Pembrolizumab with or without Concurrent Chemotherapy in Metastatic Non–Small-Cell Lung Cancer with High Programmed Death Ligand 1 Expression

Hong Kong J Radiol 2024;27:Epub 19 November 2024

 

 

¹ Department of Oncology, Princess Margaret Hospital, Hong Kong SAR, China

² Clinical Research Centre, Princess Margaret Hospital, Hong Kong SAR, China

 

  Full Article in PDF

Lung cancer is the leading cancer in Hong Kong.[1] Although epidermal growth factor receptor (EGFR) mutation and anaplastic lymphoma kinase (ALK) translocation are especially common in the Asian population, approximately half of lung cancer patients suffer from disease without an actionable driver mutation.[2] [3]

 

Since the release of the KEYNOTE-024 (KN-024) study, first-line pembrolizumab monotherapy has become one of the standards of care for metastatic non–small-cell lung cancer (NSCLC) with high expression of programmed death ligand 1 (PD-L1), defined as having a tumour proportion score ≥50%.[4] [5] Subsequently, the KEYNOTE-189 (KN-189) and KEYNOTE-407 (KN-407) trials found that a pembrolizumab-chemotherapy combination was effective regardless of the level of PD-L1 expression.[6] [7] [8] [9] [10]

 

For those with PD-L1 expression of ≥50%, pembrolizumab alone has a response rate of approximately 45% (44.8% in the KN-024 trial).[4] The KN-189 and KN-407 trials found that adding chemotherapy to pembrolizumab increased the response rates to >60% in both non-squamous (62.1% in the KN-189 trial)[7] and squamous cell carcinoma (64.4% in the KN-407 trial),[10] although the survival outcomes were comparable. From the latest 5-year update report,[8] the median overall survival (OS) and 5-year OS rates were similar with pembrolizumab monotherapy (26.3 months and 31.9%),[5] combined with pemetrexed and carboplatin in non-squamous carcinoma (27.7 months and 29.6%),[8] and combined with paclitaxel/nab-paclitaxel and carboplatin (19.9 months and 23.3%).[10] Therefore, the optimal choice of first-line treatment remains unsettled.

 

There is an unmet need to identify those who will be durable responders, as well as those who are expected to have futile and non-sustained responses to pembrolizumab treatment. Early combination with chemotherapy as an intensified treatment may benefit some patients, whereas early symptomatic care may be more appropriate. Careful selection and prudent decisions should be made together with patients after considering all available factors.

 

Clinical, biochemical, pathological, and radiological parameters are assessed for their predictive value as well as their adverse effects. Consensus on the preferred markers is difficult to arrive at in many tumours due to the complex interplay between tumour microenvironment and host immune system.[11] [12] The proinflammatory status of the patient may promote cancer cell progression and immune resistance.[13] Clinical-biochemical parameters have been proposed and some are commonly used to reflect a dysfunctional host immune state, such as a neutrophil-to-lymphocyte ratio (NLR) ≥5, a derived neutrophil-to-lymphocyte ratio (dNLR) ≥3, a platelet-to-lymphocyte ratio (PLR) ≥200, and an advanced lung cancer inflammation index (ALI) ≤18. Utilisation of such indices can serve as a simple but effective tool to assist clinical decision.[14] [15] [16] [17] [18] [19] [20]

 

We aimed to review the clinical outcomes of patients with metastatic NSCLC with high PD-L1 expression receiving first-line pembrolizumab with or without concurrent chemotherapy and to evaluate the utility of several commonly used clinical-biochemical prognostic indices.

 

This was a retrospective cohort study conducted in a single tertiary oncology centre in Hong Kong, which provides cluster-based oncology services to the most densely populated districts in the city. The cases of patients aged ≥18 years with pathologically confirmed metastatic NSCLC, without sensitising EGFR mutation or ALK translocation, and with ≥50% PD-L1 expression, who had received at least one cycle of pembrolizumab with or without concurrent chemotherapy in a first-line setting between January 2018 and December 2022 were included. Cases with Eastern Cooperative Oncology Group (ECOG) performance status score of ≥3, and baseline blood results, body weight, or height not recorded within 28 days from the first cycle of pembrolizumab were excluded. The case list was generated from the Clinical Data Analysis and Reporting System of the Hospital Authority.

 

Both electronic and physical records were reviewed. Relevant demographic, clinical, laboratory, treatment, and outcome data were extracted. The tumour proportion scores of PD-L1 were analysed by immunohistochemistry using 22C3 or SP263 antibodies. Response evaluation was performed as per the physician’s assessment and investigator’s review based on clinical symptoms, physical examination, chest radiography, computed tomography, and carcinoembryonic antigen levels.

 

The data cut-off date was 30 April 2023. The duration of follow-up was calculated from the date of initiating the first cycle of pembrolizumab to the date of death or the date of the last follow-up if the date of death was unavailable.

 

The primary endpoint was OS, which was defined as the time from the beginning of the first cycle of pembrolizumab to the date of death from any cause or the last date of follow-up if the date of death was unavailable. Only one patient was lost to follow-up.

 

The key secondary endpoints included the following: (1) time on treatment (ToT), which was defined as the time of the initiation of the first cycle of pembrolizumab to the last efficacious date after discontinuation (date of last pembrolizumab injection plus 21 or 42 days for every 3- or 6-week regimen, respectively) for any reason or death, whichever was earlier; (2) reasons for treatment discontinuation; (3) calculation of the four prognostic indices, namely NLR, dNLR, PLR and ALI, where baseline blood results within 28 days of pembrolizumab were used for analysis (online supplementary Table); and (4) negative prognostic factors for OS, including concurrent chemotherapy, male sex, age ≥70 years, ever-smoker, ECOG performance status score of 2, histology of squamous cell carcinoma, baseline brain metastasis, baseline pleural metastasis or pleural effusion, baseline liver metastasis, previous radiotherapy treatment, and NLR, dNLR, PLR and ALI at their respective cut-offs.

 

Differences in baseline characteristics between the pembrolizumab alone (immunotherapy [IO]-alone) group and concurrent platinum-based chemotherapy (IO-combination) group were compared with the Pearson’s Chi squared test or Fisher’s exact test for categorical variables and an independent t test or the Mann-Whitney U test for continuous variables. OS and ToT were calculated using the Kaplan-Meier method and compared with the log-rank test.

 

Time-dependent receiver operating characteristic (ROC) curves were used to evaluate the discriminative ability of the four prognostic indices on OS. The time-dependent area under the curve (AUC), sensitivity, and specificity with 95% confidence interval (95% CI) were calculated at 6 months, 1 year, and 2 years after therapy initiation. The cut-off of the prognostic indices was determined by the highest achievable sensitivity and specificity based on stepwise testing of incremental improvement. Sensitivity ≥70% for 6-month and 1-year OS prediction was required, and specificity ≥60% for 6-month and 1-year OS prediction was attempted. Escalation of sensitivity and specificity testing at 5% increments were attempted alternately to achieve the best cut-off. After achieving a minimum sensitivity of ≥75% and specificity of ≥65% for 6-month and 1-year OS prediction, further increments focused on 1-year OS prediction to attain the highest achievable combination of sensitivity and specificity when selecting the representative cut-off.

 

Univariate Cox proportional hazards regression analyses were performed to identify potential prognostic factors associated with OS. The association of OS with each prognostic index was examined using a multivariable Cox regression model adjusting for chemotherapy and other factors with p value < 0.1 in the univariate analyses. The variance inflation factor was used to check multicollinearity among independent variables in the regression model, and the Akaike information criterion was used to compare the performance of different regression models.

 

All statistical analyses were carried out using RStudio version 4.3.1 with packages ‘survival’, ‘timeROC’, and ‘performance’. A p value < 0.05 was considered statistically significant.

 

After excluding 9 cases with ECOG performance status score ≥3, and 31 cases without baseline blood results, body weight, or height recorded within 28 days from the first cycle of treatment, a total of 133 cases of pathologically confirmed metastatic NSCLC with ≥50% PD-L1 expression, and no sensitising EGFR mutation or ALK translocation, who had received at least one cycle of pembrolizumab with or without concurrent chemotherapy in a first-line setting were included. Molecular tests on EGFR and ALK were routinely performed for all adenocarcinoma and NSCLC but not squamous cell carcinoma as per institutional practice. Additional molecular tests were arranged based on the individual physician’s decisions. All tested patients were negative for EGFR exon 19 deletion, EGFR exon 21 L858R mutation, and ALK translocation. Fifteen patients had rare mutations detected, including ROS1 translocation (n = 1), EGFR exon 20 insertion (n = 3), human epidermal growth factor receptor 2 exon 20 insertion (n = 1), rearranged during transfection rearrangement (n = 1), KRAS (Kirsten rat sarcoma virus) mutation (G12C, G12D/S, and G12X; n = 3), MET (mesenchymal epithelial transition receptor) exon 14 skipping (n = 5), and co-mutation of KRAS G12C mutation and MET exon 14 skipping (n = 1).

 

Among the 133 consecutive cases, 112 received pembrolizumab alone and 21 received pembrolizumab with chemotherapy. Baseline characteristics are summarised in Table 1. Both groups were balanced except for age and radiotherapy treatment prior to pembrolizumab. The combination group was younger (mean age, 64.9 ± 6.1 years; p < 0.001) and had less radiotherapy before pembrolizumab (n = 2, 9.5%; p = 0.003).

 

Table 1. Baseline characteristics by treatment groups

 

The dosing and frequency of pembrolizumab varied and depended on financial and funding issues. A total of 78.9% (n = 105) of patients started one of two standard fixed-dose regimens (94 at 200 mg every 3 weeks and 11 at 400 mg every 6 weeks), while the remaining 21.1% (n = 28) received a weight-based regimen (three at 80 mg, 24 at 100 mg, and one at 120 mg every 3 weeks). Twenty-eight patients had a change of dose and frequency during the treatment course. Most patients changed from every 3 weeks to every 6 weeks regimen for easier logistics and less frequent hospital visits.

 

Twenty-one patients received concurrent chemotherapy with the choice of agents depending on histological subtypes, where 18 of them received concurrent pemetrexed and carboplatin (16 had adenocarcinoma, one had large cell carcinoma, and one had NSCLC not otherwise specified) with a median number of five cycles (range, 1-10). The remaining three patients received concurrent paclitaxel and carboplatin (one had squamous cell carcinoma, one had lymphoepithelial-like carcinoma, and one had NSCLC not otherwise specified) with a median number of five cycles (range, 3-5).

 

The median follow-up time was 10.2 months (range, 0.2-75). One patient was lost to follow-up. At the time of analysis, 69 deaths had occurred, and 88 patients had discontinued pembrolizumab. Median OS and ToT of the entire cohort were 17.8 months (95% CI = 13.4-23.5) [Figure 1a] and 8.0 months (95% CI = 5.5-12.0) [Figure 1b], respectively. No significant difference was observed between IO-alone and IO-combination groups (median OS = 18.2 months [95% CI = 12.6-38.9] vs. 17.8 months [95% CI = 5.2 to not applicable], p = 0.792; median ToT = 8.1 months [95% CI = 5.0-12.4] vs. 7.4 months [95% CI = 3.2-14.8], p = 0.863) [Figure 1c and 1d].

 

Figure 1. Kaplan–Meier estimates of overall survival (OS) and time-on-treatment (ToT) of all subjects and by treatment groups. (a) OS of all subjects. (b) ToT of all subjects. (c) OS by treatment group. (d) ToT by treatment group. No difference was observed between treatment groups. Tick marks indicate censoring of data at the last time the patient was known to be alive.

 

Reasons for discontinuation of pembrolizumab included death (n = 30); disease progression (n = 45); disease remission (n = 2); severe immune-related adverse events (n = 9) including four with pneumonitis, two with hepatitis, two with skin reactions, and one with flareup of stable autoimmune disease; deteriorated performance status (n = 2); second malignancies (n = 2; one hepatocellular carcinoma and one liposarcoma); and one case switching to capmatinib after receipt of additional molecular results. Seven patients stopped pembrolizumab due to financial reasons.

 

The time-dependent ROC curves, sensitivities, and specificities of various prognostic indices are shown in Figure 2. The ALI outperformed the other prognostic indices, with the highest AUCs of 0.797, 0.813 and 0.815 for 6-month, 1-year and 2-year OS predictions, respectively. An ALI ≤17.4 had sensitivities and specificities in 6-month (84.2% and 60.9%), 1-year (85.7% and 71.9%), and 2-year (73.1% and 95.0%) OS predictions. Cut-off values of ALI (≤17.4), dNLR (≥2.6), NLR (≥4.8), and PLR (≥232) were comparable to those identified in the literature (Table 2).[13] [14] [15] [16] [17] [18] [19] [20]

 

Figure 2. Time-dependent receiver operating characteristic curves on overall survival (OS) predictions according to prognostic indices. (a) Advanced lung cancer inflammation index. (b) Derived neutrophil-to-lymphocyte ratio. (c) Neutrophil-to-lymphocyte ratio. (d) Platelet-tolymphocyte ratio.

 

Table 2. Sensitivities and specificities for overall survival predictions of the prognostic indices at respective cut-off values.

 

The variables included in the univariate and multivariable regression analyses are shown in Table 3. Low body weight (body mass index [BMI] <18.5 kg/m²), ECOG performance status score of 2, baseline liver metastasis, and all prognostic indices at their respective cut-offs were identified as independent poor prognostic factors in multivariable regression analyses. Adjusted hazard ratios were 4.43 for ALI ≤17.4 (p < 0.001), 2.21 for dNLR ≥2.6 (p = 0.006), 2.08 for NLR ≥4.8 (p = 0.010), and 1.75 for PLR ≥232 (p = 0.037). OS curves showed clear separation with each prognostic index in which ALI performed the best (Figure 3).

 

Table 3. Univariate and multivariable analyses for prognostic factors on overall survival prediction.

 

Figure 3. Kaplan–Meier estimates of overall survival (OS) according to prognostic indices at respective cut-offs of (a) advanced lung cancer inflammation index (ALI), (b) derived neutrophil-to-lymphocyte ratio (dNLR), (c) neutrophil-to-lymphocyte ratio (NLR), and (d) platelet-to-lymphocyte ratio (PLR). ALI ≤17.4 has the greatest ability to stratify the prognosis. Tick marks indicate data censoring at the last time the patient was known to be alive.

 

The choice of the optimal first-line treatment for metastatic NSCLC with PD-L1 expression of ≥50% remains challenging. The level of PD-L1 expression appears to be indicative and was suggested as part of routine testing in non–oncogene-addicted metastatic NSCLC by international guidelines such as those from the European Society for Medical Oncology[21] and the National Comprehensive Cancer Network.[22] However, high PD-L1 expression alone was known to have limited predictive value of IO benefit, given a significant proportion of patients remained resistant to pembrolizumab treatment. Emerging biomarkers such as tumour mutation burden and tumour microenvironment appeared promising, but the best clinical criteria and biomarkers for patient selection and outcome prediction remain unsettled.[23] To our knowledge, this is the largest review in our locality of pembrolizumab with or without chemotherapy for metastatic NSCLC with high PD-L1 expression along with the comparison of four prognostic indices for OS prediction.

 

OS is a less ambiguous endpoint in the real-world setting where the timing imaging is variable, and response evaluation is not standardised. Yet, the exact treatment effectiveness could still be difficult to interpret due to susceptibility to post-baseline events. Conversely, treatment-based endpoints such as ToT are considered an approximation of progression-free survival. It can serve as a pragmatic endpoint for the evaluation of treatment benefit, especially in the context of IO or targeted therapy, by taking into consideration some other clinically relevant reasons for treatment discontinuation (such as worsened performance status, patient preference, and immune-related toxicities) or treatment continuation (such as for pseudo-progression during the early treatment phase, treatment beyond progression in the context of continued clinical benefits, and limited further treatment choices).[24] However, its interpretation should be cautioned as reasons including financial constraints and second malignancy were also included. Response rate analysis was not performed due to the lack of protocol and inconsistency in response evaluation and follow-up imaging.

 

We observed that a higher proportion of individuals in the IO-combination group were younger, required less radiotherapy, had more non-squamous histology, fewer underweight, and fewer brain metastases compared to the IO-alone group. However, only the former two factors were statistically significant (Table 1). These characteristics signified better prognosis at baseline and were compatible with the usual clinical selection criteria for combination treatment in which these patients were deemed fit enough to endure intensive combination treatment. However, despite treatment escalation, both the IO-alone and IO-combination groups showed similar median OS. Slightly longer ToT was noted in the IO-combination group, although the difference was not statistically significant. This illustrates the unmet need for optimal selection to receive IO-alone versus IO-combination treatment in metastatic NSCLC with high PD-L1 expression. There is no reason to expose patients to chemotherapy early if there is no potential benefit. There is no trial result directly comparing pembrolizumab alone and combination treatment in first-line setting for patients with metastatic NSCLC with high PD-L1 expression.

 

Unsurprisingly, our real-world OS and ToT were shorter than those in the trial setting, which is consistent with prior real-world study.[25] Inclusion of patients with ECOG performance status score of ≥2 for treatment in real-world practice (19.6% in the IO-alone group and 19.0% in the IO-combination group) [Table 1] compared with none in the three landmark trials (KN-024,[5] KN-189,[8] and KN-407[10]) is one of the explanations. Poor performance status is a well-established factor associated with poor survival and increased toxicities in the chemotherapy era, which also holds true for IO, as demonstrated in prior retrospective studies.[26] [27] [28]

 

Prior studies have shown that various prognostic indices were effective at their respective cut-off values.[13] [14] [15] [16] [17] [18] [19] [20] Still, there were no local data to compare and demonstrate their applicability in our locality. In our study, we performed ROC analyses to determine the optimal cut-off values based on local survival data. Given the prognostic nature of these indices, preference was given to sensitivity rather than specificity. We emphasised their ability to identify patients with poor prognoses and at higher risk of deterioration, in which clinicians could intensify treatment at an early stage for a suitable population or maintain closer monitoring of the patient’s condition during treatment.

 

All four prognostic indices, namely ALI, dNLR, NLR, and PLR in descending order of their prognostic abilities, effectively stratified patients’ prognoses at respective cut-off values. The identified cut-off values in our study were comparable to the results from the literature.[14] [15] [16] [17] [18] [19] [20] This served as a validation of these prognostic indices and acted as the indicative reference for local practice. ALI ≤17.4 had the largest AUCs, highest sensitivities, and specificities (Table 2). This is likely explained by the composition of ALI, which included the host’s general well-being as reflected by nutritional status (BMI and serum albumin level) and an inflammation-based marker alone (dNLR) that might have a synergistic relationship with IO.[19] Furthermore, our multivariable regression analyses also demonstrated that BMI <18.5 kg/m², ECOG performance status score of 2, and baseline liver metastases were statistically significant poor prognostic factors (Table 3). This is consistent with prior studies showing liver metastasis being an indicator of poor prognosis and poor response to IO, which may favour the use of combination treatment.[7] [29]

 

Given its prognostic value and easily collected variables in daily practice, ALI is an appealing marker to be incorporated into the treatment algorithm. However, the interpretation and application of ALI in a clinical context are important. Poor prognosis, despite pembrolizumab treatment, may result from poor baseline condition, limited treatment response, or both. For patients with high disease load and turnover rate, intensifying the treatment with an early chemotherapy combination allows a rapid cytotoxic effect on fast-growing cells and provides a synergistic immune-modulating effect, hence improving the efficacy of pembrolizumab.[30] [31] However, patients who are deemed to be poor candidates for any form of anti-cancer treatment, early introduction of best supportive care instead of proceeding with pembrolizumab and/or chemotherapy would be more appropriate. ALI can predict a patient’s prognosis with pembrolizumab, but the exercise of clinical judgement is needed to formulate a patient’s management plan.

 

Several limitations were identified in our study. First, our study was retrospective and therefore prone to biases. Other commonly utilised prognostic indices, including inflammatory markers (such as the C-reactive protein/erythrocyte sedimentation rate) and lactate dehydrogenase (LDH)–related markers (such as LDH level and the lung immune prognostic index), were unavailable in our study because they are not routine baseline blood tests. It is believed that the C-reactive protein/erythrocyte sedimentation rate may indicate host inflammation status, and LDH level may indicate tumour metabolism.[32]

 

Secondly, our sample size was relatively small, and patients receiving IO-combination represent only a minority within our cohort. This was explained by the different approval timelines of pembrolizumab alone in 2016[33] and combination treatment in 2017 (for non–squamous cell carcinoma)[34] and 2018 (for squamous cell carcinoma)[35] by the United States Food and Drug Administration and the time needed for the local working group in Hong Kong to adopt their regular uses in public health sectors. It was difficult to objectively compare disease burden and tumour turnover among these two groups, although the proportion of patients with baseline brain, pleural, and liver metastases appeared comparable among the two cohorts.

 

 

Thirdly, there is limited generalisability of our result to NSCLC with driver mutations other than common ones of EGFR and ALK due to small patient numbers. Variable responses to IO had been observed in different studies, especially concerning those harbouring rarer driver mutations.[36] [37] However, the exact interplay between biology and immunology is yet to be determined.

 

Lastly, other emerging biomarkers related to tumour genome (e.g., tumour mutation burden, microsatellite instability, and DNA repair gene) and immune microenvironment were not available for current analysis. However, these newer biomarkers are much more technology-demanding and research-based, and are not readily available in day-to-day practice for clinicians to quickly identify patients who are expected to have poor prognoses and limited benefit from pembrolizumab.

 

Despite these limitations, our results validated and suggested that patients with ALI ≤17.4 and the presence of baseline liver metastasis are at higher risk of shorter OS and earlier disease progression. Evaluation of ALI and detection of liver metastasis may guide clinicians’ decisions in formulating management plans with patients.

 

Judicious evaluation of the reasons for low ALI attributed to poor baseline general condition where little gain from any anti-cancer treatment is expected, or aggressive disease with high tumour burden and rapid tumour turnover where treatment intensification with combination treatment may be considered as essential.