Clinical and Genome-Wide Analysis of Serum Platinum Levels after Cisplatin-Based Chemotherapy

Abstract

Purpose: Serum platinum is measurable for years after completion of cisplatin-based chemotherapy (CBC). We report the largest investigation of serum platinum levels to date of 1,010 testicular cancer survivors (TCS) assessed 1-35 years after CBC and evaluate genetic contributions to these levels.

Experimental design: Eligible TCS given 300 or 400 (±15) mg/m² cisplatin underwent extensive audiometric testing, clinical examination, completed questionnaires, and had crude serum platinum levels measured. Associations between serum platinum and various risk factors and toxicities were assessed after fitting a biexponential model adjusted for follow-up time and cumulative cisplatin dose. A genome-wide association study (GWAS) was performed using the serum platinum residuals of the dose and time-adjusted model.

Results: Serum platinum levels exceeded the reference range for approximately 31 years, with a strong inverse relationship with creatinine clearance at follow-up (age-adjusted P = 2.13 × 10^-3). We observed a significant, positive association between residual platinum values and luteinizing hormone (age-adjusted P = 6.58 × 10^-3). Patients with high residual platinum levels experienced greater Raynaud phenomenon than those with medium or low levels (age-adjusted OR_high/low = 1.46; P = 0.04), as well as a higher likelihood of developing tinnitus (age-adjusted OR_high/low = 1.68, P = 0.07). GWAS identified one single-nucleotide polymorphism (SNP) meeting genome-wide significance, rs1377817 (P = 4.6 × 10^-8, a SNP intronic to MYH14).

Conclusions: This study indicates that residual platinum values are correlated with several cisplatin-related toxicities. One genetic variant is associated with these levels.

Figure 1.. Population Pharmacokinetic Modeling of Long-Term Serum Platinum Levels.

A) Serum platinum levels from 1,010 TCS were fitted to a bi-exponential model in which years since treatment completion was taken into account. Cumulative cisplatin dose was taken into account by multiplying the serum platinum levels of 400 ± 15 mg/m² patients by 0.75, enabling normalization to 300 ± 15 mg/m² patients prior to model fitting. B) Histogram of multiplicative residuals from the bi-exponential model. C) Residuals were log-transformed to fit a near normal distribution in order to examine the extent to which residual platinum values associate with cisplatin-induced toxicities. The bars denote groups of low, medium and high.

Figure 2.. Linear Regression of Residual Platinum Value and Continuous Variables.

Simple linear regression results are presented for A) age at diagnosis, B) creatinine clearance, and C) LH levels. R² and p-values for linear regression are reported for all phenotypes. A Spearman rank correlation test (r_sp) was also performed on LH levels due to its positive skew distribution, and results are shown in panel C. Fitted linear regression lines are highlighted in gray and 95% confidence intervals are indicated by the light gray shaded regions.

Figure 3.. Distributions of Cisplatin-Induced Toxicities in Testicular Cancer Survivors Based on Residual Platinum Value.

The overall distribution of A) CBM score for cisplatin-induced toxicities (p = 0.06), B) peripheral sensory neuropathy (PSN; p = 0.02), and C) Raynaud’s phenomenon (p = 0.02) in TCS based on having low, medium, and high residual platinum values is provided. Low, medium, and high groups reflect ordinal stratifications of residual platinum values based on their deviation from the mean: “medium” (regression residuals = 0 ± 1 standard deviation [SD]), “low” (residuals < −1 SD), and “high” (residuals > 1 SD). All three toxicities are divided into different degrees of severity, as indicated in the legend, with associated percentages provided in each panel. Sample sizes for each group are indicated within each panel on the x-axis. Differences between the proportions of toxicity severity observed for the low, medium, and high residual platinum value groups were evaluated for statistical significance through the Cochran-Armitage-Mantel 1df chi-square trend test (54).

Figure 4.. Genome-Wide Association Study of Residual Platinum Value as a Continuous Variable.

A) Manhattan plot of GWAS results reveals one locus meeting genome-wide significance (p ≤ 5 × 10−⁸): rs1377817 (p = 4.6 × 10−⁸). Covariates in the analysis include age at diagnosis and 10 genetic principal components accounting for population substructure. B) Quantile-Quantile plot of GWAS results.