Describing patterns of familial cancer risk in subfertile men using population pedigree data

Abstract

STUDY QUESTION

Can we simultaneously assess risk for multiple cancers to identify familial multicancer patterns in families of azoospermic and severely oligozoospermic men?

SUMMARY ANSWER

Distinct familial cancer patterns were observed in the azoospermia and severe oligozoospermia cohorts, suggesting heterogeneity in familial cancer risk by both type of subfertility and within subfertility type.

WHAT IS KNOWN ALREADY

Subfertile men and their relatives show increased risk for certain cancers including testicular, thyroid, and pediatric.

STUDY DESIGN, SIZE, DURATION

A retrospective cohort of subfertile men (N = 786) was identified and matched to fertile population controls (N = 5674). Family members out to third-degree relatives were identified for both subfertile men and fertile population controls (N = 337 754). The study period was 1966–2017. Individuals were censored at death or loss to follow-up, loss to follow-up occurred if they left Utah during the study period.

PARTICIPANTS/MATERIALS, SETTING, METHODS

Azoospermic (0 × 10⁶/mL) and severely oligozoospermic (<1.5 × 10⁶/mL) men were identified in the Subfertility Health and Assisted Reproduction and the Environment cohort (SHARE). Subfertile men were age- and sex-matched 5:1 to fertile population controls and family members out to third-degree relatives were identified using the Utah Population Database (UPDB). Cancer diagnoses were identified through the Utah Cancer Registry. Families containing ≥10 members with ≥1 year of follow-up 1966–2017 were included (azoospermic: N = 426 families, 21 361 individuals; oligozoospermic: N = 360 families, 18 818 individuals). Unsupervised clustering based on standardized incidence ratios for 34 cancer phenotypes in the families was used to identify familial multicancer patterns; azoospermia and severe oligospermia families were assessed separately.

MAIN RESULTS AND THE ROLE OF CHANCE

Compared to control families, significant increases in cancer risks were observed in the azoospermia cohort for five cancer types: bone and joint cancers hazard ratio (HR) = 2.56 (95% CI = 1.48–4.42), soft tissue cancers HR = 1.56 (95% CI = 1.01–2.39), uterine cancers HR = 1.27 (95% CI = 1.03–1.56), Hodgkin lymphomas HR = 1.60 (95% CI = 1.07–2.39), and thyroid cancer HR = 1.54 (95% CI = 1.21–1.97). Among severe oligozoospermia families, increased risk was seen for three cancer types: colon cancer HR = 1.16 (95% CI = 1.01–1.32), bone and joint cancers HR = 2.43 (95% CI = 1.30–4.54), and testis cancer HR = 2.34 (95% CI = 1.60–3.42) along with a significant decrease in esophageal cancer risk HR = 0.39 (95% CI = 0.16–0.97). Thirteen clusters of familial multicancer patterns were identified in families of azoospermic men, 66% of families in the azoospermia cohort showed population-level cancer risks, however, the remaining 12 clusters showed elevated risk for 2-7 cancer types. Several of the clusters with elevated cancer risks also showed increased odds of cancer diagnoses at young ages with six clusters showing increased odds of adolescent and young adult (AYA) diagnosis [odds ratio (OR) = 1.96–2.88] and two clusters showing increased odds of pediatric cancer diagnosis (OR = 3.64–12.63). Within the severe oligozoospermia cohort, 12 distinct familial multicancer clusters were identified. All 12 clusters showed elevated risk for 1–3 cancer types. An increase in odds of cancer diagnoses at young ages was also seen in five of the severe oligozoospermia familial multicancer clusters, three clusters showed increased odds of AYA diagnosis (OR = 2.19–2.78) with an additional two clusters showing increased odds of a pediatric diagnosis (OR = 3.84–9.32).

LIMITATIONS, REASONS FOR CAUTION

Although this study has many strengths, including population data for family structure, cancer diagnoses and subfertility, there are limitations. First, semen measures are not available for the sample of fertile men. Second, there is no information on medical comorbidities or lifestyle risk factors such as smoking status, BMI, or environmental exposures. Third, all of the subfertile men included in this study were seen at a fertility clinic for evaluation. These men were therefore a subset of the overall population experiencing fertility problems and likely represent those with the socioeconomic means for evaluation by a physician.

WIDER IMPLICATIONS OF THE FINDINGS

This analysis leveraged unique population-level data resources, SHARE and the UPDB, to describe novel multicancer clusters among the families of azoospermic and severely oligozoospermic men. Distinct overall multicancer risk and familial multicancer patterns were observed in the azoospermia and severe oligozoospermia cohorts, suggesting heterogeneity in cancer risk by type of subfertility and within subfertility type. Describing families with similar cancer risk patterns provides a new avenue to increase homogeneity for focused gene discovery and environmental risk factor studies. Such discoveries will lead to more accurate risk predictions and improved counseling for patients and their families.

STUDY FUNDING/COMPETING INTEREST(S)

This work was funded by GEMS: Genomic approach to connecting Elevated germline Mutation rates with male infertility and Somatic health (Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD): R01 HD106112). The authors have no conflicts of interest relevant to this work.

TRIAL REGISTRATION NUMBER

N/A.

Introduction

Male subfertility is posited to be a biomarker for overall somatic health. Poor semen quality and infertility are linked with several adverse health outcomes including increased risk of hospitalization and mortality from chronic conditions, decreased lifespan, and increased risk of cardiovascular disease, metabolic syndrome, and autoimmune conditions (Jensen et al., 2009; Latif et al., 2017, 2018; Choy and Eisenberg, 2018; Hanson et al., 2018; Burke et al., 2022; Chen et al., 2022). Subfertility has also been linked with increased risk for genitourinary cancers, particularly testis, and possibly linked with prostate and colon cancers (Jacobsen et al., 2000; Hanson et al., 2016, 2018; Choy and Eisenberg, 2018). There is also mounting evidence that male subfertility may serve as a cancer risk marker for family members, first- and second-degree relatives of subfertile men exhibit increased risk for testicular, thyroid, and childhood cancers (Anderson et al., 2016, 2017). However, the extent to which cancer risk patterns vary between families of subfertile men is unclear.

Familial cancer risk associations suggest underlying heritable genetic risk factors, shared environmental exposures and health behaviors, or a combination of genetic and environmental components of risk (Teerlink et al., 2012; Frank et al., 2015; Wu et al., 2018; Hemminki et al., 2021), Different etiologic factors may manifest as distinct familial multicancer patterns leading to cancer pleiotropies, a well-accepted occurrence (Lorenzo Bermejo and Hemminki, 2005; Teerlink et al., 2012; Wu et al., 2018). Characterization of these patterns is important for accurate risk predictions and patient counseling. Using breast cancer as an example, approximately 30% of hereditary breast cancer is explained by intermediate- and high-risk inherited variants, such as BRCA1/2. Following the discovery of BRCA1/2 through study of families with dense clustering of breast and ovarian cancers, unique familial multicancer configurations were described for carriers of BRCA1 (breast, ovarian, Fanconi anemia, prostate, pancreatic, fallopian tube, and peritoneal cancers) and BRCA2 (breast, male breast, prostate, pancreatic cancers, and Fanconi anemia) (Hall et al., 1990; Mérette et al., 1992; Easton et al., 1993). The BRCA1/2 multicancer configurations were only identified following the discovery of the genetic variants. However, data-driven methods can be applied to describe multicancer patterns before discovering shared genetic and/or environmental factors, paving the way toward better descriptions of cancer pleiotropies for focused gene discovery and environmental risk factor studies (Hanson et al., 2020a,b).

As it is probable that families differentially contribute to aggregate cancer risk, identifying subsets of men with subfertility and their families who share risk patterns is crucial to understanding the mechanisms underlying these associations. Familial implications of male subfertility appear to be distinct by type of subfertility. Relative to fertile controls, first- and second-degree relatives of men with azoospermia had significantly increased risk of thyroid cancer, additionally, a subsequent study among siblings and cousins of oligozoospermic men showed increased risk of childhood cancer and acute lymphoblastic leukemia (Anderson et al., 2016, 2017). Identifying families with similar multicancer patterns that could be related to shared genetic, epigenetic, and environmental factors allows for categorization into more homogeneous cancer risk subtypes and may empower the discovery of meaningful determinants of fertility and cancer risk. Classical methods for assessing familial coaggregation of cancer risk use an iterative pairwise approach leaving them unable to simultaneously exploit data from multiple cancers and limiting power to identify multicancer patterns. This analysis utilizes a novel multicancer clustering technique to simultaneously assess multiple cancer phenotypes and group families with distinct multicancer patterns (Hanson et al., 2020b).

This study leverages a unique population-level data resource, the Utah Population Database (UPDB), to identify novel multicancer patterns among the families of azoospermic and oligozoospermic men (Hurdle et al., 2013). Describing coaggregation of cancer risk patterns in families of subfertile men is an essential first step to finding meaningful determinants of fertility and cancer risk through improved understanding of shared genetic, epigenetic, and environmental risk factors.

Materials and methods

All study procedures and materials were approved by the University of Utah Institutional Review Board (IRB # 00082636).

Data sources

Subfertile men who underwent a fertility evaluation were identified using the Subfertility Health and Assisted Reproduction and the Environment (SHARE) cohort. All semen analysis (SA) results from 1996 to 2017 for the two major healthcare systems in Utah: University of Utah Health (UUH) and Intermountain Healthcare (IH) were included in SHARE. IH and UUH account for ∼80% of medical encounters in the state and maintain data warehouses that record diagnoses and clinical histories for all patients. SHARE was linked with UPDB allowing integration of demographic, genealogic, and health information and to allow for longitudinal follow-up. UPDB is a powerful statewide population registry that links demographic, residential, clinical, and vital status information across several data sources including birth and death certificates, hospitalizations, ambulatory surgeries, and driver licenses (Hurdle et al., 2013; Utah Population Database, 2021). UPDB is also record linked with healthcare encounter data from IH and UUH and to the Utah Cancer Registry (UCR). UCR is a population-based Surveillance, Epidemiology, and End Results (SEER) program cancer registry and has maintained statewide cancer records since 1966.

Semen analyses

All SA were performed in a central reference laboratory on samples obtained after 2–7 days of sexual abstinence, followed the 4th (1999) or 5th (2010) edition of the WHO manual for examination and processing of human semen depending on the time of sample collection and were in accordance with the checklist published by Björndahl et al. (2016). Men with the most extreme presentations of subfertility were selected for inclusion in the analysis: those with a sperm concentration of 0 million/ml were defined as azoospermic and those with <1.5 million/ml as severely oligozoospermic. For men with more than one SA, average parameters were used.

Study population

The study population was limited to subfertile men, those classified as azoo- or severely oligozoospermic, in SHARE. Fertile controls from UPDB were age-matched 5:1 to subfertile men. Controls were required to have at least one child to ensure they were fertile. For the familial analysis, first- through third-degree relatives were identified using UPDB pedigree data for both subfertile men and their controls. Adopted children/family members were not included in pedigrees. Families with <10 informative members were excluded from the analysis. Individuals were considered noninformative if they had <1 year of follow-up 1966–2017, the years UCR diagnoses were available for the cohort. Families without any informative second- or third-degree relatives were also excluded. Cancer diagnoses were obtained through UCR for all included individuals and grouped by SEER site recode into 34 cancer types (Supplementary Table S1).

Familial multicancer clusters

Within each family, familial risk for each cancer type was measured using standardized incidence risk ratios (SIRs), accounting for sex, age, birth cohort, and person-years of follow-up for the family members. Person-time was calculated as time from the first year residing in Utah or 1966, until: year of first cancer diagnosis, last year residing in Utah, year of death, or 2017. SIRs were placed into a cancer risk matrix and log-transformed prior to calculation of the distance matrix. A maximum value was imposed to improve robustness of results and avoid bias due to large SIRs, particularly for rare cancers, such that any SIR value above the 95th percentile was set to the 95th percentile. Cancer types were also inverse frequency weighted to compensate for rare phenotypes which can drop out if equal weights are used. A significance risk indicator matrix (SRM) was also generated using SIR P-values, families were considered to have ‘high-risk’ (SRM = 1) for a cancer if the SIR was greater than expected (P < 0.05), ‘low-risk’ (SRM = −1) if the SIR was less than expected (P < 0.05), and ‘population-risk’ (SRM = 0) otherwise. Phenotypes were inverse frequency weighted to compensate for rare phenotypes which can drop out if equal weights are used.

Use of Gower distance allows for simultaneous use of the SRM and SIR matrices to measure similarities between the familial multicancer patterns of families. Multicancer clusters were then identified from the weighted Gower distance using partitioning around medoids (PAM or k-medoids). A heuristic approach was used to determine the number of clusters, k. A series of iterative models for k = 2 through 20 clusters were considered. To select k, bootstrapped consensus clustering with 1000 random draws was used to determine cluster stability and silhouette plots were used identify the point of diminishing improvement.

Each cluster represents a familial multicancer pattern (Hanson et al., 2020a,b). The risk for each cancer type was estimated using Cox proportional hazard models. Hazard ratios (HRs) and 95% CIs were used to estimate cluster-specific differences in cancer incidence and to describe and compare clusters. HRs were also estimated for each cancer type for all azoospermia families relative to controls and for all severe oligozoospermia families relative to controls. All models controlled for birth year and sex. All analyses were carried out using the famcluster package in R (Hanson and Madsen, 2023).

Results

Cohort description

A total of 426 case families and 3105 control families were identified for the azoospermia cohort. For the severe oligozoospermia cohort, 360 case families and 2569 control families were identified (Fig. 1 and Table 1). Families of azoospermic men ranged in size from 10 to 170 members and were significantly smaller than families of controls (range: 10–723), with an average of five fewer family members out to third-degree relatives. On average, families of azoospermic men also had 1–2 fewer first- and second-degree relatives. There was no significant difference in number of third-degree relatives. Families of severely oligozoospermic men also ranged in size from 10 to 170 members and had one less first-degree relative on average (P < 0.001). There were no significant differences in overall family size or number of second- and third-degree relatives compared to families of controls (Table 1). Members of subfertile families tended to be slightly older at end of follow-up than members of control families and were also more likely to be deceased at the end of follow-up than relatives of controls (Table 1).

Figure 1.

Cohort selection diagrams. (A) Azoospermia cohort selection. (B) Severe oligozoospermia cohort selection.

Table 1.

Summary of family and individual demographic characteristics for azoospermia and severe oligozoospermia cohorts.

	Subfertile families		Fertile control families
	Mean (SD)	Range	Mean (SD)	Range	P-valuea
Azoospermia cohort	N = 426		N = 3105
Family size	51 (25.8)	10–170	55 (32.8)	10–723	0.004
First degree relatives	6 (2.6)	1–16	8 (2.9)	2–33	<0.001
Second degree relatives	14 (7.3)	2–48	15 (9.5)	1–158	0.008
Third degree relatives	29 (19.1)	0–134	31 (23.4)	0–531	0.077
Severe oligozoospermia cohort	N = 360		N = 2569
Family size	52 (26.9)	10–170	56 (34.5)	10–869	0.080
First degree relatives	7 (2.9)	1–21	8 (2.8)	2–43	<0.001
Second degree relatives	15 (8.3)	1–59	15 (9.9)	0–231	0.487
Third degree relatives	30 (19.8)	0–134	32 (24.7)	0–594	0.135

	Members of subfertile families		Members of fertile control families
	N	%	N	%	P-value
Azoospermia cohort	N = 21 361		N = 161 575
Sex					0.190
Female	10 371	48.6	79 220	49.0
Male	10 988	51.4	82 345	51.0
Missing/unknown	2	0.0	9	0.0
Age at end of follow-up					<0.001
0–9 years	2061	9.6	18 504	11.5
10–19 years	1639	7.7	14 687	9.1
20–29 years	2398	11.2	18 911	11.7
30–39 years	2931	13.7	20 462	12.7
40–49 years	2139	10.0	15 609	9.7
50–59 years	1739	8.1	13 473	8.3
60–64 years	1037	4.9	7349	4.5
65+ years	6767	31.7	47 064	29.1
Missing/unknown	650	3.0	5515	3.4
Vital status at end of follow-up					<0.001
Living	15 402	72.1	120 116	74.3
Deceased	5959	27.9	41 458	25.7
Severe oligozoospermia cohort	N = 18 818		N = 136 000
Sex					0.623
Female	9192	48.8	66 695	49.0
Male	9624	51.1	69 298	51.0
Missing/unknown	2	0.0	7	0.0
Age at end of follow-up					<0.001
0–9 years	2079	11.0	15 600	11.5
10–19 years	1512	8.0	12 253	9.0
20–29 years	2191	11.6	15 888	11.7
30–39 years	2506	13.3	16 893	12.4
40–49 years	1737	9.2	13 191	9.7
50–59 years	1462	7.8	11 603	8.5
60–64 years	849	4.5	6332	4.7
65+ years	5829	31.0	39 790	29.3
Missing/unknown	653	3.5	4450	3.3
Vital status at end of follow-up					0.002
Living	13 722	72.9	100 626	74.0
Deceased	5096	27.1	35 374	26.0

^aχ² or t-test P-value: values <0.05 are bolded.

Cancer types

For family members of azoospermic men with a cancer diagnosis, median age at first cancer was 67 years (range: 1–90). Family members of azoospermic men tended to be slightly younger at first diagnosis (median age_control = 68; P = 0.009) and showed significantly more first diagnoses among adolescents and young adults (AYAs), those diagnosed 15–39 years, than family members of controls (P < 0.001). Among family members of severely oligozoospermic men with a cancer diagnosis, the median age at first cancer was 68 years (range: 1–90). Although there was no difference in median age at first cancer diagnosis between family members of severely oligozoospermic men and controls (median age_control = 68; P = 0.438), more AYA first diagnoses were observed in the families of severely oligozoospermic men than family members of controls (P = 0.047). A summary of cancer diagnoses is shown in Table 2 with years of cancer diagnoses summarized in Supplementary Table S2.

Table 2.

Cancer diagnoses by phenotype site/site grouping for azoospermia and severe oligozoospermia cohorts.

	Azoospermia cohort		Severe oligozoospermia cohort
	Subfertile family members	Fertile control family members	Subfertile family members	Fertile control family members
	N = 2683	N = 19 146	N = 2296	N = 18 362
Phenotype	N (%)	N (%)	N (%)	N (%)
Digestive system
Esophagus	≤10	97 (0.5)	≤10	93 (0.6)
Stomach	32 (1.2)	311 (1.6)	26 (1.1)	253 (1.6)
Small intestine	14 (0.5)	72 (0.4)	10 (0.4)	76 (0.5)
Colon	270 (10.1)	1958 (10.2)	264 (11.5)	1648 (10.3)
Liver	27 (1.0)	239 (1.2)	26 (1.1)	199 (1.2)
Pancreas	60 (2.2)	436 (2.3)	61 (2.7)	364 (2.3)
Other digestive	≤10	67 (0.3)	≤10	67 (0.4)
Respiratory and oral cavity/pharynx
Extrathoracic	84 (3.1)	646 (3.4)	80 (3.5)	536 (3.3)
Lung	136 (5.1)	1119 (5.8)	106 (4.6)	935 (5.8)
Musculoskeletal
Bone/joint	18 (0.7)	52 (0.3)	13 (0.6)	45 (0.3)
Soft tissue	26 (1.0)	140 (0.7)	23 (1.0)	122 (0.8)
Skin
Melanoma	216 (8.1)	1504 (7.9)	168 (7.3)	1249 (7.8)
Other skin	17 (0.6)	77 (0.4)	≤10	83 (0.5)
Female
Breast	386 (14.4)	2777 (14.5)	351 (15.3)	2368 (14.7)
Cervical	63 (2.3)	472 (2.5)	49 (2.1)	369 (2.3)
Uterine	107 (4.0)	677 (3.5)	84 (3.7)	538 (3.3)
Ovarian	50 (1.9)	319 (1.7)	33 (1.4)	304 (1.9)
Other female genital	18 (0.7)	115 (0.6)	≤10	91 (0.6)
Male
Prostate	521 (19.4)	3662 (19.1)	458 (19.9)	3120 (19.4)
Testis	20 (0.7)	123 (0.6)	36 (1.6)	120 (0.7)
Other male genital	≤10	20 (0.1)	≤10	17 (0.1)
Urinary system
Bladder	95 (3.5)	789 (4.1)	75 (3.3)	640 (4.0)
Renal	64 (2.4)	441 (2.3)	46 (2.0)	339 (2.1)
Lymphoma
Hodgkin lymphoma	29 (1.1)	156 (0.8)	22 (1.0)	128 (0.8)
Non-Hodgkin lymphoma	125 (4.7)	823 (4.3)	91 (4.0)	663 (4.1)
Myeloma	37 (1.4)	295 (1.5)	38 (1.7)	240 (1.5)
Leukemia
Acute lymphocytic leukemia	12 (0.4)	67 (0.3)	≤10	66 (0.4)
Myeloid/monocytic leukemia	35 (1.3)	295 (1.5)	28 (1.2)	216 (1.3)
Other leukemia	37 (1.4)	320 (1.7)	40 (1.7)	259 (1.6)
Endocrine
Thyroid	80 (3.0)	396 (2.1)	55 (2.4)	362 (2.3)
Other endocrine	15 (0.6)	79 (0.4)	≤10	71 (0.4)
Nervous system
Other central nervous system	20 (0.7)	225 (1.2)	19 (0.8)	183 (1.1)
Eye	≤10	63 (0.3)	11 (0.5)	46 (0.3)
Brain	42 (1.6)	314 (1.6)	35 (1.5)	256 (1.6)

N is number of diagnoses; individuals can be diagnosed with multiple phenotypes.

Cells with 10 or fewer cases are suppressed.

Overall cancer risk

Compared to control families, significant increases in cancer risks were observed for both the azoospermia and severe oligozoospermia cohorts (Fig. 2). For all azoospermia families compared to control families, increased risk was seen for five cancer types: bone and joint cancers HR = 2.56 (95% CI  =  1.48–4.42), soft tissue cancers HR = 1.56 (95% CI = 1.01–2.39), uterine cancers HR = 1.27 (95% CI = 1.03–1.56), Hodgkin lymphomas HR = 1.60 (95% CI = 1.07–2.39), and thyroid cancer HR = 1.54 (95% CI = 1.21–1.97). Among severe oligozoospermia families, increased risk was seen for three cancer types: colon cancer HR = 1.16 (95% CI = 1.01–1.32), bone and joint cancers HR = 2.43 (95% CI = 1.30–4.54), and testis cancer HR = 2.34 (95% CI = 1.60–3.42). A significant decrease in risk for esophageal cancer HR = 0.39 (95% CI = 0.16–0.97) was observed.

Figure 2.

Overall cancer risk for families of subfertile men relative to families of fertile controls. (A) Difference in risk for all azoospermia families combined relative to fertile control families. (B) Difference in risk for all severe oligozoospermia families combined relative to fertile control families. Points correspond to the hazard ratio (HR) for each cancer phenotype, bars show the 95% confidence interval, color represents the magnitude of the HR. Estimates are plotted on a log₁₀ scale. ALL: Acute Lymphocytic Leukemia, NHL: Non-Hodgkin Lymphoma. *P < 0.05.

Familial multicancer clusters

Thirteen multicancer clusters with distinct familial cancer risk patterns were identified in the azoospermia cohort. The number of families included in each cluster ranged from 6 to 281 with 0–7 cancer types identified at elevated rates (Table 3). A majority (66%) of families in the azoospermia cohort showed population-level cancer risks (Cluster 1) with no significantly elevated cancer types. However, the remaining 12 clusters showed elevated risk for at least two cancer types (Table 3). Patterns of familial multicancer risk for these 12 clusters are shown in Fig. 3. Relative to the control families, several of the clusters with elevated cancer risks showed increased odds of cancer diagnoses at young ages. Six clusters showed a roughly 2- to 3-fold increase in odds of AYA diagnosis with two clusters also showing 3- to 12-fold increase in odds of pediatric cancer diagnosis (Table 3).

Table 3.

Summary of familial multicancer risk patterns for azoospermia and severe oligozoospermia cohorts.

Familial multicancer pattern	Families	Individuals	Pediatric diagnosisa		AYA diagnosisb		Significant phenotypes
	N (%)	N (%)	N	OR (95% CI)	N	OR (95% CI)	N
Azoospermia cohort
1	281 (66.0)	12 707 (60.9)	9	0.63 (0.32–1.23)	99	1.00 (0.81–1.23)	0
2	22 (5.2)	1138 (5.4)	1	0.61 (0.85–4.38)	24	2.30 (1.48–3.57)	2	Breast, hodgkin lymphoma
3	16 (3.8)	1057 (5.1)	5	3.64 (1.47–8.99)	25	2.88 (1.86–4.47)	7	Small intestine, bone/joint, melanoma, other skin, breast, other leukemia, thyroid
4	16 (3.8)	727 (3.5)	1	1.20 (0.17–8.70)	8	1.42 (0.68–2.95)	2	Soft tissue, ovarian
5	15 (3.5)	760 (3.6)	1	1.05 (0.14–7.52)	13	2.09 (1.16–3.78)	3	Cervical, other female genital, renal
6	14 (3.3)	944 (4.5)	0	–	17	1.96 (1.17–3.28)	3	Prostate, thyroid, other endocrine
7	13 (3.1)	566 (2.7)	0	–	13	2.24 (1.24–4.05)	2	Testis, non-Hodgkin lymphoma
8	10 (2.3)	600 (2.9)	0	–	9	1.80 (0.89–3.62)	2	Soft tissue, other skin
9	10 (2.3)	593 (2.8)	0	–	8	1.38 (0.66–2.86)	3	Stomach, small intestine, thyroid
10	10 (2.3)	537 (2.6)	7	12.63 (5.66–28.18)	9	2.26 (1.11–4.61)	2	Extrathoracic, acute lymphocytic leukemia
11	7 (1.6)	399 (1.9)	1	2.00 (0.27–14.55)	2	0.55 (0.13–2.25)	2	Other digestive, thyroid
12	6 (1.4)	461 (2.2)	1	1.56 (0.22–11.35)	4	0.88 (0.32–2.43)	2	Uterine, eye
13	6 (1.4)	393 (1.9)	1	2.00 (0.27–14.55)	1	0.27 (0.04–1.94)	2	Other male genital, thyroid
Severe oligozoospermia cohort
1	224 (62.4)	10 395 (56.1)	9	0.80 (0.41–1.57)	75	0.89 (0.70–1.13)	2	Colon, liverc
2	27 (7.5)	1348 (7.3)	1	0.53 (0.07–3.82)	34	2.78 (1.91–4.05)	1	Testis
3	18 (5.0)	999 (5.4)	0	–	7	0.66 (0.31–1.42)	2	Colon, Liver
4	17 (4.7)	956 (5.2)	1	0.87 (0.12–6.24)	12	1.45 (0.79–2.64)	3	Soft Tissue, Melanoma, Testis
5	16 (4.5)	847 (4.6)	0	–	16	2.37 (1.39–4.07)	2	Hodgkin Lymphoma, Myeloma
6	12 (3.3)	797 (4.3)	4	3.84 (1.40–10.57)	11	1.43 (0.76–2.67)	2	Bone/joint, breast
7	10 (2.8)	580 (3.1)	2	3.40 (0.82–14.06)	9	2.19 (1.08–4.47)	2	Myeloma, other endocrine
8	9 (2.5)	638 (3.4)	0	–	8	1.28 (0.62–2.66)	3	Stomach, small intestine, non-hodgkin lymphoma
9	9 (2.5)	574 (3.1)	1	1.80 (0.25–13.06)	6	1.49 (0.64–3.49)	3	Other female genital, testis, eye
10	6 (1.7)	666 (3.6)	5	9.32 (3.67–23.67)	6	1.43 (0.61–3.35)	2	Colon, acute lymphocytic leukemia
11	6 (1.7)	319 (1.7)	0	–	2	0.64 (0.15–2.66)	1	Other skin
12	5 (1.4)	399 (2.2)	0	–	3	1.01 (0.31–3.30)	1	Other male genital

^aOdds ratio (OR) and 95% CI for cancer diagnosis in pediatric age range (<15 years) relative to control families—cells are left blank when no pediatric diagnoses occurred in familial multicancer pattern.

^bOR and 95% CI for cancer diagnosis in adolescent and young adult (AYA) age range (15–39 years) relative to control families.

^cSignificantly decreased familial risk for liver cancers.

ORs with P < 0.05 bolded.

Figure 3.

Patterns of familial cancer risk across the 12 high-risk multicancer clusters for families of azoospermic men. Cluster 1 is suppressed as all cancers showed population-level risk. Points correspond to the hazard ratio (HR) for each cancer phenotype, bars show the 95% confidence interval, color represents the magnitude of the HR. Estimates are plotted on a log₁₀ scale. Phenotypes were left blank when the N was insufficient to estimate risk. ALL: Acute Lymphocytic Leukemia, NHL: Non-Hodgkin Lymphoma. *P < 0.05.

Within the severe oligozoospermia cohort, twelve distinct familial multicancer clusters were identified with 5–224 families included in each cluster. All 12 clusters showed elevated risk for at least one cancer type, with a maximum of three cancer types at elevated risk observed (Table 3). Cluster 1 harbored a significantly decreased risk for liver cancer and an elevated risk for the colon cancer. Patterns of familial multicancer risk for the 12 clusters are shown in Fig. 4. An increase in odds of cancer diagnoses at young ages was also seen in five of the familial multicancer clusters relative to the control population. Three clusters showed a 2- to 3-fold increase in odds of AYA diagnosis and an additional two clusters showed a 3- to 9-fold increase in odds of a pediatric diagnosis (Table 3).

Figure 4.

Patterns of familial cancer risk across the 12 high-risk multicancer clusters for families of severely oligozoospermic men. Points correspond to the hazard ratio (HR) for each cancer phenotype, bars show the 95% confidence interval, and color represents the magnitude of the HR. Estimates are plotted on a log₁₀ scale. Phenotypes were left blank when the N was insufficient to estimate risk. ALL: Acute Lymphocytic Leukemia, NHL: Non-Hodgkin Lymphoma. *P < 0.05.

Discussion

This analysis leveraged a unique population-based resource to describe familial multicancer patterns for families of subfertile men. Distinct patterns of cancer were observed in the families of azoospermia and severe oligozoospermia cohorts. These results suggest heterogeneity in cancer risk by type of subfertility and within subfertility type and has important implications for future studies to discover determinants of fertility and cancer risk. Categorization of families into clusters likely to be more homogeneous for cancer risk could be an important step toward discovery of shared genetic, epigenetic, and environmental factors and improved risk assessment and patient counseling. Future analyses could also assess semen quality within fertility subtypes as this could play a role in explaining heterogeneity. A majority of families in the azoospermia cohort showed a population-level cancer risk pattern with no significantly elevated cancer types. However, the remaining multicancer clusters in the azoospermia cohort, as well as all the observed clusters in the severe oligozoospermia cohort, showed elevated risk for at least one cancer type.

Several of the cancer types identified with elevated risk in this analysis are consistent with cancer types previously shown to coaggregate with subfertility at the individual and/or familial level, including testis, prostate, thyroid, ALL, and lymphoma (Choy and Eisenberg, 2018; Hanson et al., 2018). However, this analysis showed high variability in risk by subfertility type and between familial multicancer clusters within subfertility type. For example, a 2-fold increase in testis cancer risk was observed for severe oligozoospermia families compared to controls. However, elevated risk for testis cancer was only seen in one third of the oligozoospermia multicancer clusters and risk magnitude within these familial multicancer clusters ranged from a 4- to 24-fold increase. No increased testis cancer risk was observed in the overall cancer risk assessment for azoospermia families relative to fertile controls, although one familial multicancer cluster showed increased risk. However, families of azoospermic men did show an overall increase in risk for two previously reported familial cancers: lymphoma and thyroid. Five of the 12 clusters with elevated cancer risks showed increased thyroid cancer risk (Anderson et al., 2016, 2017; Hanson et al., 2018).

Heterogeneity across familial multicancer clusters could contribute to inconsistent associations observed by different studies of subfertile men and their families for some cancers, including prostate cancer, colon cancer, and melanoma (Choy and Eisenberg, 2018; Hanson et al., 2018). Within the azoospermia cohort here, elevated familial prostate cancer risk was observed in only one of the multicancer clusters and increased melanoma risk in another. Similarly, within the severe oligozoospermia cohort, melanoma risk was elevated in a single multicancer cluster, while colon cancer risk was elevated overall and in three familial multicancer clusters. Based on the observed variability in the magnitude of risk for individual cancer types across familial multicancer clusters, it is expected that overall cancer risk patterns could vary greatly depending on the included families.

In addition to cancer types known to associate with subfertility, here we identified several novel coaggregations. Myeloma, breast, skin, small intestine, and soft tissue cancers were observed in several familial multicancer clusters which have not previously been associated with male subfertility. Several of the clusters with elevated cancer risks also showed significantly increased odds of diagnosis at younger ages, particularly for cancers typically seen in AYAs, including breast, thyroid, melanoma, testis, cervical, sarcoma, and lymphoma. AYA cancers have been shown to have distinct intrinsic and extrinsic risk factors, tumor biologies, and prognosis, with some studies showing unique mutational and genomic profiles relative to older adults (Bleyer et al., 2008; Tricoli et al., 2016, 2018; Wang et al., 2022). Additional study of why these cancers appear to coaggregate in the families of subfertile men could lead to improved understanding of AYA cancer and infertility etiology.

Cancer and subfertility are both complex diseases with heterogeneous presentation and etiology. Identifying novel pleiotropies could provide avenues to improve etiologic understanding and uncover shared genetic, epigenetic, and environmental risk factors. There are some commonalities in the genetic mechanisms underlying subfertility and cancer: microsatellite instability, dysfunction in mismatch repair genes, recombination abnormalities leading to aneuploidy, and germline mutations (Gonsalves et al., 2004; Bonadona et al., 2011; Hanahan and Weinberg, 2011; Ji et al., 2012). Commonalities exist in environmental risk factors as well. Exposure to endocrine disrupting compounds has been associated with both subfertility and risk for some cancers including breast, prostate, vaginal, and thyroid cancers (Burki, 2019; Sharma et al., 2020; Rocha et al., 2021; Rodprasert et al., 2021; Ramsay et al., 2023). The familial multicancer clusters described in this study could represent shared genetics, environment, or, most likely, a combination of both. Future focused analyses of these possibly more homogeneous families, with particular subfertility/cancer patterns, has the potential to improve power to discover and define risk determinants and thereafter develop screening tools to better determine individual and familial risk.

Although this study has many strengths, including population data for family structure, cancer diagnoses and subfertility, there are limitations. First, semen measures are not available for the sample of fertile men. Second, there is no information on medical comorbidities or lifestyle risk factors such as smoking status, BMI, or environmental exposures. Third, all of the subfertile men included in this study were seen at a fertility clinic for evaluation. These men were therefore a subset of the overall population experiencing fertility problems and likely represent those with the socioeconomic means for evaluation by a physician. The families studied are those with events in Utah, which has less racial and ethnic diversity than other parts of the country. Additionally, several of the included phenotypes show wide confidence intervals as they are rare and, by definition, occur infrequently in the population. Finally, men were not categorized by subfertility etiology, such as those with Klinefelter syndrome, which could also be linked with cancer risk. However, this is a rare condition and more likely to impact individual cancer risk than familial risk patterns.

Conclusions

Leveraging a combination of unique population-level data resources, this study described novel multicancer patterns observed in the families of azoospermic and severely oligozoospermic men. Distinct overall multicancer risk and familial multicancer patterns were observed in the azoospermia and severe oligozoospermia cohorts, suggesting heterogeneity in cancer risk by type of subfertility and within subfertility type. Describing families with similar cancer risk patterns provides a new avenue to increase homogeneity for focused gene discovery and environmental risk factor studies. Such discoveries will lead to more accurate risk predictions and improved counseling for patients and their families.

Data availability

The data underlying this article cannot be shared publicly for the privacy of individuals who participated in the study.

Authors’ roles

J.M.R.: conception, data creation, analysis, writing; M.J.M.: analysis, writing; J.J.H.: conception, data creation, writing; H.A.H.: conception, data creation, analysis, writing; N.J.C.: analysis, writing; B.R.E.: data creation, writing; K.I.A.: data creation, writing; E.F.: writing; J.M.H.: conception, writing.

Funding

This work was funded by GEMS: Genomic approach to connecting Elevated germline Mutation rates with male infertility and Somatic health (Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD): R01 HD106112).

Conflict of interest

The authors have no conflicts of interest relevant to this work.