Predicting the Prognosis of Patients with Pneumocystis Pneumonia by the (1,3)-β-D-Glucan Test Change

Background: To investigate the effects of the (1, 3)-β-D-glucan test (G test) to predict the outcomes of the patients with Pneumocystis pneumonia ( PCP ). Methods: The clinical data of PCP patients diagnosed from 2010 to 2018 in Peking Union Hospital collected. The imaging changes, the independent risk factors of death, changes of the G test and the patients’ outcomes were used as a receiver operating the characteristic (ROC) curves, and the area under the curve (AUC) calculated. Results: A total number of 104 patients enrolled in. The independent risk factors of death for PCP included pulmonary interstitial fibrosis (P=0.003), Cytomegalovirus co-infection (P=0.006), severe complication during treatment (P=0.001). The reduction of the G test was less than 9.13% of the initial G test (P=0.005), and the second imaging was not better than the first imaging (P=0.001). The efficiency of predicting the prognosis by the declined G test (AUC=0.740) is similar (P=0.893) to that of improving the imaging (AUC=0.731). The combination of the risk factors was the most effective predictor of the patients’ outcomes (AUC=0.938, CI 0.873 - 0.976, P < 0.001). Conclusion: The changes of G test can predict patient prognosis, and their efficiency is comparable to that of the changed imaging.


Introduction
As early as in 1996, researchers have proposed that the (1, 3)-β-D-glucan test (G test) could diagnose Pneumocystis pneumonia (PCP) through some animal experiments.
Moreover, the changes of the G test could be used to monitor the effects of the treatment (1,2). Later, the researchers have confirmed the role of the G test in the diagnosis of PCP in the clinical studies (3)(4)(5), but the G test was considered to be independent of the severity of the disease (6), and it doesn't reflect the patients' response to the anti-PCP drugs (6,7). A prospective study of 31 patients with PCP in 2011 suggested that the G test could serve as a marker for monitoring the effects of the treatment (8). Another study conducted showed a dynamic review of the results of the G test in 18 patients with PCP, and found that patients with a dynamically reduced results of the G test were mostly improved, but the dynamically elevated results of the G test didn't mean a worse clinical performance (9). One study in 2012 indicated that changes of the G test could not predict clinical mortality in 6 or 12 weeks (10). In conclusion, the G test can diagnose PCP, however, whether it can judge the prognosis of patients is still controversial. This study was designed to analyze whether the changes of the G test can be used as a predictor for the patients' outcomes, and how is its efficiency comparable to the other clinical indicators.

Methods
Research design: This study is a retrospective, single-center, clinical study. The protocol was approved by our institutional review board(IRB)(Peking Union Medical College Hospital(PUMCH), IRB S-K 643).
Population*: (1) The diagnosis of PCP was confirmed on the first page of the medical record in Peking Union Hospital from 2010 to 2018, with or without fungal infections. (2) There were at least 2 results of G test (Endosafe Glucan Assay Cartridges) during hospitalization. The interval between the first G test (G1) and the second G test (G2) was ≥ 5 days. The results of theG1 ≥105pg/ml (11). (3) There are at least 2 results of the CT imaging or the X-Ray imaging during the hospitalization. The interval between the first imaging and the second imaging was ≥ 5 days. (4) The patient didn't receive anti-PCP treatment outside the hospital.
Data collection: see figure-1 *Considering that period of the anti-PCP treatment may be too short to affect the results of the G test, the interval of the 2 G tests was required ≥ 5 days. According to the literature, 105pg/ml as a cutoff value is able to distinguish between Pneumocystis jirovecii colonization and P. jirovecii infection. Therefore, the result of the G1 was required ≥ 105pg/ml (11).

Data processing and statistical methods used
The study used the IBM@ SPSS@ Statistics: Version 23 and the MedCalc Statistical Software: version

for the statistical analysis
We compared the baseline data using the t-test, the Mann-Whitney U test and the chi-squared test to determine whether the data differences were significant between the survival group and the death group, based on the data type and its distribution.
The binary logistic regression was used for multivariate analysis to derive the independent risk factors of death. (1) The clinical characters were analyzed by a single factor analysis, and the characters whose P value <0.05 was included in the candidate variables.  Figure 3D).
The patients with the different prognoses had significant differences in the APACHE II score, the pulmonary fibrosis, the CMV infection, the A. baumannii infection and many complications. There were no significant differences in age, sex, length of hospital stay and hypertension (Table 1).

The independent risk factors of death and the effects of the different methods judging the prognosis
The independent risk factors of death: Pulmonary interstitial fibrosis, diabetes, co-infection with the CMV, the complication occurred during the treatment, and the imaging didn't improve: are the 5 risk factors of the PCP. The risk of death in the patients with (G1-G2)/G1≤9.13% was 1.50 (CI 1.13-1.993, P = 0.005) times the risk of death in the patients with (G1-G2)/G1 9.13%. The risk of death in the patients with the pulmonary interstitial fibrosis was 26.11 (CI 3.06-222.86, P = 0.003) times that in the patients without the pulmonary fibrosis. The risk of death in patients with the CMV infection was 9.71 (CI 1.89-49.78, P = 0.006) times that in the patients without the CMV infection. The risk of death in the patients occurred complication was 10.38 (CI 2.51-42.82, P = 0.001) times that in the patients without the complication occurred. The risk of death in the patients with the unimproved imaging was 9.86(CI 2.46-39.51, P = 0.001) times that in the patients whose imaging was improved. See Table 2 for details.
The efficiency of the different methods judging the prognosis: The results are shown in Table 3 and Figure 4. The AUCs corresponding to the 3 methods is 0.74, 0.73 and 0.938, respectively, which are statistically significant. Among them, the AUC value of the third method is the largest. There was no significant difference (P = 0.893) between the efficacy(efficiency) of predicting by the changes of the G test and that by the changes of imaging. But there was a significant difference (P = 0.001) between the AUC of the G test and that of the independent risk factors of death. The efficacy of predicting by risk factors is the most effective one.

Discussion
The efficacy of the G test predicting the outcomes The aim of the study is exploring the clinical value of the changes of the G test in predicting the outcomes in the PCP patients. The reference system was established by being compared with the changes of imaging, the APACHE II score, and the independent risk factors of death, for it is well-known that the changes of the imaging was the most common way used to judge the patients' condition, the APACHE score is related to mortality, and the risk factors are the most closely related to the prognosis. The multivariate logistic regression showed that the APACHE II score was not one of the independent risk factors of death. However, the changes of the G test was included in the 5 risk factors. Moreover, it is suggested by the ROC analysis that the efficacy of predicting outcomes by changes of the G test is comparable to that by the changes of imaging, lower than that of the combination of the risk factors.
The independent risk factors of death in this study is different from that of other studies.
The authors believe that independent risk factors of death in the PCP patients can be divided into 2 categories: the one is disease-specific. For examples, low FiO2, CMV coinfection and pulmonary interstitial fibrosis are the risk factors of the PCP rather than the risk factors of gastrointestinal bleeding. The other one is common for all diseases, like low level of serum albumin, elder, and underlying disease, etc. For instances, the higher the APACHE II score is, the higher probability of death of the patient admitted to the ICU, regardless of what kind of disease the patient suffered. The changes of the G test reflect whether the P. jirovecii were decreased in lung after anti-PCP treatment was received. It is a disease-specific risk factor, and it could not completely predict patients' outcomes because there are interferences of other disease-common risk factors.
The comparison of the AUCs showed that capacity of predicting outcomes by the changes of the G test was similar to that of the lung imaging. However, the G test and the lung imaging have different characteristics and uses. The G test reflects the concentration of (1, 3)-β-D-glucan in patient's blood, and the pulmonary imaging provides more information from the distribution and the morphology of the lesions. When the G test is declining but the lung imaging is not improved, the clinicians need to consider whether the patient is infected by other pathogens. When the G test and the lung imaging both were not improved, the clinicians need to consider whether the P. jirovecii is resistant to drugs. The G test and the pulmonary imaging could be mutually assisted in the clinic.

Possible reason of complication
The data showed that bleeding in different parts was the most common complication, including gastrointestinal bleeding, hemoptysis, cerebral hemorrhage, diffuse alveolar hemorrhage, and abdominal hematoma in 12 (12/104, 11.54%) cases. The second complication is gas accumulation in chest, including mediastinal emphysema, subcutaneous emphysema, and pneumothorax, accounting for 10/104 (9.62%). Since the patients' severe respiratory failure required mechanical ventilation, it's easily understand that gas accumulation in chest is the second common complication. But the reason of that bleeding is the most common complication remains unknown.
There are 2 possible reasons as it follows: one is a deficiency of the vitamin K -it is a case report that the patient had bleeding caused by a deficiency of the vitamin K after taking sulfonamide, PT and APTT returned to normal value after oral administration of sufficient vitamin K (31). Approximately 100% of the PCP patients received the SMZ-TMP. Is it truly the reason of bleeding complication? Another is of the anticoagulant drugs: the patients lying on bed for long time are prone to have deep vein thrombosis and need to use anticoagulant drugs. Was such a bleeding induced by the anticoagulant drugs? Because the study did not collect serum concentration of the Vitamin K and the incident rate of deep vein thrombosis, the real reason needs some further study. It's suggested that occurring complication is one of independent risk factors of death, so reduction of complication may benefit for the PCP patients.

Limitation
The study is a retrospective study and unable to fix the interval between the 2 results of the G tests and pulmonary imaging. The APACHE II score was scored retrospectively, which may bring a little bias.
There is a heterogeneity in the sample population collected in the study, for a significant difference of the APACHE II score showed between the death group and the survival group.
Because the diagnosis of the fungal infection is difficult, the study did not analyze whether the fungal infection is the independent risk factor of death.
The study did not collect the data of serum levels of albumin and the LDH that was mentioned as an independent risk factors in other literature, and did not assess their capacity of predicting prognosis compared with the G test.
Only the 5 variables were chosen and submitted to the binary logistic regression, due to the limitation of the sample size (104). If the sample size was expanded, the other variables might have been more meaningful.

Conclusion
The changes of G test and the changes of imaging are both independent risk factors of PCP, and their efficiency are comparable. Combined application of them could improve the accuracy of predicting prognosis. Clinical doctors could appropriately refer to those results.

Declarations
Ethics approval and consent to participate

Authors' contributions
Xu Jun put forward the idea and designed the experimental framework; Yao Hui collected data, processed data, and drafted the manuscript;  Supplementary Files This is a list of supplementary files associated with the primary manuscript. Click to download.