Intraoperative flurbiprofen alters the expression of PD-1, a molecule of immune checkpoint pathway, in patients undergoing elective thoracoscope resection of lung cancer: a randomized controlled trial

Background: This study aimed to determine the influence of intraoperative use of non-steroidal anti-inflammatory drugs (NSAIDs) flurbiprofen on postoperative level of programmed death 1 (PD-1) in patients undergoing thoracoscope surgery. Methods: In this prospective double-blind trial, patients were randomized to receive intralipid (Control group, n=34, 0.1ml/kg, i.v.) or flurbiprofen axetil (Flurbiprofen group, n=34, 1mg/kg, i.v.) before the induction of anesthesia and 6 hours after the initial injection. PD-1 level on T-cell subsets, inflammation and immune markers in peripheral blood were examined before induction of anesthesia (T0), and after surgery (24 hours (T1), 72 hours (T2) and 1 week (T3)). A linear mixed model was used to examine whether the changes from baseline values (T0) between groups were different during our study. Results: The increases in the percentages of PD-1(+)CD8(+) T-cell observed at T1 and T2 in the control group were higher than in the flurbiprofen group (T1:12.91%±1.65% versus 7.86%±5.71%, P=0.031; T2:11.54%±1.54% versus 8.75%±1.73%, P=0.004) while no difference was observed at T1 and T2 between the groups in terms of change in percentages of PD-1(+)CD4(+) T-cell. Moreover, extensive changes in the percentages of lymphocytes subsets and the concentrations of inflammatory markers was observed at T1 and T2 after surgery, and flurbiprofen seemed to attenuate the most of changes. Conclusion: Perioperative administration of flurbiprofen attenuated postoperative PD-1 increase on CD8(+) T-cell up to 72 hours, but not after this time. The clinical relevance of changes with PD-1 to long-term outcome of surgery is still unknown.

Interestingly, PD-1 and its ligands were markedly inducible by cyclooxygenase (COX) enzyme activity and downstream prostaglandins (PGs) that are prominent in tumorsustaining inflammatory mediators [17,18]. Additionally, non-steroidal anti-inflammatory drugs (NSAIDs), which exert anti-inflammatory and analgesic properties via inhibiting COX enzyme, pharmacologically act in cooperating with anti-PD-1 treatment efficiency in the preclinical cancer model [19]. In clinical settings, perioperative use of flurbiprofen, regularly prescribed as perioperative analgesic, efficiently elicited a short-term increase of innate and adaptive immune cells in postoperative peripheral blood from patients [20].
Because less is known about how NSAIDs act to adjuvant synergize with anti-PD-1 therapy in clinical settings, we aimed to determine if flurbiprofen had an effect on postoperative PD-1 level in blood circulating T-cell subset with patients undergoing elective thoracoscope resection of NSCLC. 2) history of peptic ulceration; 3) blood coagulation disorder; 4) severe cardiac, hepatic or renal dysfunction; 5) perioperative blood transfusion; 6) bronchial asthma; 7) current or recent receipt of radiotherapy, chemotherapy, immunedepressant, glucocorticoid; 8) autoimmune disease or acute inflammation; 9) severe hypertension or diabetes mellitus; 10) pregnancy; 11) use of enoxacin, lomefloxacin, or norfloxacin; 12) duration of operation less than 120 min.
Patients were mechanically ventilated with suitable ventilation parameters to maintain the end-tidal carbon dioxide in a range of 35-45 mmHg. In addition, an appropriate depth of anesthesia was monitored by Narcotrend. Postoperatively, all patients received the same regimen of patient-controlled intravenous analgesia (PCIA) with sufentanil (100µg, diluted to a total volume of 100ml with 0.9% sodium chloride). PCIA was performed with a loading dose of 2ml, a background infusion of 2ml/h, a bolus of 2ml, and a lockout time of 15 minutes. In addition, tramadol as the component of rescue analgesic was required postoperatively for unbearable pain when VAS (Visual Analogue Scale) score greater than or equal to 5-point.

Intervention and Randomization
Using a computer-generated random number sequence, patients were allocated in a 1:1 ratio to receive treatment with either flurbiprofen (flurbiprofen axetil, 1mg/kg, i.v., Bei Jing Tai De Medicine Co., Ltd. Batch number: 1E016R; Beijing, China) or placebo (intralipid, 0.1ml/kg, i.v.) before the induction of anesthesia and 6 hours after the initial injection. Data collection was performed by an independent researcher who was not involved in the trial. In addition, another researcher was in charge of the preparation of study drugs. The study drugs were placed into unmarked syringes and treatment assignments concealed in sealed opaque envelopes; all of which was blinded to patients, anesthetists and other investigators involved in the study. The statistician was unaware of the assignments until all data analyses were completed.

Outcomes
Venous blood (2ml) was obtained from the non-infused peripheral vein before the induction of anesthesia (T 0 ), postoperative 24 hours (T 1 ), 72 hours (T 2 ), and 1 week (T 3 ) after surgery. The blood was preserved in EDTA anticoagulation tube at 4°C for subsequent testing within 24 hours. The primary outcome was counts of circulating PD-1 (+ ) CD8 (+ ) T-cell or PD-1 (+ ) CD4 (+ ) T-cell at each time point perioperatively. Secondary outcome was the percentages of lymphocyte subsets in peripheral blood mononuclear cells (PBMCs) and concentrations of inflammatory markers in peripheral blood including tumor necrosis factor-α (TNF-α); interferon-γ (IFN-γ); interleukin-6 (IL-6); C-reactive protein (CRP). Moreover, the data in the full blood count test including platelet count, total WBC (white blood cell) count, hemoglobin, neutrophil count, lymphocyte count, monocyte count, eosinophil count, basophil count and serum vascular endothelial growth factor (VEGF)-A were also tracked simultaneously.

Measurement of Inflammatory Cytokines
The concentrations of TNF-α, IFN-γ, IL-6 and VEGF-A were measured by enzyme-linked immunosorbent assay (ELISA) using a commercially available kit (R&D, Minneapolis, MN, USA) as per the manufacturer's protocol.

Full blood count, Serum glucose and CRP Measurement
The analysis of full blood count was performed on the EDTA peripheral blood sample using a hematology analyzer instrument (BC-6900, Mindray, China). In addition, the concentrations of serum glucose and CRP were tested from peripheral blood using the Clinical Chemistry System (ADVIA 2400, SIEMENS) and i-CHROM laser fluorescence reader, respectively.

Sample Size Calculation
The number of patients recruited for groups was based on the primary outcome by testing percentage of PD-1 (+ ) CD8 (+ ) T-cell from a previous pilot study. According to these pilot data, a difference in mean PD-1 (+ ) CD8 (+ ) T-cell percentages between two groups was 2.08 % and a pooled SD was 2.57 %. To detect a difference of this magnitude between study groups, 32 patients were required in each group to provide 90% power at an alpha of 0.05.
However, considering a loss-to-follow-up rate of about 10% in our study, we aimed to enroll a total sample size of 70 patients.

Statistical Analyses
A Shapiro-Wilk test was applied to assess the normality of continuous data. The normally distributed continuous data were recorded as mean (SD) and analyzed using the independent t-test. The skewed data were presented as the median (interquartile range), and the comparison were treated with the Mann-Whitney U test or Wilcoxon rank-sum test.
The categorical variables were reported as frequency and compared using the chi-square or Fisher's exact test. Changes in percentages of PD-1 (+ ) CD8 (+ ) T-cell or PD-1 (+ ) CD4 (+ ) Tcell and concentrations of inflammatory markers were calculated between baseline values at T 0 and those at T 1 , T 2 and T 3 . Changes in both groups were analyzed with a mixed linear regression model adjusted for age and ASA status (i.e., a random effect was introduced in an effort to account for the repeated measures) in order to validate the findings of the univariable analysis. In our model, the group and time were applied as predictors, the interaction between group and time as fixed effects and patient as a random effect, and an interaction term was taken into account to test whether the change over time was different in both groups. In the univariable analyses, a Bonferroni correction was applied to post hoc analysis to compare different time changes. In addition, a sensitivity analysis using a linear mixed model with scale identity correlation matrices was applied to affirm vital differences between groups. All statistical analyses were performed on SPSS 19.0 software and all reported P values less than 0.05 were indicated to be of statistical significance.

Study Population
The CONSORT diagram is shown in Figure 1. A total of 70 patients were screened for our study, with 68 patients ultimately participating. The 2 patients not included were 1 patient who refused to participate and 1 patient who did not meet inclusion criteria. 4 patients (2 in flurbiprofen group, 2 in control group) were withdrawn from the statistical analysis although they had been randomized into two groups (one patient in flubiprofen group refused phlebotomy for laboratory testing after the operation; moreover, one patient was withdrawn because of more than 1.5 liter blood loss in the flurbiprofen group; two patients in the control group due to the tumor metastasis invasion into the pleura that did not receive lung parenchyma resection operation procedure). Baseline and intraoperative characteristics of patient in both groups are shown in Table 1. No significant differences were observed between groups in terms of age, body mass index, gender, and ASA status.
There were also no significant between-group differences in surgery type, duration of surgery, the use of antibiotic and propofol consumption intraoperatively. The consumption of remifentanyl in the control group was higher compared with the flurbiprofen group.
Outcome values at baseline as shown in supplementary material (supplementary material, Table 1 and Table 2) were compared between groups with no significant differences observed in terms of percentage of PD-1 (+ ) CD8 (+ ) T-cell, PD-1 (+ ) CD4 (+ ) T-cell, T lymphocyte subsets, VEGF-A and other inflammatory markers concentrations. Table 2, the increases in the percentages of PD-1 (+ ) CD8 (+ ) T-cell observed at 24h postoperatively (T 1 ) and 72h postoperatively (T 2 ) in the control group were higher than in the flurbiprofen group (T 1, 12.91%±1.65% versus 7.86%±5.71%, P=0.031; T 2, 11.54%±1.54% versus 8.75%±1.73%, P=0.004)(Supplementary material, Figure 2, the original images regarding the PD-1 (+ ) CD8 (+ ) T-cell at different time points) while no difference was observed at T 1 and T 2 between the groups in terms of change in percentages of PD-1 (+ ) CD4 (+ ) T-cell (Supplementary material, Figure 3, the original images regarding the PD-1 (+ ) CD4 (+ ) T-cell at different time points). However, changes in the percentages of PD-1 (+ ) CD8 (+ ) T-cell from baseline (T 0 ), as well as PD-1 (+ ) CD4 (+ ) T-cell, were similar between groups at 1 week after surgery (T 3 ) . Moreover, as shown in Table 3, the decreases observed for all lymphocyte subsets in the control group were markedly greater at T 1 and T 2 compared to those in the flurbiprofen group except CD8 (+ ) T-cell (significant change only observed at T 1 ). In addition, the observed significant increases in concentrations of TNF-α, IFN-γ, IL-6, CRP and VEGF-A at T 1 and T 2 was much greater in the control group than those in the flurbiprofen group. For longer postoperative follow-up (T 3 ), no significant differences were observed in the above-mentioned data from baseline.

As shown in
Except above time points, we also collect the 3 weeks (T 4 ) data after the surgery, however, there was no significance between groups regarding all relevant indicates (Table   data not shown). We just show the expression of PD-1 (+ ) on CD8 (+ ) T-cell and CD4 (+ ) T-cell in two groups at T 4 (Supplementary material, Figure 2 and Figure 3).
There were no significant differences between groups with respect to the proportion of patients who were unexpectedly subjected to respiratory depression after surgery, defined as a respiratory rate less than 8 breaths per minute and oxygen saturation either below 92% or a decrease of more than 5% from baseline in patients with a baseline of SPO 2 less than 90% [21] .There was also no significant between-group difference in the rate of nausea and skin pruritus. Eight patients in the control group suffered vomiting and retching, compared with one patient in the flurbiprofen group (P=0.026). Furthermore, the use of anti-emetics in the control group was higher than those in the flurbiprofen group (P=0.011). A significant reduction in 24h postoperative score for cough was observed in the flurbiprofen group but no difference was observed between groups 72h postoperatively. Moreover, ten patients in the control group required postoperative rescue analgesia (tramadol) for unbearable pain, as compared with three patients in the flurbiprofen group (P=0.030). In addition, no difference was observed between the groups in terms of wound infection and the length of hospital stay postoperatively.

Sensitivity analysis
Linear mixed models confirmed the significant differences in the percentages of PD-1 (+ ) on CD8 (+ ) T-cell, CD4 (+ ) T-cell, NK cells and the concentrations of TNF-α, IL-6, CRP between patients who received flurbiprofen and placebo (Supplementary material, Tables 3 to 8), and the significant differences was observed at T 1 and T 2 .

PD-1, expressed in tumor-infiltrating T-cells and circulating T-cells, has been shown to
predict prognosis as well as a target-therapy candidate in several malignant tumors, including NSCLC [8,12,13,16,[22][23][24][25][26]. In peripheral blood, the latest evidence indicates a higher level of PD-1 on circulating CD8 (+ ) T-cell to be correlated with a worse clinical outcome and shorter overall survival [10]. Interestingly, it is has been reported that COX inhibitors can act synergistically with immune-checkpoint blockade, implying that NSAIDs commonly used as perioperative analgesic could be a useful adjuvant for anti-PD-1/anti-CTLA-4 therapies in cancer patients [19]. To the best of our knowledge, this is the first clinical study providing evidence that NSAIDs alter postoperative level of PD-1 resulting in a smaller increase of PD-1 on CD8 (+ ) T-cell in peripheral blood from lung cancer patients undergoing resection surgery.
Since the COX activity and COX-dependent inflammatory mediators such as PGs facilitate the increase of inhibitory immune-checkpoints PD-1/CTLA-4, and low levels of PD-1/CTLA-4 were confirmed in COX-2 MEC knock-out mice bearing tumors, suggesting COX activity and downstream PGs have some potential link with PD-1 [18]. Although NSAIDs could act in cooperating with anti-PD-1 blockade in inducing eradication of tumors in preclinical study [19], our study primarily identified flurbiprofen could alter PD-1 level on the circulating CD8 (+ ) T-cell population from NSCLC patients until 72 hours postoperative, without any change for the percentage of circulating PD-1 (+ ) CD8 (+ ) T-cell observed after this time point. Flurbiprofen had little influence on postoperative percentage of PD-1 (+ ) CD4 (+ ) T-cell during the overall perioperative period. The lymphocyte percentage after surgery in the control or flurbiprofen groups was, in general, lower compared with hospital reference values. It is interesting to note that there were extensive changes in lymphocyte subsets and inflammatory markers following administration of flurbiprofen in the short-term up to postoperative 72 hours.
A possible explanation for the observed change between the groups is that prominent tumor-sustaining inflammatory factors were considered as a potent stimulator to induce PD-1/CTLA-4. This highlights that PD-1 and other inhibitory checkpoints levels involved in CD8 (+ ) T-cell exhaustion were markedly enhanced with high levels of VEGF produced by a pro-angiogenic factor in the tumor's microenvironment [27]. Moreover, COX-derived PGE2 promotes tumor progression by sustaining angiogenesis through induction of VEGF that is largely required for development of a stable blood supply for tumor growth [28,29]. We speculate that inhibition of COX and PGE2 by flurbiprofen alleviates the increase of PD-1 partly due to abrogating the induction of VEGF.
There are several limitations. First, further studies are needed to focus on the exact mechanism of how NSAIDs down regulate the anti-tumor immunity and immune escape.
Second, although the randomization in our study was strict, some baseline and perioperative factors were not equal between the two groups. Third, we didn't subdivide the CD4 (+ ) T-cell which should be discriminated between conventional CD4 (+ ) T-cell and regulatory CD4 (+ ) T-cell. Fourth, our study is a single-centre investigation, a large multicenter study would be ideal to confirm our findings.

Conclusion
In conclusion, perioperative administration of flurbiprofen appears to attenuate PD-1 increase on CD8 (+ ) T-cell until 72 hours postoperative, with no effect identified after this time.

Availability of data and materials
The datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Authors' contributions
The conception and the design of the study: XQ Chai, D Wang. Patient's data collection: LX Xie, SH Shu. The statistical analysis: JC Hu, SS Hu. Interpretation of results and wrote the manuscript: JC Hu, D Wang. Review of this manuscript: XQ Chai, C. M.

Ethics approval and consent to participate
The protocol of this study was approved by the Biomedical Research Ethics Committee of Anhui Medical University and and written informed consents have been obtained from all patients.  All continuous data are presented as mean (SD) or median (interquartile range, IQR) and compared using the independent t-test or the Mann-Whitney U test, respectively.

Consent for publication
Categorical data are presented as frequency and analyzed using the chi-square or Fisher's exact test. BMI=body mass index. ASA=American Society of Anesthesiologists. VAS= visual analogue scale. NA=not applicable. Table 2. Changes in perioperative PD-1 (+ ) on CD4 (+ ) and CD8 (+ ) T-cell after receiving flurbiprofen or placebo at 24 hours, 72 hours and 1 week from the baseline values before induction of anesthesia.

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download.