Lung cancer (LC) is the leading cause of cancer mortality worldwide, accounting for an estimated incidence of 2.4 million cases per year (12.4% of all cancer cases).1 This disease is classified into small cell lung cancer (SCLC) and non-small-cell lung cancer (NSCLC), with differing prognosis and treatments.2 NSCLC accounts for 80–90% of all lung tumors and is subcategorized into two histological types: squamous and non-squamous.2

For years, first-line treatment for advanced non-mutated NSCLC involved platinum-doublet chemotherapy; however, high relapse rates often required second-line treatment with docetaxel.3 Accumulating evidence linking specific genetic alterations to targeted therapeutic strategies has profoundly reshaped the treatment algorithm for NSCLC. In this context, molecular profiling (EGFR, BRAF, KRAS, HER2 gene mutations; ALK, ROS, RET rearrangements or NTRK fusions)2 has become essential, as immunotherapy has emerged as a cornerstone of treatment.

Immunotherapy is aimed at boosting the immune system activity against the tumor through the use of immune checkpoint inhibitors (ICIs). For such a purpose, a range of monoclonal antibodies targeting the programmed death-1 (PD-1) receptor or its ligand (PD-L1) has been developed. Current clinical guidelines recommend these agents as second-line treatment for patients with advanced/metastatic NSCLC following platinum-based chemotherapy.2,4,5 Cumulative evidence supporting the superiority of nivolumab, pembrolizumab and atezolizumab over docetaxel found in phase III studies3,6–11 led to their approval by the Spanish Agency for Medicines and Medical Devices.3,6–11 Consequently, immunotherapy has positioned as the preferred first-line treatment for NSCLC. Nevertheless, a proportion of patients still continue to progress to second-line therapy without having received prior immunotherapy. In this setting, any of the three agents can be considered, given the comparable level of evidence supporting their efficacy, despite the lack of direct comparative analyses.4

Immunotherapy has demonstrated lower toxicity compared to standard cytotoxic agents. However, PD-1/PD-L1 inhibitors also entail some risks. Immune checkpoint pathways play a critical role in maintaining self-tolerance. Accordingly, ICIs have the potential to alter the immune homeostasis, thereby triggering immune-related adverse events (irAEs) at any level.12 In a recent meta-analysis, the incidence of irAEs was estimated to reach 40%, with high-grade irAEs accounting for 19.7% of all irAEs.13 Common manifestations include arthralgia, endocrinopathies, dermatitis, hepatitis, colitis and pneumonitis.14

The role of irAEs as a surrogate marker of tumor response to ICI is controversial. A variety of studies and meta-analyses have revealed higher survival and response rates in patients who develop irAEs, as compared to those not experiencing any.15–20

The purpose of this study was to explore the role of irAEs as a predictive factor of response to ICI in NSCLC.

Although direct comparison of the three agents was not performed, they exhibited similar clinical benefits. No statistically significant differences were observed in PFS (7.8 months for nivolumab; 7.8 months for pembrolizumab; 5.0 months for atezolizumab) or OS (12.9 months for nivolumab, 22 months for pembrolizumab, 12.9 months for atezolizumab). The main limitations of this study include its sample size, retrospective design and single-centre nature, which limit the generalization of results to other populations and clinical settings.

The OS obtained with nivolumab is consistent with those achieved with second-line treatment in NSCLC patients in pivotal clinical trials and other real-world studies, with OS ranging from 9.7 to 11.0 months.9,21–23 However, the ORR obtained in our series (28%) was superior to that reported by other authors (19.7%),9 thereby resulting in a higher PFS as compared to previous studies (2.3–5.3 months).9,21–23 Such differences may stem from various factors, such as the retrospective nature of the study and the absence of an independent radiological review committee to assess response.

The median OS for pembrolizumab (22 months) differs substantially from that reported in the phase III study (10.4 months).8 Such a difference could be attributed to the small size of our sample (n = 11 patients) and the high percentage of patients with PD-L1  ≥  50%, which has been associated with improved OS outcomes (16.9 months).10 The PFS outcomes in our study are not consistent with those reported in previous real-world studies (16.7–24 months).24,25 Interestingly, a substantial difference was observed in ORR (45.5%) as compared to the pivotal study (18.0%).10 This inconsistency could be due to the detection of two CRs, which has a remarkable impact considering our small sample.

Regarding the atezolizumab cohort, OS rates are aligned with those obtained in the pivotal study (13.8 months).3,11 There is broad variability in OS outcomes in real practice, ranging from 6.5 months26 to not achieved. Nevertheless, the latter study involved a shorter follow-up period and number of patients.25 However, a higher level of agreement with other case-series studies was observed in terms of PFS (2–2.8 months)3,17 and ORR (19% in our cohort versus 14%).3

The incidence of AEs (93% of patients) is not consistent with that previously reported for nivolumab, pembrolizumab and atezolizumab.3,6–8 However, a comparable toxicity profile was identified concerning AE type. The most common AEs included anorexia and asthenia. No other common AEs, such as nausea were detected, probably due to underdiagnosis. Interestingly, a high proportion of patients developed hyperglycemia, since it is easily identified on laboratory tests. Higher consistency with pivotal trials was observed regarding severe AEs (24% in our cohort vs. 10.8% nivolumab; 13.0% pembrolizumab and 15.0% atezolizumab) and toxicity-related immunotherapy discontinuance (14% in our cohort vs. 6.5% for nivolumab, 4% pembrolizumab and 8% for atezolizumab).3,6–8

In our series, the most common irAEs included hepatotoxicity, arthromyalgia and colitis, whereas the most frequent irAEs in pivotal trials included cutaneous and gastrointestinal toxicity for nivolumab6,7; thyroid gland dysfunction and pneumonitis for pembrolizumab8; and pneumonitis, hepatitis and colitis for atezolizumab.3

The most relevant finding in this study was the significant association observed between irAEs and clinical benefit. The patients who developed two or more irAEs exhibited improved OS (28.4 vs 11.9 months) and PFS (24.5 vs. 5.2 months), as compared to the patients who developed one or less AEs. The pathophysiological mechanisms that explain this association are not fully understood; however, the onset of irAEs could reflect a robust activation of the immune system against the tumor. This association has been previously described in the literature, with a higher benefit observed in OS, PFS, ORR and CET in patients who developed irAEs while receiving ICI.16,17 Similar results were obtained in previous studies in smaller cohorts.18–20,27,28 Nevertheless, given the observational design of the study, the results allow for the identification of statistical associations but do not establish causal relationships. Furthermore, the presence of time-dependent bias cannot be excluded. Specifically, survival time needed to be long enough for patients to develop two or more irAEs; therefore, survival for this cohort may have been overestimated. Other limitations include the limited sample and number of events, potentially influencing the accuracy of estimations, and the potential influence of confounding factors (age, comorbidities, performance status, previous treatments, molecular profile, etc.).

The comparability of results is hindered by the lack of a standard definition for irAEs. The vast majority of studies assessed immune-related toxicity as a dichotomous variable (presence versus absence).17–19,27,28 Only two studies established the cut-off point of ≥2 irAEs, thereby using a similar approach to ours.16,20 This patient subgroup exhibited a durable tumor response.16 Stratification by accumulated number of irAEs facilitates a more accurate evaluation of the association between immune-related toxicity and the observed clinical benefit.

Interestingly, not all irAEs seem to play the same role as predictive of effectiveness. Some AEs, such as exanthema and temperature, were found to be associated with improved PFS, whereas others, such as diarrhea and hepatitis, have shown no association.29 Conversely, an association has been reported between pneumonitis and increased mortality. However, this finding could be influenced by the severe impact of pneumonitis-induced hypoxia in NSCLC patients, in whom tissue oxygenation is already impaired.30 The small sample hindered any comparative analysis of the different types of AEs.

In conclusion, real-world data suggest a superior clinical benefit of PD-L1 inhibitors in patients who develop two or more irAEs. The median OS doubles and PFS is five-fold higher in this subgroup of patients, thereby supporting a potential role of irAEs as a potential predictive factor of response. IrAE control has progressively improved as experience with the management of these toxicities has been gained. At present, irAEs lead to treatment discontinuance in only a small proportion of patients. In this setting, the potential of irAEs as markers of treatment efficacy suggests that active monitoring of these AEs would facilitate early irAE detection. Such an approach would help establish supportive care, thereby enabling the maintenance of dose intensity without compromising patient safety. Nonetheless, these findings should be interpreted with caution, in view of the limitations described above. Larger prospective studies are needed to confirm results and perform a more robust subgroup analysis.

Lung cancer is the leading cause of cancer mortality worldwide. As clinical outcomes have been shown to improve with immunotherapy, this approach has progressively moved into earlier lines of treatment, thereby displacing chemotherapy. However, a subgroup of patients continues to receive second-line therapy without previous immunotherapy.

The contribution of this study lies in the provision of real-world evidence regarding the significant association between immune-related toxicity and clinical benefit in patients with NSCLC treated with PD-1/PD-L1 inhibitors. The results obtained suggest the role of immune-related toxicity as a surrogate marker of effectiveness. These findings open an avenue for future research involving larger cohorts aimed at developing protocols that maintain dose intensity despite the occurrence of toxicities.

This study did not receive any funding.

All authors accept full responsibility as defined by the International Committee of Medical Journal Editors (available at http://www.icmje.org/). In the event of publication, the authors hereby exclusively transfer the rights of reproduction, distribution, translation, and public communication (by any means or through any sound, audiovisual, or electronic medium) of our work to Farmacia Hospitalaria and the SEFH. To this end, a copyright transfer agreement will be signed at the time of submission through the online manuscript management system.

The authors used ChatGPT for style and format editing. After using this AI tool, the authors revised and edited, where appropriate, the contents of the manuscript, and assume full responsibility for its contents.

Maria Susana Fortes-Gonzalez: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Silvia Vazquez-Blanco: Writing – review & editing, Writing – original draft, Formal analysis, Data curation. Leticia Herrero-Poch: Writing – review & editing, Writing – original draft, Formal analysis, Data curation. Lucia Santome-Couto: Writing – review & editing, Writing – original draft, Supervision, Formal analysis, Conceptualization. Ana Cristina Cercos-Lleti: Writing – review & editing, Writing – original draft, Supervision, Formal analysis, Conceptualization.