CD4/CD8 Ratio of Pleural Effusion Is A Prognostic Predictor for Non-Small Cell Lung Cancer Patients Under Immune Checkpoint Inhibitor Treatments

Pleural effusion is a rare immune-related adverse event for lung cancer patients receiving immune checkpoint inhibitors (ICIs). We enrolled 281 lung cancer patients treated with ICIs and 17 were analyzed. We categorized the formation of pleural effusion into 3 patterns: type 1, rapid and massive; type 2, slow and indolent and type 3, with disease progression. CD4/CD8 ratio of 1.93 was selected as the cutoff threshold to predict survival. Most patients of types 1 and 2 effusions possessed pleural effusion with CD4/CD8 ratios > = 1.93. The median OS time in type 1, 2, and 3 patients were not reached, 24.8, and 2.6 months. The median PFS time in type 1, 2, and 3 patients were 35.5, 30.2, and 1.4 months. The median OS for the group with pleural effusion CD4/CD8 > = 1.93 and < 1.93 were not reached and 2.6 months. The median PFS of those with pleural effusion CD4/CD8 > = 1.93 and < 1.93 were 18.4 and 1.2 months. In conclusion, patients with type 1 and 2 effusion patterns had better survival than those with type 3. Type 1 might be interpreted as pseudoprogression of malignant pleural effusion. CD4/CD8 ratio > = 1.93 in pleural effusion is a good predicting factor for PFS.


Introduction
Immune checkpoint inhibitors (ICIs) have become promising agents against a variety of cancers. However, in some patients, concomitant immune-related adverse events (irAEs) develop. Among organs affected by immune checkpoint blockade, pleural involvement is rare. Under ICI treatment, pseudoprogression may develop, with a transient increase in the tumor size before regression 1 . Pseudoprogression in lung-cancer patients occurs not only in the solid part of the tumor, but also has been reported in malignant spread to pleural and pericardial space with the presentation of rapidly accumulating recurrent effusions 2 . The clinical course and outcomes of patients receiving ICIs followed by pleural effusion development are poorly known.
IrAEs involving different organs may result from various mechanisms 3 . For example, in myocarditis, the in ammatory in ltration of T cells is predominantly CD8 4 , whereas in pericardial involvement, T cell in ltration is predominantly CD4 5 . Which types of lymphocytes are involved in pleural effusion under ICIs remained unknown.
In the present study, we aimed to categorize the clinical presentations of ICI-related pleural effusion and analyze the lymphocyte components in the pleural effusion in relation to the clinical outcomes of non-small cell lung cancer (NSCLC) patients receiving ICIs.

Methods
Study design NSCLC patients were enrolled retrospectively at Taichung Veterans General Hospital from Oct 2015 to Dec 2019, during which ICI treatments were initiated.
The last follow-up was on May 31, 2020. Eligible patients all had non-infectious pleural effusion after ICI use. The exclusion criteria were as follows: no cytology results of pleural effusion, mortality of unknown etiology, specimens for lymphocyte analysis not from pleural effusion, the duration from last dose of ICI to development of pleural effusion exceeded 12 months. Patients receiving ICIs and docetaxel were excluded, since docetaxel was known to cause pleural effusion 6 . This study was approved by the Institutional Review Board of Taichung Veterans General Hospital (IRB No. CF16018A). Written informed consents for clinical data records, genetic and immunological testing were obtained from all patients. All methods were carried out in accordance with the relevant approved guidelines and regulations.
Identi cation of driver mutations and PD-L1 assay Tumor specimens were procured for oncogenic mutation analyses as previously reported 7 . Five oncogenic drivers, including EGFR, KRAS, BRAF, HER2 and   EML4-ALK, were tested. For patients with squamous cell carcinoma, oncogenic mutation analyses were not routinely performed.   Three commercial Programmed Death-ligand 1 (PD-L1) IHC assays, 22C3, SP142, and SP263, were performed for all patients when adequate specimens were  available. The PD-L1 IHC 22C3 pharmDx was conducted on the DAKO Autostainer Link 48, while the Ventana PD-L1 SP142 and SP263 assays were conducted on the Ventana BenchMark platform.

Data records and response evaluation
Clinical data of individual patients included age, gender, Eastern Cooperative Oncology Group Performance Status (ECOG PS), tumor stage, smoking status, and thyroid function. The age, ECOG PS, and tumor stage were evaluated while ICIs were initiated. The overall survival (OS) and progression free survival (PFS) were analyzed from the beginning of ICI treatment. TNM (tumor, node, and metastases) staging was performed according to the 8th edition of the American Joint Committee for Cancer (AJCC) staging system. We adopted here unidimensional measurements as de ned by Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1.

Statistical methods
Fisher's exact test and Mann-Whitney U test were used to compare inter-group differences for categorical and continuous variables as appropriate. Univariate and multivariate Cox proportional hazard regression models were used to estimate the hazard ratio. The OS and PFS were estimated using the Kaplan-Meier method, whereas the between-group differences were assessed using the strati ed log-rank test. Two-tailed tests with p values < 0.05 were considered statistically signi cant.
All analyses were performed with the IBM SPSS Statistics package, version 23 (IBM Corporation, Armonk, NY).

Patient characteristics
We included a total 281 advanced (stage IIIb/IV) NSCLC patients with ICIs initiated. Among these patients, 168 patients were treated with pembrolizumab, 43 with nivolumab, 47 with atezolizumab, and 23 with durvaluamb. Among them, 27 developed pleural effusion after ICI use, with 10 excluded. Among the remaining 17 patients, three were categorized as type 1, 5 as type 2, and 9 as type 3 ( Fig. 1).
Their descriptive characteristics are summarized in Table 1. All the patients had reached advanced stages of lung cancer before ICI use. Adenocarcinoma was diagnosed in 13 patients, and three of them harboring EGFR mutations. Negative PD-L1 expression was found in 6 patients, low PD-L1 expression in 4 patients, and high PD-L1 expression in 5 patients. Of the 17 patients, 13 received lymphocyte subset analyses of their pleural effusions within one month after rst thoracentesis. Clinical data and outcomes of the enrolled patients are shown in Table 2. The disease course and treatment timeline for patients of types 1 and 2 are shown in Fig. 2. Type 1 patients within 2 weeks after ICI use developed pleural effusion with or without pericardial effusion (one patient also received pericardial drainage), and pigtail catheters were applied to two patients. All these patients presented with pleural effusion before ICI use and one of them was found to have malignant pleural effusion. Malignant cells were found in pleural effusion after the initial treatment, but were then absent in the following serial thoracentesis. Type 2 patients developed pleural effusion one month after ICI use. Malignant pleural effusion was not documented before or after ICI use.
Among type 3 patients, malignant pleural effusion persisted in 6 patients while disease progression to other organs was found in three patients. Expression levels of cytokines in pleural effusion were shown in online supplemental Fig. 4. IL-8 levels in pleural effusion of patients with CD4/CD8 ratios < 1.93 were higher than those with ratio > = 1.93 (online supplemental table 3A). Expression levels of cytokines were however similar across effusion types (online supplemental table 3B).

Discussion
We have here found different presentations of lung cancer patients developing pleural effusion after receiving ICI. Three effusion developmental patterns were identi ed. Type 1 patients developed massive effusion within one month after initiating ICI treatment, usually within two weeks. The rst time cytological examinations of thoracentesis after treatment revealed positive for malignancy in all these patients. Their development of effusion could be interpreted as "pseudoprogression", because the cytological examinations turned negative in the serial thoracentesis afterwards. Fulminant effusion development was resolved within two months after ICI use.
Most researchers reported survival bene ts of pseudoprogression markedly better than that of typical progression [10][11][12][13] . In our study, type 1 patients had longer PFS and OS than those of type 3 and type 2 patients. Some studies reported that the malignant pleural effusion present before anti-PD-1 treatment is associated with shorter PFS and OS 14 . In our study, if pseudoprogression occurred as type 1 pleural effusion, long-term survival could be achieved. Therefore, ICI should still be considered in patients with malignant pleural effusion.
Kolla et al. reported similar cases in which pseudoprogression was suspected after nivolumab administration 2 . One patient developing massive pleural effusion had frequent thoracentesis for 8 weeks after nivolumab use. Cytological examination from pleural effusion was positive for malignancy. Nivolumab continued and there was a complete response. No more drainage was recorded after the rst two months of therapy. That case shared a similar clinical presentation with our type 1 patients and may be categorized as "type 1" pleural effusion.
Type 2 pleural effusion developed one month after ICI treatment had begun. The time from treatment to rst thoracentesis was as long as 25 months (case no.8). It is not surprising that the occurrence of irAEs was delayed, since Nigro et al. reported earlier that late-irAEs (after 12 months) are common (incidence 30.3%) in long responders to ICIs 15 . The cytology of pleural effusion was never documented positive before or after ICI use so this type was not categorized as pseudoprogression. Thoracentesis was usually infrequent. The exception was an atypical case no.7 who developed massive pleural effusion. Cytological results were variable, with alteration of positive or negative ndings of malignancy, and subsequent infection of the effusion was also noted.
We identi ed those with CD4/CD8 ratios of pleural effusion > = 1.93 were well predicted for their survival. Though the initial CD4/CD8 ratio of patient no.1 (type 1) was only 1.89, serial pleural effusion CD4/CD8 ratios examined showed progressively elevated ratios, with the highest reaching 13.8. It is worth noting that the B cell ratio of pleural effusion was also elevated from 0.4-22.1% in serial analyses (Fig. 2). For such patients presented with the typical type 1 or type 2 pleural effusions with CD4/CD8 ratios < 1.93, serial follow up is recommended, because elevation of the ratio may indicate a good response to the ICI treatment. In type 1 patients, elevated B cell percentage is the feature distinguishing them from types 2 and 3 patients. CD4/CD8 ratio of which is higher than the peripheral blood. This may be a defensive mechanism against cancer, and ICI likely reinforces the mechanism.
Regarding irAEs, cytokines or chemokines in response to ICIs have been studied. Khan et al. reported irAEs patients have initially low levels of CXCL9, 10, 11 and 19, but levels of CXCL 9, and 10 remarkably increase after treatment compared with those patients without irAEs 18 . Lim et al. found elevations of 11 cytokines in patients with severe irAEs, and even introduced a cytokine toxicity score 19 . IL-17 and IL-6 levels were reported as biomarkers in predicting irAEs 20,21 . In our study, we found IL-8 levels in patients with pleural effusion CD4/CD8 ratio < 1.93 were higher than those with ratio > = 1.93. IL-8, a chemokine produced by cancer cells, could play a role in cancer microenvironment. Higher IL-8 levels are correlated with poor prognosis 22 . Only one patient from the type 1 group has a higher level of IL-17. We also examined several other cytokines including IL-1, IL-2, IL-4, IL-6, IL-12p70, INF-γ, and TNF-α. However, the levels of these cytokines are either under detection limit or demonstrate no signi cant difference among the three types of patients.
There were several limitations of our study. First, its sample size was small, and was conducted retrospectively in a single medical center. Second, not all patients had their CD4/CD8 ratios determined at the initial thoracentesis. Also, their CD4/CD8 ratios were not determined before ICI treatment nor their ratio in the peripheral blood. Third, 11 of 17 patients received both chemotherapy and ICI, presenting a confounder on response evaluation. However, no patient was lost during follow up and all required clinical information was collected. We are the rst to report two distinct types of pleural effusions after ICI use. These two types of patients both had relatively good prognosis. Our study is also the rst to use the CD4/CD8 ratio in pleural effusion to predict patient survival after ICI use.
In conclusion, beside pleural effusion due to disease progression (type 3), two distinct effusion types were identi ed after ICI use: type 1, rapid (develop < 1 month) and massive and type 2, slow (develop > = 1 month) and relative indolent. Both types showed better overall and progression free survival than type 3.
Type 1 could be interpreted as pseudoprogression of malignant pleural effusion. CD4/CD8 ratio > = 1.93 in pleural effusion after ICI use is a good predicting factor in PFS. In most patients of types 1 and 2, their CD4/CD8 ratios > = 1.93 in pleural effusion. In those patients presented with typical type 1 or type 2 pleural effusion but with CD4/CD8 ratios < 1.93, serial follow up is recommended because elevating ratio may indicate a good response to ICI.