Factors associated with longitudinal progression of the cumulative burden of morbidity and overall mortality after cisplatin-based chemotherapy for testicular cancer

Abstract

Background

To comprehensively evaluate the longitudinal progression of cumulative burden of morbidity (CBM) in testicular cancer survivors (TCS) following standard-dose cisplatin-based chemotherapy and the impact of modifiable risk factors on morbidity and early mortality.

Methods

Participants completed first-line chemotherapy at or longer than 6 months before baseline assessments with comprehensive questionnaires and physical examinations. Based on follow-up assessments (median: 7 years later), longitudinal progression of adverse health outcomes (AHOs) and CBM score (encompassing AHO number and severity) were examined. Baseline health behaviors and AHOs were evaluated for associations with mortality using mixed-effects parametric proportional-hazards regression to identify modifiable risk factors.

Results

Among 616 TCS longitudinally assessed, 23% experienced worsening CBM postchemotherapy (median = 11 years, interquartile range = 7-15). Declines were driven by worsening treatment-related AHOs: tinnitus (29.7%), hearing loss (24.4%), Raynaud’s disease (22.6%), neuropathy (18.5%), and neuropathic pain (10.7%). Baseline factors associated with worsening neuropathy included lack of aerobic physical activity (odds ratio [OR] = 1.98, 95% confidence interval [CI] = 1.06 to 3.72), and obesity (OR = 1.85, 95% CI = 1.17 to 2.92). These were also related to worsening neuropathic pain (OR = 2.82, P = .009 and OR = 2.29, P = .023). Twenty-nine deaths occurred among 1830 5-year TCS (4.2% cumulative hazard) (median age = 48 years, range = 22-74). Participants reporting neuropathic pain (hazard ratio [HR] = 3.64, 95% CI = 1.45 to 9.10), no aerobic (HR = 6.56, 95% CI = 2.73 to 15.8), or no low-impact physical activity (HR = 3.96, 95% CI = 1.40 to 11.2) had significantly higher mortality, as did TCS indicating fair (HR = 9.23, 95% CI = 3.08 to 27.8) or poor (HR = 18.5, 95% CI = 3.30 to 103) health. Relationships between pain and mortality were mediated through lowered physical activity (P = .036).

Conclusions

Clinically actionable factors associated with early mortality identify high-risk TCS in need of closer monitoring and targeted interventions. The significant relationship between neuropathic pain and mortality, mediated by low physical activity, is the first to our knowledge in TCS.

Introduction

Testicular cancer (TC) is the leading cancer in young men but has excellent long-term survival,¹ making TC survivors (TCS) an important group in which to characterize and quantify the late effects of cancer and its treatment. Cisplatin-based chemotherapy is standard-of-care and responsible for high cure rates but contributes to morbidities that can impact health-related quality of life and may adversely affect long-term survival.^2-4 We previously characterized the cumulative burden of morbidity (CBM) in TCS by quantifying the prevalence and severity of 21 adverse health outcomes (AHOs).⁵ Even at a young median age (37 years), almost 1 in 5 TCS had a high-to-severe CBM score; only 5% reported no AHOs. As median time from chemotherapy completion was less than 5 years, AHOs with longer latency periods, such as cardiovascular outcomes,² were likely not fully captured. Moreover, the severity of acute-onset adverse effects, such as neuropathy and ototoxicity, may worsen over time.⁶ Thus, there is a critical need for longitudinal assessment to understand late effects and CBM score trajectories and to identify modifiable factors that could delay or prevent AHO development.

Large population-based studies have shown that overall mortality is significantly higher among TCS than expected based on population rates in comparable age groups.^7-9 Wang et al. estimated a standardized mortality ratio of 1.14 (95% confidence interval [CI] = 1.04 to 1.24) among TCS in the SEER program (2000-2019).⁹ However, these analyses^7-9 were not able to consider detailed information on cancer treatment, health behaviors, and AHOs. This is important because several AHOs contributing to the CBM score have been associated with early TCS mortality.³

In this multicenter study, we comprehensively evaluate and quantify for the first time the longitudinal progression of the CBM score following standard cisplatin-based chemotherapy among a well-characterized TCS cohort.^5,10,11 We additionally assess the impact of clinical, sociodemographic, and behavioral factors on progression of cisplatin-related neuropathy, neuropathic pain, and overall mortality. Longitudinal characterization is important to develop risk-stratified, evidence-based follow-up recommendations and identify intervenable factors to reduce morbidity and early mortality.

Methods

Study population

The Platinum Study¹¹ was approved by institutional review boards and required written informed consent. Eligible participants had histologic/serological diagnosis of germ cell tumor, were aged 60 years or younger at diagnosis, completed first-line cisplatin-based chemotherapy at or longer than 6 months previously, and were undergoing routine follow-up at participating sites; 93% of all eligible patients enrolled, with participants referred to as TCS.¹¹ Baseline assessments at enrollment included physical examinations and comprehensive health questionnaires, health behaviors, symptoms, conditions, current prescription medications with indication (Supplementary Material S1). Cancer diagnosis/treatment data were abstracted from medical records. Follow-up assessments collected similar health questionnaires, with longitudinal results reported here for 616 TCS with complete assessments through January 16, 2024 (Figure S1).

Health behaviors and adverse health outcomes

Health behaviors included smoking status, alcohol consumption, and physical activity. Ten types of physical activity^12,13 categorized into 3 nonmutually exclusive dichotomous (yes/no) variables (ie, “aerobic,” “low impact,” “strength based”) were used to calculate kcals expended/week¹⁴ (Supplementary Material S1). Health conditions and prescription-medication use with indication were mapped to individual AHOs and graded with a 0-4 severity scale, as described previously⁵ (Supplementary Material S1). The SF-36¹⁵ captured self-reported health (SRH): excellent, very good, good, fair, or poor. CBM scores were calculated based on number and severity of AHOs, as described previously⁵ following methods adapted from Geenen et al.¹⁶ For longitudinal analysis, peripheral sensory neuropathy (PSN) was defined with the EORTC QLQ-Chemotherapy-Induced Peripheral Neuropathy-20 (CIPN-20)¹⁷ questionnaire items “a-f,” “I,” “j” (Supplementary Material S1) and prescription-medication use if the indication was neuropathy. Neuropathic pain was defined by EORTC-CIPN-20 items “e,” “f,” and prescription-medication use. Individual AHOs were selected for survival analysis based on prior literature.¹⁸

Statistical analysis

Descriptive statistics and regression analyses were performed using Stata-v15.1 (StataCorp.) unless otherwise noted. All tests were 2-sided (α = 0.05). Discrete and continuous data were described using numbers/(percentages) and medians/(ranges), respectively. Wilcoxon rank-sum tests compared medians. Sociodemographic, health behavior, cancer treatment, and AHO variables were individually tested for associations with worsening PSN and worsening neuropathic pain among TCS with data available at baseline and follow-up assessments and reporting no worse than grade-1 baseline neuropathy. Odds ratios (ORs), 95% confidence interval, and P values were calculated from mixed effects logistic regression, testing each variable for association with the dichotomous outcome of any vs no worsening neuropathy, adjusting for time since chemotherapy completion and enrollment site as fixed and random effects, respectively. Mixed-effects models were used to account for the fact that study participants are clustered within enrollment site, which was treated as a random effect, whereas other variables were modeled as fixed effects as in standard regression.

Analyses of overall mortality included all 5-year TCS. Deaths and cause of death (COD) were ascertained via electronic health records, communication with study coordinators at enrollment sites, and National Death Index searches. Survival time was defined as years from baseline assessment to date of non-TC death or last follow-up; TC deaths were treated as competing risks, as were 2 deaths due to accidents, and were censored at time of death. Sociodemographic, health behavior, cancer treatment, and AHO variables were each tested individually for association with overall mortality using a mixed-effects parametric survival model with a Weibull distribution, proportional hazards parameterization, and adjusting for age at baseline assessment and enrollment site as fixed and random effects, respectively, with an independent variance-covariance structure of random effects. Physical activity and body mass index (BMI) were assessed as potential mediating variables in the relationship between pain and mortality in separate path models (Mplus-software)¹⁹ (statistical methods are detailed in Supplementary Material S1).

Results

Among 1906 TCS with baseline assessments, 616 completed longitudinal assessments at a median of 7 years later (range = 3-10) (median age = 45 years, range = 22-79) (Table 1; Figure S1). Compared with baseline, greater proportions at follow-up were married/living as married (66.3% vs 77.6%) and had completed college (73.6% vs 79.2%), likely reflecting the older follow-up age. Baseline and follow-up health behaviors were largely similar with only slightly lower proportions reporting low-impact, strength-based, or aerobic physical activity at follow-up (Table 1).

Table 1.

Sociodemographic characteristics and health behaviors among 616 survivors of cisplatin-treated germ cell tumors who completed assessments at baseline and follow-up.

	Baseline assessment	Follow-up assessment
	(2012-2018)	(2022-2024)
Sociodemographic characteristic
Age, years	38 [18-74]	45 [22-79]
Time (years) since completion of chemotherapy
<5	377 (61.2)	38 (6.2)
5-9	110 (17.9)	241 (39.1)
10-14	64 (10.4)	176 (28.6)
$\ge$15	65 (10.5)	161 (26.1)
Racea		—
Asian	21 (3.4)
Black or African American	7 (1.4)
White	556 (90.5)
Other	30 (4.9)
Ethnicityb		—
Hispanic	27 (4.4)
Non-Hispanic	588 (95.6)
Marital statusc
Married or living as married	406 (66.3)	467 (77.6)
Single, never married	172 (28.1)	99 (16.4)
Divorced or separated	34 (5.6)	36 (6.0)
Educational attainmentd
College graduate or graduate education	452 (73.6)	482 (79.2)
Posthigh school education or training	122 (19.9)	94 (15.4)
High school or less	40 (6.5)	33 (5.4)
Employment statuse
Employed part- or full-time	547 (89.2)	525 (89.4)
Unemployed	38 (6.2)	7 (1.2)
Retired	12 (2.0)	38 (6.5)
On disability	16 (2.6)	17 (2.9)
Health insurancef
Yes	597 (96.9)	589 (95.8)
No	19 (3.1)	26 (4.2)
Health behavior
Smoking statusg
Never	405 (65.8)	369 (64.7)
Former	183 (29.8)	180 (31.6)
Current	27 (4.4)	21 (3.7)
Alcohol consumptionh
One drink per day or less	549 (89.1)	539 (87.6)
More than one drink per day	67 (10.9)	76 (12.4)
Low-impact physical activityi
Yes	587 (96.1)	579 (94.4)
No	24 (3.9)	34 (5.6)
Strength-training physical activityj
Yes	313 (52.2)	296 (48.8)
No	287 (47.8)	311 (51.2)
Aerobic physical activityk
Yes	560 (91.3)	530 (86.2)
No	53 (8.7)	85 (13.8)
kcals/week of physical activityl	2360 [0-28 790]	1910 [0-17 920]

Continuous variables are represented as the median (range), and categorical variables are represented as the number (percentage). The percents indicate the column percentages.

^aRace was not reported for 2 participants. Among the 30 participants for whom race is categorized as “Other,” 14 selected more than 1 race and 16 selected “Other” on the questionnaire.

^bEthnicity was not reported for 1 participant.

^cMarital status was not reported for 4 participants at baseline and for 14 participants at follow-up.

^dEducational attainment was not reported for 2 participants at baseline and for 7 participants at follow-up.

^eEmployment status was not reported for 3 participants at baseline and for 29 participants at follow-up. Among those who reported being on disability leave at the follow-up evaluation, the reasons were all health-related: pain and/or neuropathy, n = 6; anxiety and/or depression, n = 4; fatigue, n = 3; other. Among those who reported being unemployed, 4 participants indicated the reason was related to health.

^fHealth insurance status was not reported for 1 participant at the follow-up assessment.

^gSmoking status was not reported for 1 participant at baseline and for 46 participants at follow-up.

^hAlcohol consumption was not reported for 1 participant at the follow-up assessment. Among those who reported more than 1 drink per day, 10%, 1%, and 1% reported 2-3/day, 4-5/day, and 6+/day, respectively.

ⁱLow-impact physical activity was not reported for 5 participants at baseline and for 3 participants at follow-up.

^jStrength-based physical activity was not reported for 16 participants at baseline and for 9 participants at follow-up.

^kAerobic physical activity was not reported for 3 participants at baseline and for 1 participant at follow-up.

^lTotal kcals per week of physical activity was not available for 4 participants at baseline and for 6 participants at follow-up.

Approximately 23% of TCS experienced worsening CBM scores at a median of 11 years (interquartile range = 7-15) since chemotherapy (Figure 1; Table S1). Self-reported health declined in 38.2% (Figure 1; Table S1), reflecting worsening CBM scores. Declines in CBM score were driven by worsening treatment-related AHOs (Table 2) including tinnitus (29.7% TCS), hearing loss (24.4%), Raynaud’s phenomenon (22.6%), PSN (18.5%), and neuropathic pain (10.7%). After adjusting for enrollment site and time (years) from completion of chemotherapy to the baseline assessment, TCS either with no college degree or on disability leave had increased odds of worsening PSN (OR = 1.54, 95% CI = 1.00 to 2.35 and OR = 4.48, 95% CI = 1.38 to 14.5, respectively; Table 3). Modifiable factors associated with worsening PSN included obesity (OR = 1.85, 95% CI = 1.17 to 2.92) and lack of aerobic physical activity (OR = 1.98, 95% CI = 1.06 to 3.72). The same factors were associated with worsening neuropathic pain, with greater OR than seen for overall PSN. Compared with BEP×3, chemotherapy regimens including 4 cycles were associated with increased odds of worsening neuropathic pain: EP×4 (OR = 2.67, P = .027); BEP×4 (OR = 2.42, P = .017). As this longitudinal study is ongoing, we assessed potential selection bias by comparing baseline CBM scores for 616 TCS completing follow-up surveys by January 16, 2024 vs all other TCS, with no significant difference observed (χ²  P = .81).

Figure 1.

Longitudinal course of (A) cumulative burden of morbidity (CBM) score and (B) self-reported health (SRH) among 616 survivors of cisplatin-treated germ cell tumors (GCTs). TCS = testicular cancer survivor.

Table 2.

Adverse health outcomes (AHOs) at baseline, follow-up, and longitudinal course for 616 survivors of cisplatin-treated germ cell tumors who completed both assessments.

	Baseline assessment	Follow-up assessment	Longitudinal course
	(2012-2018)	(2022-2024)
Platinum-related AHOs
Peripheral sensory neuropathy
Grade 0 or 1	436 (70.8)	485 (78.7)	—
Grade 2+	180 (29.2)	131 (21.3)	—
Individual course:
Worsened	—	—	114 (18.5)
Remained the same	—	—	325 (52.8)
Improved	—	—	177 (28.7)
Neuropathic pain
Grade 0 or 1	562 (91.2)	573 (93.0)	—
Grade 2+	54 (8.8)	43 (7.0)	—
Individual course:
Worsened	—	—	66 (10.7)
Remained the same	—	—	477 (77.4)
Improved	—	—	73 (11.9)
Hearing damage
Grade 0 or 1	520 (84.4)	511 (83.0)	—
Grade 2+	96 (15.6)	105 (17.0)	—
Individual course:
Worsened	—	—	150 (24.4)
Remained the same	—	—	392 (63.4)
Improved	—	—	74 (12.0)
Tinnitus
Grade 0 or 1	515 (83.6)	452 (73.4)	—
Grade 2+	101 (16.4)	164 (26.6)	—
Individual course:
Worsened	—	—	183 (29.7)
Remained the same	—	—	369 (59.9)
Improved	—	—	64 (10.4)
Kidney disease
Grade 0	600 (97.4)	590 (95.8)	—
Grade 1+	16 (2.6)	26 (4.2)	—
Individual course:
Worsened	—	—	12 (2.0)
Remained the same	—	—	604 (98.0)
Improved	—	—	0
Cardiovascular AHOs
Hypertension
Grade 0a	558 (90.6)	507 (82.3)	—
Grade 2+	58 (9.4)	109 (17.7)	—
Individual course:
Worsened	—	—	12 (2.0)
Remained the same	—	—	600 (97.4)
Improved	—	—	4 (0.6)
Hyperlipidemia
Grade 0a	559 (90.7)	517 (83.9)	—
Grade 2+	57 (9.3)	99 (16.1)	—
Individual course:
Worsened	—	—	60 (9.7)
Remained the same	—	—	538 (87.3)
Improved	—	—	18 (2.9)
Thromboembolic event
Grade 0a	564 (91.6)	556 (90.3)	—
Grade 2+	52 (8.4)	60 (9.7)	—
Individual course:
Worsened	—	—	10 (1.6)
Remained the same	—	—	606 (98.4)
Improved	—	—	0
Heart disease
Grade 0 or 1	604 (98.1)	589 (95.6)	—
Grade 2+	12 (1.9)	27 (4.4)	—
Individual course:
Worsened	—	—	16 (2.6)
Remained the same	—	—	600 (97.4)
Improved	—	—	0
Peripheral artery disease
Grade 0 or 1	610 (99.0)	599 (97.2)	—
Grade 2+	6 (1.0)	17 (2.8)	—
Individual course:
Worsened	—	—	25 (4.1)
Remained the same	—	—	591 (95.9)
Improved	—	—	0
Other AHOs
Obesity
Grade 0b	177 (28.7)	197 (32.0)	—
Grade 2+	439 (71.3)	419 (68.0)	—
Individual course:
Worsened	—	—	70 (11.4)
Remained the same	—	—	466 (75.6)
Improved	—	—	80 (13.0)
Raynaud’s disease
Grade 0 or 1	511 (83.0)	507 (82.3)	—
Grade 2+	105 (17.0)	109 (17.7)	—
Individual course:
Worsened	—	—	139 (22.6)
Remained the same	—	—	375 (60.9)
Improved	—	—	102 (16.6)
Depression and/or anxiety
Grade 0a	573 (93.0)	556 (90.3)	—
Grade 2+	43 (7.0)	60 (9.7)	—
Individual course:
Worsened	—	—	38 (6.2)
Remained the same	—	—	557 (90.4)
Improved	—	—	21 (3.4)
Diabetes
Grade 0a	601 (97.6)	586 (95.1)	—
Grade 2+	15 (2.4)	30 (4.9)	—
Individual course:
Worsened	—	—	17 (2.8)
Remained the same	—	—	598 (97.1)
Improved	—	—	1 (0.2)c
Cerebrovascular disease
Grade 0	612 (99.3)	610 (99.0)	—
Grade 1+	4 (0.6)	6 (1.0)	—
Individual course:
Worsened	—	—	2 (0.3)
Remained the same	—	—	614 (99.7)
Improved	—	—	0

Data are represented as the number (percentage) of participants, where the percentages are column percentages.

^aData needed to assign grade 1 according to the Common Terminology Criteria for Adverse Events (version 4.03) definitions were not captured in the current study.

^bThe Common Terminology Criteria for Adverse Events (version 4.03) does not include a grade 1 for obesity.

^cThis participant reported taking insulin for diabetes at the baseline assessment (grade 3) and at the follow-up assessment reported taking metformin (grade 2), resulting in an improvement, although the condition was still present.

Table 3.

Association of sociodemographic characteristics, health behaviors, and adverse health outcomes at baseline with worsening neuropathy^a at follow-up.

	Worsening of any peripheral sensory neuropathy		Worsening of neuropathic pain
	OR (95% CI)	P	OR (95% CI)	P
Sociodemographic characteristic at baseline
Age (per year)	1.00 (0.98 to 1.03)	0.998	1.01 (0.98 to 1.04)	0.425
Race
White	Referent	—	Referent	—
Non-White	1.41 (0.75 to 2.66)	0.286	1.71 (0.75 to 3.92)	0.204
Ethnicity
Non-Hispanic	Referent	—	Referent	—
Hispanic	1.59 (0.66 to 3.86)	0.306	1.28 (0.36 to 4.56)	0.706
Educational attainment
College or graduate school	Referent	—	Referent	—
Less than college graduate	1.54 (1.00 to 2.35)	0.048	2.58 (1.46 to 4.55)	0.001
Employment statusa
Employed part- or full-time	Referent	—	Referent	—
Unemployed	1.38 (0.64 to 2.98)	0.418	1.25 (0.42 to 3.75)	0.689
Retired	0.32 (0.04 to 2.57)	0.283	NAb	NAb
On disability	4.48 (1.38 to 14.5)	0.013	7.30 (2.11 to 25.3)	0.002
Health behavior at baseline
Smoking status
Never	Referent	—	Referent	—-
Former	1.29 (0.85 to 1.95)	0.230	1.29 (0.72 to 2.33)	0.390
Current	2.12 (0.91 to 4.91)	0.080	2.22 (0.78 to 6.36)	0.136
Alcohol consumption
One drink per day or less	Referent	—	Referent	—
More than one drink per day	0.75 (0.40 to 1.44)	0.392	0.76 (0.29 to 1.98)	0.570
Aerobic physical activity
Yes	Referent	—	Referent	—
No	1.98 (1.06 to 3.72)	0.033	2.82 (1.30 to 6.13)	0.009
Strength-training physical activity
Yes	Referent	—	Referent	—
No	1.47 (1.00 to 2.17)	0.052	1.69 (0.96 to 2.96)	0.069
Low-impact physical activity
Yes	Referent	—	Referent	—
No	1.30 (0.51 to 3.30)	0.581	2.01 (0.64 to 6.32)	0.234
kcals per week of physical activityc	0.87 (0.75 to 1.01)	0.076	0.81 (0.66 to 0.99)	0.041
Cancer and treatment characteristics
Age at GCT diagnosis (per year)	1.00 (0.98 to 1.02)	0.894	1.01 (0.98 to 1.04)	0.555
Cumulative cisplatin dose (per 100 mg/m2)	1.24 (0.97 to 1.58)	0.090	1.23 (0.88 to 1.72)	0.234
Chemotherapy regimend
BEP×3	Referent	—	Referent	—
EP×4	1.68 (0.90 to 3.11)	0.101	2.67 (1.12 to 6.40)	0.027
BEP×4	1.56 (0.92 to 2.63)	0.097	2.42 (1.17 to 5.00)	0.017
Other	1.21 (0.64 to 2.29)	0.549	1.70 (0.67 to 4.31)	0.267
Adverse health outcome at baseline e
Obesity
Grade 0f	Referent	—	Referent	—
Grade 2+	1.85 (1.17 to 2.92)	0.008	2.29 (1.12 to 4.69)	0.023
Diabetes
Grade 0d	Referent	—	Referent	—
Grade 2+	0.99 (0.26 to 3.75)	0.988	0.89 (0.11 to 7.17)	0.911
Peripheral artery disease
Grade 0 or 1	Referent	—	Referent	—
Grade 2+	3.11 (0.61 to 15.9)	0.173	NAb	NAb
Hypothyroidism
Grade 0 or 1	Referent	—	Referent	—
Grade 2+	0.75 (0.08 to 6.78)	0.795	2.12 (0.23 to 19.8)	0.508

Analysis includes 573 TCS who completed both assessments and reported no worse than grade 1 neuropathy at baseline. ORs compare TCS with any increase in neuropathy grade to those TCS whose neuropathy grade remained grade 1 or 0, or improved from grade 1 to grade 0. ORs are from separate logistic regression models of the listed variable, adjusted for enrollment site as a random effect and time (years) from completion of chemotherapy to the baseline assessment as a fixed effect (refer to the “Methods” section). Statistically significant values are in bold font.

Abbreviations: AHO = adverse health outcome; CI = confidence interval; GCT = germ cell tumor; OR = odds ratio; TCS = testicular cancer survivor.

^aAmong the 7 participants who reported worsening of neuropathy and were on disability leave, the reason for each of them was health-related: pain and/or neuropathy, n = 4; fatigue, n = 2; anxiety and/or depression, n = 1.

^bOR and 95% CI were not estimable due to small numbers of participants.

^cKcals per week was log transformed due to right-skewed distribution.

^dData needed to assign grade 1 according to the Common Terminology Criteria for Adverse Events (version 4.03) definitions were not captured in the current study.

^eSelected AHOs were chosen based on reported associations with neuropathy in the literature.^20-23

^fThe Common Terminology Criteria for Adverse Events (version 4.03) does not include a grade 1 for obesity.

Among all 1906 individuals with baseline assessments, 38 deaths (cumulative hazard = 4.6%; median age = 43, range = 21-74) occurred at a median of 9 years (range = 3-41) since TC diagnosis. COD among 5-year survivors were 8—second malignant neoplasms, 4—infection, 5—TC, 3—cardiovascular disease, 2—suicide, 2—accidents, and 1—vascular intestinal disorder; 4—missing. COD within 5 years of diagnosis were 6—TC, 1—infection, 1—acute leukemia, and 1—complications of dermatomyositis. As our objective was to identify factors related to non-TC mortality, we restricted analyses to 5-year survivors and treated the few TC deaths and accidental deaths occurring after this time as competing risks. Baseline sociodemographic characteristics significantly associated with mortality included marital/relationship status, education, and employment status (Table 4). Lack of physical activity was a strong modifiable factor related to subsequent mortality, with the strongest association observed for lack of aerobic activity (hazard ratio [HR] = 6.56, 95% CI = 2.73 to 15.8; Table 4); this strong effect remained after adjustment for sociodemographic factors related to mortality (adjusted HR = 2.95, 95% CI = 1.03 to 8.45) (Table S2). Kilocalories expended per week during any type of exercise showed a strong dose-dependent association with subsequent mortality with 6-fold, 12-fold, and 33-fold risk reductions observed, respectively, for 1-499, 500-999, and 1000 kcal/week or more vs no physical activity (ie, HR = 0.16, P = .008; HR = 0.08, P = .002, and HR = 0.03, P < .001, respectively) (Table 4).

Table 4.

Association of baseline sociodemographic characteristics, health behaviors, and adverse health outcomes with subsequent mortality^a among 1830 5-year survivors of cisplatin-treated germ cell tumors.

	HR (95% CI)b	P
Sociodemographic characteristic at baseline b
Race
White	Referent	—
Non-White	0.83 (0.19 to 3.66)	0.808
Ethnicity
Non-Hispanic	Referent	—
Hispanic	NA	NA
Marital status
Married or living as married	Referent	—
Single, never married	3.27 (1.08 to 9.88)	0.036
Divorced or separated	6.98 (2.51 to 19.3)	<0.001
Educational attainment
College or graduate education	Referent	—
Less than college graduate	3.15 (1.28 to 7.71)	0.012
Employment status
Employed part- or full-time	Referent	—
Unemployed	4.48 (1.24 to 16.2)	0.022
Retired	4.89 (0.86 to 27.7)	0.073
On disability	17.9 (6.36 to 50.5)	<0.001
Health insurance
Yes	Referent	—
No	1.44 (0.40 to 5.25)	0.579
Health behavior at baseline c
Smoking status
Never	Referent	—
Former	1.39 (0.55 to 3.52)	0.492
Current	2.78 (0.86 to 8.95)	0.087
Alcohol consumption
One drink per day or less	Referent	—
More than one drink per day	1.35 (0.45 to 4.01)	0.588
Aerobic physical activity
Yes	Referent	—
No	6.56 (2.73 to 15.8)	<0.001
Strength-training physical activity
Yes	Referent	—
No	1.47 (0.60 to 3.61)	0.405
Low-impact physical activity
Yes	Referent	—
No	3.96 (1.40 to 11.2)	0.009
kcals/week of physical activityd	0.61 (0.46 to 0.80)	0.001
kcals/week of physical activity
None	Referent	—
1-499	0.16 (0.04 to 0.62)	0.008
500-999	0.08 (0.01 to 0.39)	0.002
$\ge$1000	0.03 (0.01 to 0.13)	<0.001
Cancer and treatment characteristic
Age at GCT diagnosis (per year)	0.95 (0.90 to 1.01)	0.121
Cumulative cisplatin dose (per 100 mg/m2)e	1.47 (0.93 to 2.33)	0.103
Chemotherapy regimenf
BEP×3	Referent	—
EP×4	2.30 (0.50 to 10.7)	0.286
BEP×4	2.54 (0.76 to 8.54)	0.131
Other	2.77 (0.76 to 10.2)	0.124
Adverse health outcome at baseline c
Obesity
Grade 0g	Referent	—
Grade 2+	0.85 (0.33 to 2.21)	0.745
Thromboembolic event
Grade 0f	Referent	—
Grade 2+	1.20 (0.28 to 5.14)	0.807
Heart disease
Grade 0	Referent	—
Grade 1+	6.53 (1.88 to 22.6)	0.003
Cerebrovascular disease
Grade 0	Referent	—
Grade 1—transient ischemic attack	9.04 (1.14 to 71.9)	0.037
Grade 2—stroke	NA	NA
Depression and/or anxiety
Grade 0g	Referent	—
Grade 2+	2.36 (0.68 to 8.20)	0.175
Neuropathic pain
Grade 0 or 1	Referent	—
Grade 2+	3.64 (1.45 to 9.10)	0.006
Prescription opioid use for neuropathic painh
No	Referent	—
Yes	5.62 (1.23 to 25.6)	0.026
Hearing loss
Grade 0 or 1	Referent	—
Grade 2+	1.27 (0.46 to 3.51)	0.651
Tinnitus
Grade 0 or 1	Referent	—
Grade 2+	1.06 (0.35 to 3.15)	0.921
Raynaud’s disease
Grade 0 or 1	Referent	—
Grade 2+	0.96 (0.32 to 2.83)	0.936
Self-reported health
Excellent or very good	Referent	—
Good	1.89 (0.63 to 5.67)	0.255
Fair	9.23 (3.08 to 27.8)	<0.001
Poor	18.5 (3.30 to 103.9)	0.001
Cumulative burden of morbidity score
None or very low	Referent	—
Low or medium	2.08 (0.27 to 16.2)	0.484
High, very high, or severe	4.83 (0.59 to 39.3)	0.141

HRs are from separate mixed-effects proportional hazards models of the listed variable, adjusted for enrollment site as a random effect and age at baseline as a fixed effect (refer to the “Methods” section). Bold values are those with P < .05.

Abbreviations: BEP×3 = 3 cycles of bleomycin, etoposide, and cisplatin; BEP×4 = 4 cycles of bleomycin, etoposide, and cisplatin; CI = confidence interval; EP×4 = 4 cycles of etoposide and cisplatin; GCT = germ cell tumor; HR = hazard ratio; TCS = testicular cancer survivor.

^aCause of death among 5-year survivors were 8—second malignant neoplasms (SMN), 4—infection, 5—testicular cancer (TC), 3—cardiovascular disease, 2—suicide, 2—accidents, and 1—vascular intestinal disorder; 4—missing. Of the 8 SMN, 2—were acute myeloblastic leukemia, 1—malignant neoplasm of kidney, 1—malignant neoplasm of bladder, 1—non-Hodgkin lymphoma, 1—metastatic adenocarcinoma of unknown primary site, 1—parietal-occipital gliosarcoma, and 1—colorectal adenocarcinoma stage IV. The 5 TC deaths and 2 accidental deaths were censored at time of death in analyses of overall mortality.

^bHRs and P values are adjusted for age at baseline assessment and enrollment center and are based on 30 deaths that occurred 5 or more years after GCT diagnosis.

^cBaseline data were collected at the time of the first assessment.

^dkcals per week was log-transformed due to right-skewed distribution.

^eParticipants who received carboplatin for 1 or more cycles and, therefore, received fewer than 3 cycles of cisplatin, are excluded from analysis of the cumulative cisplatin dose and the number of cycles of cisplatin.

^fChemotherapy regimens administered to the cohort of 1830 5-year TCS are as follows: BEP×3 includes 678 patients who received the standard 3 cycles of BEP, and 13 patients who received 3 cycles of a combination of BEP and EP. BEP×4 includes 239 who received the standard 4 cycles of BEP, 91 patients who received 4 cycles of a combination of BEP and EP. EP×4 includes 515 patients who received the standard 4 cycles of EP. The “other” group includes 29 who received fewer than 3 cycles of BEP, 6 who received more than 4 cycles of BEP, 42 who received fewer than 4 cycles of EP, 8 who received more than 4 cycles of EP, 48 who received carboplatin for 1 or more cycles as part of their cisplatin-based chemotherapy, 14 who received an unknown number of cycles of BEP or EP, and 147 who received another combination of cisplatin-based chemotherapy.

^gThe Common Terminology Criteria for Adverse Events (version 4.03) does not include grade 1 for obesity, and data needed to identify grade 1 for thromboembolic events and depression/anxiety were not collected.

^hNon-opioid medications reported for use in treating neuropathic pain included gabapentin (n = 26), pregabalin (n = 9), duloxetine (n = 3), carbamazepine (n = 2), amitriptyline (n = 2), and nortriptyline (n = 1).

Among baseline AHOs, cerebrovascular disease (HR = 9.04, 95% CI = 1.14 to 71.9), heart disease (HR = 6.53, 95% CI = 1.88 to 22.6), and neuropathic pain (HR = 3.64, 95% CI = 1.45 to 9.10) conferred significantly elevated risks of subsequent mortality (Table 4). Treatment with prescription opioids for neuropathic pain was associated with significantly increased 5-fold mortality (HR = 5.62, 95% CI = 1.23 to 25.6). Mediation analysis showed a significant indirect effect of pain on mortality mediated through overall quantity of physical activity (ie, HR = 3.64 for pain without physical activity as a mediator, HR = 3.03 with physical activity as a mediator; indirect effect P = .036). No evidence existed for a mediating effect of BMI (indirect effect P = .99). There was no statistically significant increase in mortality among TCS given higher cumulative cisplatin doses (HR per 100 mg/m² = 1.47, 95% CI = 0.93 to 2.33). CBM score showed a nonsignificant 4- to 5-fold association with overall mortality: HR = 4.83 (95% CI = 0.59 to 39) for those with baseline CBM scores of high, very high, or severe. Self-reported health showed a more striking and significant association (Table 4, Figure 2); mortality was elevated 9-fold and 18-fold, respectively, among those indicating fair (HR = 9.23, 95% CI = 3.08 to 27.8) or poor (HR = 18.5, 95% CI = 3.30 to 103) health vs excellent or very good health.

Figure 2.

Overall mortality among 1830 5-year survivors of cisplatin-treated germ cell tumors by self-reported health at baseline assessment.

Discussion

Our results are based on the largest study to date of TCS given standard-dose cisplatin-based chemotherapy for whom detailed clinical information and longitudinal assessments are available. For the first time, to our knowledge, we show that nearly 1 in 4 TCS experienced worsening CBM scores at a median of 11 years since chemotherapy and that neuropathic pain was associated with an increased risk of mortality. Although CBM scores capture the incidence and severity of many AHOs, those for which the highest proportion of TCS experienced worsening scores were cisplatin-related, that is, neurotoxicity and ototoxicity.

Neuropathy is a particularly troublesome AHO given the lack of effective treatments,²⁴ with neuropathic pain associated with significantly worse TCS global mental health.²⁰ Here, we extend these findings to show that it may be prognostic for overall mortality: TCS reporting moderate or worse neuropathic pain at baseline subsequently had significantly increased 3.6-fold mortality risks mediated by low or no physical activity. These results are plausible because pain could limit participation in different types of physical exercise, and we found a strong inverse dose-dependent association of total kilocalories expended/week with mortality risk. Postulated mechanisms for associations between chronic pain and mortality in noncancer populations vary, but pain-related limitations in physical activity are frequently mediators.^25,26 Among 500 434 UK-Biobank patients, excess mortality from chronic pain (ie, >3 months)²⁵ was substantially attenuated after adjustment for low physical activity, high BMI, smoking, and poor diet, suggesting these as possible mediators. Mediation analyses in a population-based study in England of 6324 adults age $\ge$50 years with self-reported pain (ie, “Are you often troubled by pain? yes/no”)²⁶ found the strongest, statistically significant mediating factors between “troubling pain” and mortality were functional limitations (HR = 1.31), symptoms preventing walking 1/4 mile (HR = 1.45), physical inactivity (HR = 1.14), and poor SRH (HR = 1.32). In a large population-based investigation, US patients prescribed long-acting opioids for moderate-to-severe chronic pain (vs alternative medications) had significantly increased 1.6-fold risks of all-cause mortality, with cardiovascular disease (CVD) the most frequent nonoverdose COD.²⁷ Assessments of risks and benefits of opioid use by cancer survivors are complex,²⁸ and there is an urgent unmet need for safe and effective alternative management approaches for neuropathic pain.²⁴

It is critical to note that obesity and lack of physical activity are potentially modifiable factors associated with worsening neuropathy and neuropathic pain that might be targeted in complementary approaches to pharmaceutical management strategies. A systematic review of randomized controlled trials of exercise for reducing CIPN symptoms²⁹ showed that exercise is beneficial, consistent with our findings, but did not include TCS studies; although obesity per se was not addressed, typically adherence to an exercise regimen is also beneficial for weight loss.³⁰ Cumulative cisplatin dose did not emerge as a significant predictor of worsening neuropathy. It is plausible that the effect of cisplatin dose on neuropathy is already captured in the baseline survey, given the acute onset (baseline survey at a median of 3 years after chemotherapy), thus, neuropathy may not worsen further at the later survey as a function of cisplatin dose.

Associations of lack of physical activity at baseline with both worsening morbidity (eg, neuropathy) and mortality underscore an important opportunity to modify TCS outcomes. Our finding that aerobic physical activity is protective independent of sociodemographic factors highlights the potential survival benefit of this modifiable behavior to all TCS. This finding agrees with a more general mortality analysis in Norwegian TCS (diagnosed 1980-1994) that broadly evaluated a composite variable of “unhealthy vs healthy” lifestyle¹⁸ (smoking, alcohol use, obesity, low physical activity, probable depressive disorder), but where physical activity was assessed with one question. With only 11 deaths among 124 Norwegian TCS given standard-dose chemotherapy, the effect of this composite variable on overall mortality was not evaluated.¹⁸ A subsequent analysis of the 1 physical activity question³¹ did not report the effect after standard-dose cisplatin. Treatment in this cohort was also heterogeneous, participants were older, and consequently mortality was higher. Our cohort is considerably younger, treated more recently (>90% in 2000 or later) with contemporary standard-dose cisplatin-based chemotherapy, and assessed with a validated bank of physical activity questions (Supplementary Material S1).^12,13 Our larger sample size revealed that physical activity is prognostic for overall survival even at a median of 11 years after treatment when modification of this behavior might alter the course of survival.

Lifestyle interventions to reduce the known elevated risk of CVD, stroke, and metabolic syndrome, and risk factors (eg, obesity/hypertension/hyperlipidemia) in TCS, were recently reviewed.² Underlying mechanisms of physical activity’s protective effect include reduction of inflammation,^32-34 which plays a role in the development/progression of many chronic health conditions, including ischemic heart disease, stroke, diabetes, and others elevated in TCS.^32-34 Benefits of exercising with and after a cancer diagnosis are numerous. In cancer survivors, exercise can improve health-related quality of life, reduce fatigue, increase physical function, and decrease anxiety/depression.³⁵ Importantly, our data now show that even in this relatively young population, it may be associated with reduced mortality. In fact, TCS reporting as little as less than 500 kcal of exercise per week had significantly improved overall survival vs those not exercising at all, suggesting that benefits can be gained even with small behavioral changes. In turn, many institutions are now instituting formalized cardiac-rehabilitation programs after cancer therapy³⁶ that could be recommended for TCS. Physical activity counseling could also be tailored for delivery by primary-care providers following TCS long term.

Worsening CBM scores coincided with worsening SRH, demonstrating that the contributing factors are not subclinical, but rather those that directly impact patients’ self-perception of health. CBM scores showed a marginal association with higher overall mortality, with SRH showing a more striking and significant association (eg, 14-fold increased mortality risk among those reporting poor health at baseline). The relationship between SRH and mortality is consistent with findings among older adults in the general population^37,38 and we have now extended the finding to younger TCS. Importantly, this patient-reported outcome can be easily queried by health-care providers during survivorship clinic visits. Follow-up questions which ascertain the specific AHOs contributing to poor health perception can be addressed among those most in need (ie, those reporting fair or poor health) as part of precision survivorship care.

Major study strengths include longitudinal assessment of health behaviors and a comprehensive set of AHOs, treatment with contemporary chemotherapy at major cancer centers, and minimal selection bias due to the 93% participation rate.¹¹ Additional strengths include the large sample size and availability of cause of death, enabling overall mortality analyses even at a relatively early median follow-up (11 years), and clinically actionable findings. Inherent limitations are that the impact on mortality of relatively rare AHOs and those with long latency periods, such as CVD,² may not be fully evident. Although TCS older than 60 years at diagnosis were not included given our long-term follow-up plans, these account for only about 4% of all TC.³⁹ Limitations common to studies of cancer survivors are lack of pretreatment assessments and cohort entry at different times since diagnosis.^40-44 We accounted for the latter by adjusting for age at baseline assessment in all regression models.

In conclusion, almost 1 in 4 TCS experienced worsening CBM scores at a median of 7 years since first assessment, driven by worsening treatment-related AHOs. Neuropathic pain was associated with worse overall survival, as were sociodemographic characteristics, whereas physical activity was protective. The finding that physical activity could lessen the apparent influence of sociodemographic factors is notable, as this represents a modifiable health behavior that could be reinforced to TCS. Factors associated with progression of morbidity and early mortality, including neuropathic pain, identify high-risk TCS in need of closer monitoring and targeted interventions through precision survivorship care. Though the relationship between neuropathic pain and mortality was mediated through physical activity, it nonetheless serves as a useful symptom that can be readily monitored in the clinic. Importantly, as reviewed elsewhere,²⁴ it will be critical to develop better treatments and management strategies for neuropathic pain.

Author contributions

Sarah L. Kerns (Conceptualization, Formal analysis, Writing—original draft, Writing—review & editing), Paul C. Dinh (Conceptualization, Data curation, Writing—original draft, Writing—review & editing), Patrick O. Monahan (Formal analysis, Methodology, Writing—review & editing), Timothy Stump (Formal analysis, Writing—review & editing), Chunkit Fung (Data curation, Writing—review & editing), Howard D. Sesso (Methodology, Writing—review & editing), Darren R. Feldman (Data curation, Writing—review & editing), Robert J. Hamilton (Data curation, Writing—review & editing), David J. Vaughn (Data curation, Writing—review & editing), Robert Huddart (Data curation, Writing—review & editing), Christian Kollmannsberger (Data curation, Writing—review & editing), Neil E. Martin (Data curation, Writing—review & editing), Kathryn Nevel (Data curation, Writing—review & editing), John Kincaid (Data curation, Writing—review & editing), Lawrence H. Einhorn (Data curation, Writing—review & editing), and Lois B. Travis (Conceptualization, Data curation, Funding acquisition, Project administration, Supervision, Writing—original draft, Writing—review & editing).

Supplementary material

Supplementary material is available at JNCI: Journal of the National Cancer Institute online.

Funding

This work was supported by the National Institutes of Health National Cancer Institute 2R01 CA157823 (L.B.T.), K07 CA187546 (S.L.K.), and Cancer Center Support Grant P30 CA008748 (D.R.F.). The funder did not have any role in the design of the study, data collection and analysis, the decision to publish, or preparation of the manuscript.

Conflicts of interest

S.L.K., D.R.F., D.J.V., and L.B.T. report research funding (institutional) from the National Institutes of Health. C.F. reports Open Payments Link: https://openpaymentsdata.cms.gov/physician/450635. D.R.F. reports Consulting or Advisory Role: Telix Pharmaceuticals, BioNTech, Renibus; Research Funding: Telix Pharmaceuticals, Decibel Therapeutics, Astellas Pharma, Exelixis; Royalties for authorship of topic review: UpToDate. K.N. reports Personal Fees: Aptitude Health, LLC; payment for answering survey on glioma management. J.K. reports Subinvestigator for Ionis Pharmaceuticals in a treatment trial of a compound that treats familial amyloid neuropathy. All other authors report no conflicts of interest related to the work in this publication.

Data availability

Deidentified data that support the findings of this study are available upon request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.