Clinical and Genome-Wide Analysis of Multiple Severe Cisplatin-Induced Neurotoxicities in Adult-Onset Cancer Survivors

Abstract

Purpose: Cisplatin is a first-line chemotherapeutic for many cancers, but causes neurotoxicity including hearing loss, tinnitus, and peripheral sensory neuropathy. However, no study has comprehensively characterized risk factors for developing multiple (>1) severe neurotoxicities.

Experimental design: The relationship between multiple severe neurotoxicities and age, cumulative cisplatin dose, medical history, and lifestyle/behavioral factors was evaluated in 300 cisplatin-treated testicular cancer survivors using logistic regression. Case-control genome-wide association study (GWAS; cases, n = 104 and controls, n = 196) was also performed.

Results: Age at clinical examination (P = 6.4 × 10^-16) and cumulative cisplatin dose (P = 5.4 × 10^-4) were positively associated with multiple severe neurotoxicity risk, as were high serum platinum levels (P = 0.02), tobacco use (ever smoker, P = 0.001 and current smoker, P = 0.002), and hypertension (P = 0.01) after adjustment for age and cumulative cisplatin dose. Individuals with multiple severe neurotoxicities were more likely to experience dizziness/vertigo (P = 0.01), Raynaud phenomenon (P = 3.7 × 10^-9), and symptoms consistent with peripheral motor neuropathy (P = 4.3 × 10^-14) after age and dose adjustment. These patients also reported poorer overall health (P = 2.7 × 10^-5) and a greater use of psychotropic medications (P = 0.06). GWAS identified no genome-wide significant SNPs. Gene-based association analysis identified RGS17 (P = 3.9 × 10^-5) and FAM20C (P = 5.5 × 10^-5) as near genome-wide significant. Decreased FAM20C expression was associated with increased cisplatin sensitivity in tumor cell lines.

Conclusions: Certain survivors are more susceptible to cisplatin-induced neurotoxicity, markedly increasing likelihood of developing numerous neuro-otological symptoms that affect quality of life. Genome-wide analysis identified genetic variation in FAM20C as a potentially important risk factor.

Figure 1.. Relationships Between Cisplatin-Induced Hearing Loss, Tinnitus, and Peripheral Sensory Neuropathy (PSN) in Testicular Cancer Survivors.

Associations between A) hearing loss and tinnitus (p<2x10⁻¹⁶), B) hearing loss and peripheral sensory neuropathy (p=3.94x10⁻¹²), and C) peripheral sensory neuropathy and tinnitus (p<2x10⁻¹⁶) are highly significant, forming the basis for combining the most severe forms of all three toxicities into a single phenotype. Statistical significance was assessed through ordinal logistic regression.

Figure 2.. Effects of Cumulative Cisplatin Dose and Residual Platinum Value on Proportion of Patients with Multiple Severe Cisplatin-Induced Neurotoxicities.

The overall proportion of testicular cancer survivors with multiple severe cisplatin-induced neurotoxicities is shown based on A) cumulative cisplatin dose and B) residual platinum value. Cumulative cisplatin dose (age-adjusted OR/100 mg/m²=1.9, 95% CI: 1.3-3.0, p=0.003) and residual platinum values (age and dose-adjusted OR=3.0, 95% CI: 1.3-7.7, p = 0.01) were significantly associated with multiple severe neurotoxicities. Patients who received 400 mg/m² cisplatin had a notably increased likelihood of developing multiple severe neurotoxicities when compared to patients who received 300 mg/m² (41.2% vs. 24.6%, p=0.007). Patients with high residual platinum values also had a significantly greater incidence of multiple severe neurotoxicities (58.8%) than those with medium (31.3%) or low (16.7%) serum platinum levels (p=0.02). Statistical significance is based on the two-proportions z-test, and sample sizes for each group are indicated within each panel on the x-axis.

Figure 3.. Distributions of Comorbidities in Testicular Cancer Survivors Based on Multiple Severe Cisplatin-Induced Neurotoxicities.

The overall distribution of A) Raynaud phenomenon, B) peripheral motor neuropathy, and C) self-reported health in testicular cancer survivors based on the occurrence of multiple severe neurotoxicities is provided. Patients with multiple severe neurotoxicities were more likely to experience Raynaud phenomenon (age and dose-adjusted OR=3.5, 95% CI: 2.4-5.5, p<0.0001), report symptoms of peripheral motor neuropathy (age and dose-adjusted OR=14.3, 95% CI: 7.4-29.0, p<0.0001), and report poorer overall health (age and dose-adjusted OR=2.2, 95% CI: 1.5-3.3, p<0.0001). All three comorbidities are divided into different degrees of severity, as indicated in the legend. Patients with multiple severe neurotoxicities reported more severe forms of Raynaud phenomenon (p<0.0001), peripheral motor neuropathy (PMN; p <0.0001), and self-reported health (p<0.0001). Sample sizes for each group are indicated within each panel on the x-axis. Differences between the proportions of toxicity severity observed for cases and controls were evaluated for statistical significance through the Cochran-Armitage-Mantel 1df chi-squared trend test (38).

Figure 4.. Gene-Based Genome-Wide Association Analysis of Multiple Severe Cisplatin-Induced Neurotoxicities.

Summary statistics for SNP-based GWAS were uploaded to FUMA to run a gene-based association analysis based on a multiple linear principal components regression to determine the aggregated effect of all SNPs within a gene. Inputted SNPs were mapped to 18,404 protein coding genes, producing a significance threshold of p=0.05/18,404 (2.7x10⁻⁶). A) Manhattan plot of the gene-based association analysis identified RGS17 (p=3.9x10⁻⁵) and FAM20C (p=5.5x10⁻⁵) as near genome-wide significant. B) Quantile-Quantile plot of results from the gene-based association analysis. C) Scatter plot of cisplatin sensitivity as a function of normalized FAM20C expression are provided for CNS (ρ=0.29, p=0.06; R²=0.04, p=0.20). Cisplatin sensitivity, measured as the area under the cisplatin dose-response curve, for all CNS tumor cell lines was extracted from CancerRX, and gene expression data were downloaded from the Cancer Cell Line Encyclopedia. Expression data were ranked normalized to fit a normal distribution prior to analysis. Correlation was assessed nonparametrically using the Spearman rank method, as well as by linear regression.