant reasons, a total of 19 studies [32-50] were included in this meta-analysis. Details of the literature search for obtaining these articles are shown in Figure 1. Of the 19 studies, 16 studies [32-34,37-39, 41-50] were case-control studies, 1 study [40] was retrospective studies, and 2 studies [35-36] were cohort studies. Two studies [44,47] investigated various cutoff values of serum BDG in definite, probable, possible/colonized patients and controls, respectively. Three studies [40,42,50] investigated the various cutoff values of serum BDG in definite/probable, possible/colonized patients and controls, individually. There were eleven studies [34-38,41-43,46,48-49],which only analyzed definite/probable patients and controls. Another four studies’ [32-33, 39,49] participants consisted of 2 populations, definite patients and controls. Finally, a total of 2,310 participants, of which 170 definite cases, 847 definite/probable cases,66 probable cases, 83 possible/colonized cases were available from the included studies. Characteristics of the included studies are shown in Table 2. Details of cutoff levels and their corresponding 2 x 2 contingency data in each included study are shown in Table 3.
Risk of bias/quality assessment
Table 4 provides results of assessing the risk of bias, with most studies demonstrating high risk of bias.Only 2 studies [35-36] had an unclear risk of bias. In Boer’ study [35], it was unclear whether the reference standard correctly classified the target condition. And there was no available data about the interval between the index test and reference standard in Passos’ study [36].
The results of Meta-analysis
Diagnostic accuracy of serum BDG for PJP patients versus controls
Forest plots of sensitivity and specificity with 95% CI are shown in Figure 2a-2e for the diagnosis of Pneumocystis jirovecii pneumonia in PJP patients versus controls at various cutoff values of serum BDG 80,200,300,400 and 500 pg/mL, individually. Table 5 shows pooled point estimates of sensitivity, specificity,LR-, LR+, DOR, Study heterogeneity Q2, study inconsistency I2 ,and P value of publication bias for the study-specific different BDG cutoff values for diagnosing PJP based on a bivariate random-effects model. There were 19, 9, 8,9,and 12 studies included in our study at different cutoff values of serum BDG 80 pg/mL [32-50], 200 pg/mL [32-33,40,42,44-47,50], 300 pg/mL[32-34,40,42,44-45,50], 400 pg/mL[32-33,39-40,42,44-46,50] and BDG 500 pg/mL [32-34, 40-47, 50], respectively. Evidence of significant publication bias was identified by Egger’s test when the thresholds of serum BDG were 80,400,500 pg/mL (P=0.007, 0.037 and 0.037, respectively), but there were no evidence of significant publication bias when the thresholds of serum BDG were 200,300 pg/mL (P=0.114 and 0.086, individually). Forest plots of sensitivity and specificity for the diagnosis of Pneumocystis jirovecii pneumonia (PJP patients vs controls) at various cutoff values are shown in Figure 2a-2e. Pooled point estimates with 95% confidence regions of sensitivity/specificity and the hierarchical summary receiver-operating characteristic(HSROC) curve for the different cutoff values of serum BDG are shown in Figure 3.
Diagnostic accuracy of serum BDG for HIV-negative PJP patients versus controls subgroup
Five studies [32-33,42,44-45] reported the diagnostic accuracy of serum BDG at various cutoffs of 80,200,300,400,and 500 pg/mL in HIV-negative PJP patients versus controls subgroup, individually. The detailed results of stratified analysis of the point estimates of sensitivity, specificity, LR+, LR-, and DOR are shown in Table 5. Except for 80mg/mL stratum (P=0.397), the heterogeneity studies in the other strata were consistently significant (P<0.001, for any other strata). Except for the threshold of serum BDG 80mg/mL stratum (P=0.087),no significant publication bias was performed by Egger’s test (P>0.10, for any other strata).
Diagnostic accuracy of serum BDG for HIV-positive PJP patients versus controls subgroup
A total of seven studies [32-33, 39, 41, 44-46] allowed reproduction of the 2 x 2 contingency tables for HIV-positive PJP patients versus controls. From these studies, 7 BDG cutoff value of 80mg/mL [32-33, 39, 41, 44-46] , 5 BDG cutoff value of 200mg/mL [32-33, 44-46] , 4 BDG cutoff value of 300mg/mL [32-33, 44-45] , 6 BDG cutoff value of 400mg/mL [32-33, 39, 44-46] , and 6 BDG cutoff value of 500mg/mL[32-33, 41, 44-46] datasets were obtained,individually. As to various thresholds of serum BDG in HIV-positive PJP individuals, overall point estimates of sensitivity, specificity,LR-, LR+, DOR, Study heterogeneity Q2, study inconsistency I2 ,and P value of publication bias based on a bivariate random-effects model are also shown in Table 5. Compared with HIV-negative patients in subgroup, LR- was lower for detecting HIV-positive persons who would develop PJP. Except for the cutoff value of serum BDG was 80mg/mL stratum (P=0.202), evidence of significant publication bias was identified by Egger’s test (P<0.10, for any other strata).
Diagnostic accuracy of serum BDG for PJP versus colonized patients subgroup
Six studies allowed the analysis of the diagnostic ability between PJP and colonized patients at various serum BDG cutoff values [40,42,44-45,47,50]. From these studies, 6 BDG cutoff value of 80mg/mL[40,42,44-45,47,50] , 6 BDG cutoff value of 200mg/mL [40,42,44-45,47,50], 5 BDG cutoff value of 300mg/mL[40,42,44-45,50], 6 BDG cutoff value of 400mg/mL [40,42,44-45,47,50], and 6 BDG cutoff value of 500mg/mL [40,42,44-45,47,50] datasets were available. The results of stratified analyses of point estimates of sensitivity, specificity, LR+, LR-, and the DOR for serum BDG based on a bivariate random-effects model are also shown in Table 5. Despite evidence of significant publication bias was performed (Table 5),the pooled overall LR- was low to 0.17 at a cutoff value of serum BDG 80 mg/mL, resulting in the strong ability to exclude PJP when the result of PJ-qPCR is positive in BAL. No evidence of significant publication bias was performed by Egger tests (P>0.10, for any other strata).