Randomized clinical trials (RCTs) are undoubtedly the best way to highlight the benefits of a novel technology or intervention in healthcare under ideal circumstances and, therefore, avoid confounding factors that could either enhance or cover the effects arising from its implementation. The results from an RCT are very important considering that they can dramatically change the standard practices and could improve therapeutic offers for patients (1).

Not less important is verifying whether the RCT results can improve outcomes in real-world practice, outside the ideal context of a clinical trial, in which patients are selected with rigorous inclusion and exclusion criteria and procedures and timing are strictly defined (2–6).

A.L. Cochrane defines effectiveness as the measure of whether an intervention provides more benefits than drawbacks when applied in regular healthcare settings (7). Therefore, the most reliable method to assess the effectiveness of novel technology is to conduct a randomized pragmatic study in regular clinical practice. This study would compare the outcomes of two treatments used to address the same medical condition (8–11).

However, it is not ethical to deny a group of patients a treatment that has been proven effective through a randomized controlled trial. Additionally, scientific advancements and news spread rapidly across the globe, making information about the latest approved treatments easily accessible to patients and their caregivers.

The only way to assess treatment effectiveness is through real-world studies even though they are not designed to evaluate comparative effectiveness and suffer from bias and confounding factors (12). Real-world studies are nevertheless important for the following:

Immunotherapy has become the most impactful technology in the field of oncology over the past decade (15). It has redefined treatment guidelines for most forms of cancer, especially for drugs that are effective on the programmed cell death protein 1/programmed cell death ligand 1 (PD-1/PD-L1) signaling pathway, since nivolumab has been approved by the National Health System for the treatment of metastatic melanoma and second-line therapy of non-small cell lung cancer (NSCLC) (16, 17). Soon after, other drugs like pembrolizumab, atezolizumab, avelumab, durvalumab, cemiplimab, and dostarlimab became available for many other indications beyond the first two registered (18). These drugs have made it possible to treat cancers like NSCLC that a decade ago had a poor prognosis but now have a better life expectancy (19). Pembrolizumab was the first drug active on the PD-1/PD-L1 axis available for the first-line treatment of metastatic NSCLC. Firstly, the availability was limited to the treatment as monotherapy for patients with a tumor proportional score (TPS) of PD-L1 ≥ 50% without the driver mutations epidermal growth factor receptor (EGFR) and anaplastic lymphoma kinase (ALK) (20). Subsequently, pembrolizumab in combination with pemetrexed and platinum became available and reimbursed for patients with nonsquamous NSCLC with TPS PD-L1 <50%. This could be possible thanks to the outcome of the Keynote-189 trial, in which pemetrexed and platinum combined with pembrolizumab achieved better results than that combined with placebo in OS [median: 19.4 months vs. 11.3 months; hazard ratio (HR): 0.60; 95% confidence interval (CI): 0.50–0.72] and PFS (median 9.0 months vs. 4.9 months; HR: 0.50; 95% CI: 0.42–0.60) both in the intention-to-treat population and in subgroups stratified for TPS of PD-L1 (21).

The Italian Medicinal Agency (AIFA) reimburses patients with central nervous system metastasis, whereas they would have been exclusion criteria for the Keynote-189 study unless they were previously treated for brain metastases, were clinically stable for at least 2 weeks, showed no evidence of new or enlarging brain metastases, and had been off steroids 3 days before dosing with pembrolizumab. In standard clinical practice, patients were treated only if they were not symptomatic for central nervous system metastasis, although with less caution regarding the stability period post-radiation treatment and the steroid washout period.

Moreover, patients with Eastern Cooperative Group Performance Status (ECOG PS) 2, who would be excluded from Keynote-189, are eligible for National Health Service payments.

On the other hand, although two meta-analyses proved that chemo-immunotherapy was more effective than immunotherapy alone (22, 23), only patients with PD-L1 < 50% met the reimbursement criteria for the combined pembrolizumab, pemetrexed, and platinum treatment. In this subgroup, Keynote-189 demonstrated the superiority of immunotherapy as compared to placebo when combined with pemetrexed and platinum. The results were based on a post-hoc analysis with a sample not calibrated to reject the null hypothesis with a predefined alpha error. Moreover, as always, the exclusion criteria for patients with a life expectancy of less than 3 months present in the clinical trial protocol are not applicable in real-world practice. Key features of patients treated in real-life settings and the registrative clinical trials are summarized in Supplementary Table 1 .

To evaluate the real benefit of pembrolizumab in a real-world context, considering the Italian Medical Agency’s reimbursability criteria, we designed an observational retrospective cohort study (PEMBROREAL) in 16 Italian oncology institutions to assess the effectiveness and safety of pembrolizumab in combination with pemetrexed and platinum used for first-line therapy of NSCLC. The study aims to describe the benefit not only for the general population but also for specific subgroups of interest.

To our knowledge, PEMBROREAL is the first real-world study reported in literature that focused on the use of pembrolizumab as a first-line treatment of NSCLC combined with pemetrexed and platinum in patients without driver mutation of EGFR and ALK and PD-L1 expression < 50%. The first consideration coming from the observation of the distribution of demographic and prognostic characteristics of patients included in the PEMBROREAL study is that the population normally treated in current clinical practice had a distribution of these characteristics significantly different from that enrolled in the clinical trial Keynote-189 (21, 24). As discussed in the Introduction in the real-world setting, we included patients with ECOG PS 2 usually excluded from clinical trials, patients with brain metastasis, and independently of their life expectancy. Another important difference regarding the population enrolled was the absence of patients with PD-L1 greater than or equal to 50% because of the limitation imposed by the Italian Medicinal Agency about reimbursability of the combination for these patients whom pembrolizumab was available as monotherapy. Moreover, we can observe some other differences, for example, relative to sex balance between Keynote-189 and PEMBROREAL. As can be seen in Table 1 , 67.0% of the PEMBROREAL study participants were men and 33.0% were women. If compared with Keynote-189, the female sex usually has a more favorable prognosis (25) and is more represented in the pivotal clinical trial (38%) than in the real-life setting. Another relevant difference is the rate of patients with PD-L1 TPS <1% (44.1%) treated in the real-world setting compared to patients enrolled in Keynote-189, which included only 31% of patients with PD-L1 < 1% (21, 24). Considering only patients with PD-L1 <50% treated with pembrolizumab combined with pemetrexed and platinum, this percentage rose to 49.8%.

The percentage of patients with ECOG PS 0 was lower in the PEMBROREAL study (39.5%) compared to patients treated with pembrolizumab in the Keynote-189 study (45,5%). These two variables, PD-L1 expression and ECOG PS, are particularly relevant if we consider that PD-L1 end ECOG PS emerged as important prognostic factors from the univariate analysis conducted by our study team and from brief literature research (26–28). Between the pull of variables analyzed, smoker status seems to be the most relevant difference between patients treated in current clinical practice and in clinical trials. Never-smoker patients recruited in the Keynote-189 trial were only 11.2% versus 23.4% in our real-world study, but we have to consider that smoker status was unknown for 138 (49.5%) of the 279 patients included in the study. Nevertheless, this variable is very interesting considering the known differences in terms of response to immunotherapy documented in scientific literature (29–33).

The median RwPFS was 8.0 months, and this is consistent with findings from Keynote-189 in which PFS was 9.0 months in patients treated with pembrolizumab administered with pemetrexed and platinum. However, as previously stated, the results reached in the pivotal clinical trial were referred to as the intention-to-treat population, which also included patients with PD-L1 ≥ 50%, excluded from the PEMBROREAL study.

If we compare the RwPFS considering PD-L1 expression between the real-world setting and the clinical trial setting, PFS in the subgroup of patients with PD-L1 < 1% and with PD-L1 ≥ 1% and < 50% was very similar (5.8 months versus 6.2 and 9.4 months versus 9.9 months, respectively).

Despite the follow-up of 19.7 months, we did not reach the RwOS for patients included in our study. However, the survival rate was consistent with the findings from the Keynote-189 trial. A 12-month survival rate in the intention-to-treat population of Keynote-189 was 69.8%, not so different if compared with the survival rate obtained in PEMBROREAL (66.1%). A greater difference could be observed if we consider the 24-month survival rate from Keynote-189 (45.7%) compared with PEMBROREAL (52.5%). This result seems to be better for patients treated in current clinical practice. Furthermore, if we consider the survival rate separately in the subgroup of patients with PD-L1 <1% and those with PD-L1 ≥ 1 and <50%, survival rate was very close between the pivotal clinical trial and the real-world study. Twelve-month and 24-month survival rates were 71.1% and 63.4% in Keynote-189 versus 71.8% and 59.4% in the PEMBROREAL study for patients with PD-L1 ≥ 1 or <50%, respectively, while for patients with PD-L1 < 1%, survival rates were 63.4% and 39.3% versus 59.0% and 43.2%.

From the analysis conducted on subgroups of interest, the PD-L1 expression and ECOG PS emerged as important prognostic factors influencing significant outcomes such as PFS and OS. Particularly remarkable were the PFS and OS results of patients with ECOG PS 2 (4.6 months) excluded from Keynote-189. This result is comparable with the PFS obtained for patients with ECOG PS 0–1 enrolled in the control arm of Keynote-189 (PFS: 4.9 months) and in the registrative study of pemetrexed combined with platinum as first-line treatment of NSCLC (PFS: 4.8 months) (21, 34) that did not include patients with ECOG PS 2. OS in the same subgroup as well was comparable to OS results from Keynote-189 in patients with better ECOG PS treated in the control arm (12.0 months vs. 10.8 months) and in patients treated in the registrative study of pemetrexed combined with platinum in the first-line treatment of NSCLC (OS 10.3 months). This is relevant to underscore because this result suggests that adding pembrolizumab to standard platinum-based chemotherapy made it possible to achieve the same results of chemotherapy alone for patients with ECOG PS 2 that generally were the subgroup with the worst prognosis normally excluded from recruitment in clinical trials. However, the limit of these considerations about this subgroup of interest included in the PEMBROREAL study is the underrepresentation of patients with ECOG PS 2 (n = 16, 5.9%); thus, this result should be interpreted with caution. Nevertheless, the value of poor ECOG PS as a prognostic factor for patients treated with immunotherapy is well explored in literature (35, 36), and our finding is consistent with what was reported about this subpopulation. The worst prognosis for these patients could be related to disease burden and comorbidities. Moreover, patients with poor ECOG PS are more likely to experience illness and mortality caused by the toxicity related to the immunotherapeutic regimen with three drugs combined and are less likely to receive further line of therapy aimed to prolong survival in case of treatment failure.

Response to treatment was very interesting and consistent with results from clinical trials. However, these results could be biased from the method of progression assessment as explained in Methods and Materials. Nevertheless, if we consider the control rate from Keynote-189 (84.6%), it was not so different from the value of the DCR estimated in PEMBROREAL (79.2%).

The results shown above seem to be very close to that obtained in Keynote-189. This finding could be unexpected if we consider that populations treated in the clinical trial cohorts are typically better selected by stringent inclusion and exclusion criteria and are generally healthier compared with patients treated in clinical practice (37). However, several factors can contribute to the overestimation of clinical outcomes in the real-world setting. First of all, the institutions selected to participate in our study are strongly oriented to cancer treatment. As stated previously, 13 out of the 16 involved institutions had a recognized role in cancer research. Additionally, PFS is generally overestimated in real-world studies. This issue may have been exacerbated by the COVID-19 pandemic contest, which could have resulted in fewer hospital visits (38).

Moreover, progression had to be determined radiologically as per RECIST criteria in Keynote-189 and was subject to blinded independent central review. On the other hand, patients in PEMBROREAL could have progression determined based on either radiological or clinical evidence. Future analyses to investigate the impact of this limitation on PFS would be of interest. OS is also generally overestimated in real-world studies. In terms of the effect of patients lost to follow-up because the reason for the loss “may” be related to the underlying patient health status. However, to avoid this confounding factor, we decided to cap the survival of patients lost at follow-up to the last contact documented in clinical records.

Regarding safety, findings from the PEMBROREAL study reveal that pembrolizumab associated with chemotherapy was well tolerated. If we compared AE incidence in patients enrolled in the experimental arm of Keynote-189 (99.8%) to the incidence reported in the real-world setting (44.4%), we could see that outside of the clinical trials in which all AEs are reported carefully in clinical records and a case report form, the incidences of AEs that physicians assess as important to report as meaningful are fewer than half. Gastrointestinal and blood and lymphatic system disorders are the most reported AEs, which is in accordance with findings from clinical trials but probably were related to the well-known effects of chemotherapy rather than pembrolizumab (39–41). Grade 3 and 4 AEs were rare in patients treated in current clinical practice (27 events). Most of the AEs that occurred did not require specific treatment of the event, and the others were easily managed by treating physicians pharmacologically or with dose reduction or temporary suspension, after which the patient was considered fit to restart treatment. Most concerns regarding treatment safety are related to adverse reactions going unreported, especially if they are mild or not clinically relevant.

The findings from the PEMBROREAL study confirm that pembrolizumab, when used in combination with pemetrexed and platinum salts, is effective and safe in treating metastatic non-squamous NSCLC in patients with PD-L1 < 50%. It is especially important to keep in mind the restrictions on reimbursement for the treatment discussed earlier. These limitations are based on a post-hoc analysis of the Keynote-189 trial, which was conducted with a sample size that was not adequate to determine the impact of the treatment on PFS and OS in this subgroup analysis. PFS and OS outcomes were broadly consistent with the Keynote-189 trial despite limitations associated with assessing disease progression in the context of a retrospective observational study that could cause an overestimation of RwPFS. The treatment was generally well tolerated with only a few cases of definitive treatment interruption due to toxicity, but perhaps also due to the general absence of a recognized standard second-line treatment for NSCLC patients without driver mutation previously treated with chemotherapy and immunotherapy (41–43). This fact could be the reason behind physicians not giving up on treatment toxicity but rather persevering until the toxicity becomes unmanageable (44).

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving humans were approved by Ethics Committee of Vast Area of Romagna. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

AC: Writing – original draft, Supervision, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. FF: Writing – review & editing, Visualization, Software, Methodology, Formal analysis, Data curation. ON: Writing – review & editing, Supervision, Methodology, Conceptualization. MCh: Funding acquisition, Writing – review & editing, Supervision, Software, Resources. MCo: Writing – review & editing, Supervision, Investigation, Conceptualization. PB: Writing – review & editing, Supervision, Conceptualization. SO: Writing – review & editing, Investigation. FE: Writing – review & editing, Supervision, Investigation, Conceptualization. VL: Writing – review & editing, Supervision, Investigation, Conceptualization. RL: Writing – review & editing, Supervision, Investigation, Conceptualization. PN: Writing – review & editing, Supervision, Investigation, Conceptualization. PM: Writing – review & editing, Supervision, Investigation, Conceptualization. FG: Writing – review & editing, Supervision, Investigation, Conceptualization. AG: Writing – review & editing, Supervision, Investigation, Conceptualization. GC: Writing – review & editing, Supervision, Investigation, Conceptualization. RP: Writing – review & editing, Supervision, Investigation, Conceptualization. PC: Writing – review & editing, Supervision, Investigation, Conceptualization. AP: Writing – review & editing, Supervision, Investigation, Conceptualization. SV: Writing – review & editing, Supervision, Investigation, Conceptualization. MV: Writing – review & editing, Investigation. ADF: Writing – review & editing, Supervision, Conceptualization. LC: Writing – review & editing, Visualization, Validation, Supervision, Project administration, Conceptualization. AD: Writing – review & editing, Visualization, Validation, Supervision, Project administration, Conceptualization. CM: Funding acquisition, Writing – review & editing, Supervision, Software, Resources, Project administration, Conceptualization.