This noninterventional, retrospective cohort study used data from the international TriNetX Global Collaborative research data network. The primary goal of this research is to evaluate the effect of ICIs treatment for metastatic lung cancer (MLC) on the likelihood of developing a SPC outside the lungs in a real-world scenario. We studied patients from TriNetX with MLC aged 18 years or older. TriNetX is a global collaborative research network that provides real-time access to electronic medical records (diagnoses, procedures, medications, laboratory values, and genomic information) from more than 135 million patients from 114 health care organizations (HCOs), primarily from North America and Western Europe [12]. All data collection, processing, and transmission is done in compliance with all Data Protection laws applicable to the contributing HCOs, including the EU Data Protection Law Regulation 2016/679, the General Data Protection Regulation (GDPR) on the protection of natural persons with regard to the processing of personal data, and the Health Insurance Portability and Accountability Act (HIPAA), the US federal law that protects the privacy and security of health care data. The Global Collaborative Network is a distributed network, and analytics are performed at the HCO, with only aggregate results being returned to the platform.  Individual personal data does not leave the HCO. TriNetX is ISO 27001:2013 certified and maintains a robust IT security program that protects both personal data and health care data. Data collection and quality control methods have been described [11]. As a federated network, research studies using TriNetX are compliant with requests from ethics committees of all contributing countries [13]. To comply with legal frameworks and ethical guidelines guarding against data reidentification, the identity of participating HCOs and their individual contribution to each dataset are not disclosed.

Selected patients were men or women, aged 18 or older, with a diagnosis of MLC between February 1, 2004, and February 1, 2024, to capture the full evolution of ICI use in clinical practice. This extended timeframe allows for robust comparisons across different treatment eras and reflects real-world changes in therapeutic strategies over time. SPCs were identified using International Classification of Diseases codes for new primary malignancies at anatomical sites other than the lung, with a minimum latency of 6 months after the initial lung cancer diagnosis, to distinguish them from recurrences or metastases [14]. Due to data limitations, most SPCs identified were metachronous, and synchronous cases could not be reliably assessed. Patients were excluded when (1) a patient died or had a non–lung cancer diagnosis within the first 6 months after the metastasis diagnosis, (2) patients never had a visit to the HCO before the metastasis diagnosis, (3) patients never had a follow-up visit after 6 months of the metastasis diagnosis, (4) patients had lung cancer as the SPC given the difficulty in clinical practice in differentiating a second primary lung cancer from a new metastatic localization of the treated lung cancer, and (5) patients did not have any data of treatment after the metastasis. The initial group of participants was then divided into 2 cohorts. The “ICI” group included patients treated with ICI after the metastasis diagnosis and the “non-ICI” group included patients never treated with ICI. For this study, data were collected until February 25, 2024 (ie, database extraction date). The details on the codes and criteria used to define the cohorts can be found in the Multimedia Appendices 1 and 2.

All the statistical analysis was done using the TriNetX platform’s built-in analysis features. ICI and non-ICI cohorts were propensity score matched (PSM) before the comparison on age at the moment of the metastasis diagnosis, gender, race (when available), de novo MLC, smoking, and patient dependency status [15,16]. The index event of both cohorts was the metastasis diagnosis, and the follow-up time window was 10 years.

Baseline characteristics before and after PSM are presented with the number and percentage for categorical variables and mean and SD when variables are numeric. Categorical variables are compared between cohorts with the Fisher exact test P value and the standardized mean differences. Numeric variables are compared with the t test P value and standard difference.

Kaplan-Meier analysis was performed on the matched cohorts to calculate the time between the MLC diagnosis and the SPC. Cox proportional hazard ratios (HRs) and 95% CIs are presented.

This study was conducted using data from the TriNetX Global Collaborative Network, which provides access to aggregated and deidentified electronic health records. As such, it was exempt from institutional ethics review [17]. All data handling complied with relevant data protection regulations, including the GDPR in the European Union and the HIPAA in the United States. Data collection and quality control methods have been described [11], and research studies using the TriNetX federated network are compliant with ethics committee requirements of all contributing countries [13]. To safeguard participant privacy, only aggregated counts and deidentified information are used, and no protected health information or personal data is made available to platform users. The process of deidentification has been formally attested by a qualified expert in accordance with Section §164.514(b)(1) of the HIPAA Privacy Rule (last updated in December 2020). As such, additional institutional review board approval or informed consent is not required for studies performed within TriNetX. This formal determination by a qualified expert, refreshed in December 2020, supersedes the need for TriNetX’s previous waiver from the Western Institutional Review Board.

Our study, encompassing 2844 patients with metastatic primary lung cancer from the TriNetX network, indicates a significant impact of ICIs on the incidence of secondary primary cancers and overall survival, showcasing the power of large-scale data in generating valuable insights. This study highlights the critical role of big data and digital health in advancing cancer treatment and research. The TriNetX network, encompassing numerous international centers, exemplifies how big data can be leveraged to conduct extensive and robust studies. The analysis, balanced post-PSM with 685 patients in each cohort, revealed that ICI treatment was associated with a reduced risk of developing SPCs. Specifically, the incidence of SPCs excluding lung cancer was 1.5% in the ICI group compared to 3.6% in the non-ICI group, translating into a risk difference of −0.028 and a favorable HR of 0.49. In addition, our findings demonstrated a reduction in the risk of SPC or death in patients treated with ICIs, with a significant HR of 0.78. This suggests not only the efficacy of ICIs in treating primary tumors but also their potential role in preventing secondary malignancies. The median time to MLC or death was notably higher in the ICI-treated group, further emphasizing the survival benefits of ICIs. The interpretation of the composite endpoint “SPC or death” is limited by the potential confounding effect of overall survival, as death may censor the observation of SPCs. Moreover, patients treated with ICIs may live longer, thereby increasing their window of risk for developing SPCs, which could paradoxically elevate SPC incidence in this group despite a protective effect of immunotherapy.

Our findings align with prior studies suggesting a decreased incidence of patients with SPCs treated with ICIs, such as those by Heudel et al [4-7]. However, our work extends these observations by analyzing a larger, more recent, and real-world cohort of patients with MLC (N=2844) with PSM and an updated follow-up through 2024. Unlike previous studies that included various primary tumor types, we focused specifically on metastatic or locally advanced lung cancer, enhancing the interpretability within a defined clinical context. Moreover, our study leverages the robust infrastructure of the TriNetX platform [18], enabling federated access to real-world data across multiple international HCOs and offering methodological transparency and reproducibility. Recent immunogenetic research supports a biological rationale for ICI-mediated protection: a large UK Biobank study identified specific HLA alleles associated with either increased or decreased SPC risk, suggesting that the adaptive immune system, and thus immune modulation via ICIs, may play a key role in SPC prevention [18]. In addition, population-level evidence from Sweden highlighted strong bi-directional associations between myeloid malignancies and various SPCs, likely involving immune dysfunction or treatment-induced changes [19]. Taken together, these studies reinforce the hypothesis that immune surveillance mechanisms are central to SPC susceptibility. Our results, combined with this growing body of evidence, support the concept that ICIs may offer not only survival benefits but also a protective effect against SPCs, especially in patients with immune-related predispositions.

Of course, our study of many limitations began with its retrospective nature. Even if the TriNetX network is extremely important, with numerous international centers, it clearly appears that the scope of the available data remains limited since we do not have, for example, the histological subtypes, or the smoking status, and even less the molecular characteristics of cancers. These missing data constitute a major limitation to the conclusions of our study and to the question of the accessibility and the sharing of these data, which are necessarily present in the electronic medical records [20]. Learned societies have recently established recommendations on real-world data studies, the quality of the conclusions of which is directly linked to the quality of the data obtained [21].

However, the effectiveness of big data and digital health hinges on the quality and comprehensiveness of the data. We must not forget the importance of real-world data that are complementary to clinical trials [22], particularly on subjects such as the appearance of a SPC for which these patients are often excluded from traditional clinical research studies. It therefore appears fundamental to continue collaboration within, for example, the TriNetX network because the greater the volume and completeness of the data, the greater the confidence in the conclusions of these retrospective studies [12]. For this, it seems necessary for each hospital to define its digital data strategy, which must include a data collection strategy (by aligning them from the start with international benchmarks), for the qualification of this data, for their analysis with quality control, and a sharing strategy [23]. This entire strategy must, of course, respect the regulations in force and integrate a priori the question of interoperability to facilitate data sharing and enhance research outcomes [24,25].

To maximize the potential of big data and digital health, continued collaboration within networks such as TriNetX is essential. The greater the volume and completeness of the data, the higher the confidence in the study conclusions. Recommendations from learned societies on real-world data studies emphasize the importance of data quality, suggesting that the robustness of study conclusions is directly linked to the quality of the data obtained.

In conclusion, our study showed that patients treated with ICIs for a first lung cancer had a reduced risk of an SPC and underscores the significant benefits of using big data and digital health in cancer research. The ability to conduct real-life studies on an international network such as TriNetX allows for the inclusion of diverse patient populations and the generation of real-world evidence that complements clinical trials. Despite current data limitations, continued network collaborations and strategic data management can enhance the relevance and impact of real-world data studies, ultimately improving patient outcomes and advancing medical knowledge.