This study used data from the Merative MarketScan Research Database (1 January 2014–30 September 2021). This large private health insurance database contains paid inpatient (IP), outpatient (OP), and pharmaceutical claims generated each year by nearly 51 million employees and their dependents in the USA with employer-sponsored insurance and enrolled in various fee-for-service and managed care plans. Also included are claims for Medicare-eligible retirees and their dependents with employer-sponsored supplemental insurance. All census regions in the USA are represented, although there is a slightly higher representation of beneficiaries from the South and North Central (Midwest) regions.

This was a retrospective longitudinal cohort study. The index date was defined as the initiation date of first-line systemic treatment for aRCC (i.e., nivolumab plus ipilimumab, pembrolizumab plus axitinib, sunitinib monotherapy, pazopanib monotherapy, cabozantinib monotherapy, nivolumab plus cabozantinib, and pembrolizumab plus lenvatinib) on or after the date of aRCC diagnosis (Fig. 1). For combination therapy, the initiation date of the first agent of the combination therapy was defined as the index date and the second agent had to be initiated within 28 days after the initiation of the first agent. No other RCC-related treatments were allowed within the 28-day time window. The 28-day time window was selected based on the most commonly used dosing cycles (2–4 weeks) for combination therapies [6, 12, 13], previous studies which evaluated lines of therapies in cancer [14, 15], and clinical input.

Study design. aRCC could be diagnosed before or during the 6-month baseline period. 1L first line, AE adverse event, aRCC advanced renal cell carcinoma

Patient characteristics were evaluated during the baseline period, defined as the 6 months preceding the index date. The AE identification period spanned from the index date to the earliest of the following events: (1) 30 days after the date of discontinuation for the index first-line treatment—i.e., the first date that the first-line therapy was no longer available to the patient, followed by a gap of ≥ 90 days of therapy supply; (2) initiation of second-line therapy; (3) end of continuous enrollment of the health insurance plan; or (4) end of data availability.

The study population comprised adult patients with aRCC receiving first-line treatments. Inclusion criteria were as follows: (1) At least two medical claims ≥ 30 days apart with a diagnosis for RCC (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 189.0 or ICD-10-CM code: C64). (2) At least two medical claims with a diagnosis of secondary malignant neoplasm on different days on or after the initial RCC diagnosis date (Appendix 2); the date of the first secondary malignant neoplasm diagnosis was defined as the aRCC diagnosis date. (3) Receipt of any of the aRCC systemic therapies of interest as first-line treatment on or after the aRCC diagnosis date. (4) Continuous health insurance eligibility for ≥ 6 months before and ≥ 1 month after the index date. Patients were excluded if they were enrolled in any health maintenance organization or capitated point-of-service health insurance plan due to incomplete cost data. In addition, patients were excluded if they had ≥ 2 claims for any other primary cancers or were enrolled in a clinical trial during the baseline period.

AEs of interest were selected based on a review of product labels of the first-line treatments included in this study. To be included in this study, corresponding grade 3/4 AEs had to have occurred in ≥ 5% of the study population in any of the product labels. In total, 27 AEs of interest were identified and confirmed by clinical experts. The AEs were categorized as follows: general (fatigue/asthenia, pain, weight loss/decreased appetite); endocrine (hypothyroidism, pancreatitis); gastrointestinal (diarrhea, nausea/vomiting); nervous system (syncope); renal and urinary (acute kidney injury, proteinuria); respiratory, thoracic, and mediastinal (dyspnea); skin and subcutaneous tissue (palmar–plantar erythrodysesthesia/hand–foot syndrome, rash, stomatitis); vascular (hypertension, hypotension); laboratory abnormality—chemistry (hyperglycemia, hyperkalemia, hypomagnesemia, hyponatremia, hypophosphatemia, increased alanine aminotransferase [ALT]/aspartate aminotransferase [AST]); and laboratory abnormality—hematology (anemia, leukopenia, lymphopenia/lymphocytopenia, decreased neutrophils/neutropenia, decreased platelets/thrombocytopenia). These AEs were identified in the administrative claims data using ICD-9/10-CM codes (Appendix 1).

Costs associated with AEs of interest were analyzed for cases and controls during the AE identification period. Cases were patients with each specific AE of interest in the primary or secondary position of a claim. The date of first occurrence of the AE was defined as the AE event date; the 30 days that followed constituted the AE cost assessment period. Controls were patients who did not have the specific AE during the AE identification period. For example, in the analysis of dyspnea, eligible patients with dyspnea were selected as cases, and eligible patients without dyspnea were selected as controls. A ‘shadow’ date before the end of the AE identification period was randomly assigned to control patients based on a normal distribution with the same mean and standard deviation as the time from initiation of first-line treatment to the occurrence of the first AE of interest for cases. The AE cost assessment period for controls was the 30-day period after the shadow date. Note that the same patient could be classified as a case for a given AE and as a control for another AE. For example, if a patient had dyspnea and hypertension (and no other AEs) during their AE identification period, this patient would be classified as a case for these two AEs but would also be classified as a control for other AEs (e.g., fatigue). Additionally, patients were required to have continuous eligibility for ≥ 30 days after the AE event date (for cases) or shadow date (for controls). Patients were excluded if the AE event occurred during the baseline period.

All-cause healthcare costs incurred during the AE cost assessment period were evaluated for cases and controls; these included total healthcare costs, total medical service costs (IP, emergency room [ER], OP, and other costs such as durable medical equipment, dental, and vision costs), and total pharmacy costs. Costs related to RCC treatments were excluded so that only costs associated with AEs—and not the cost difference for first-line treatments between cases and controls—were captured. Costs were inflated to 2022 USD using the medical care index of the US Bureau of Labor Statistics Consumer Price Index [16].

Baseline characteristics were described and compared between patients with vs. without any AE during the AE identification period. Continuous variables are reported as mean (standard deviation [SD]) or median (interquartile range) and were compared using the Wilcoxon rank-sum test. Categorical variables are reported as number (percentage) and were compared using the chi-squared test. Incremental healthcare costs associated with an AE were estimated as the difference in total healthcare costs incurred during the AE cost assessment period between cases and controls. Unadjusted and adjusted cost differences were estimated using a two-part modeling approach: the probability of observing a non-zero cost was first modeled by logistic regression, and a generalized linear model with a gamma distribution and log link was then used to model costs among patients with positive costs (Appendix 7).[17] For the adjusted cost difference, relevant baseline characteristics were adjusted in the modeling, including age, sex, geographic region, health plan type, year of index date, time from RCC diagnosis to index date, type of first-line treatment, time from index date to AE or shadow date, presence of metastasis at a visceral site (lung, liver, or brain), presence of bone metastasis, number of metastatic sites, and Charlson Comorbidity Index (CCI) score. The list of variables was selected based on clinical and statistical relevance. A nonparametric bootstrap procedure with 499 replications was used to estimate 95% confidence intervals and p values. Sensitivity analyses using shorter AE cost assessment periods (7 and 14 days) were performed to reduce the possibility of capturing non-target AE costs and evaluate the robustness of the results.