Sex Disparities in Cancer Mortality and Survival

Sex is known to be an important factor in the pathogenesis, diagnosis, and prognosis of many diseases (1). Cancer is a stark example of such—the risk of malignancy is much higher in males, relative to females, for a majority of cancers at most ages (2). Exposure factors implicated in these sex disparities include hormones (3), body mass index (kg/m²; ref. 4), viral infections (5, 6), carcinogenic susceptibility (7), and health care access and utilization (8).

Less information on potential sex differences in cancer mortality and cancer survival is available. Prior reports suggest that certain cancers are disproportionate by sex in these metrics (9–14), but no study has explicitly and systematically tested for sex differences in cancer mortality and cancer survival. These are important questions because disparities in cancer mortality result from the combined effects of incidence and survival; if sex differences exist in cancer mortality and are the result of sex differences in cancer incidence, and not cancer survival, then such evidence may suggest etiologic clues for future analytic studies. If cancer survival is also highly disproportionate between men and women, then reasoning becomes more complex because, in addition to etiologic factors, this metric may suggest sex differences in the natural history of disease, access to or use of medical care, response to treatment, or some combination of these.

In a previous report, we described sex disparities in cancer incidence rates in the United States using Surveillance Epidemiology and End Results (SEER) in an attempt to bring attention to sex as an important consideration in studies of cancer (2). In this complementary article, we now utilize SEER and National Center for Health Statistics (NCHS) data to assess sex differences in cancer mortality and cancer survival.

The April 2009 submission of the SEER Program (15) mortality database in SEER*Stat v6.6.2 was used to calculate cancer mortality counts and rates per 100,000 person-years (age-adjusted in single years to the 2000 U.S. standard population) for each cancer, stratified by sex for the periods 1977–1986, 1987–1996, 1997–2006, and 1977–2006. The National Vital Statistics System makes this data available through NCHS via SEER*Stat. Investigated cancers were all malignant cancers, all malignant cancers excluding sex-specific (cervix uteri, corpus uteri, uterus NOS, ovary, vagina, vulva, other female genital organs, prostate, testis, penis, other male genital organs), all malignant cancers excluding sex-specific and breast, and individual cancers according to the SEER Cause of Death Recode (16). Male-to-female mortality rate ratios (MRR) were calculated for cancers which had at least 10 deaths in each sex by using the male age-adjusted mortality rate as the numerator and the female age-adjusted mortality rate as the denominator. CIs for the male-to-female MRRs were generated in SEER*Stat (17). Graphs of male-to-female MRRs and sex-specific mortality rates plotted by age at death (10-year age groups) were produced for each cancer. Each data point (age group) of these plots was required to have at least 25 cases for each sex. Cancers that had the most extreme (>10%) and consistent changes in male-to-female MRR over the 3 decades studied and a mortality rate of at least 3 per 100,000 in one of the sexes are illustrated and discussed herein (graphs for other cancers can be accessed online as Supplementary Material).

SEER 17 (18) incidence data were extracted for survival analyses. The geographic area and calendar period covered by each of these registries is available online (19). For each cancer, we extracted the following variables: patient id, sex, primary site, histologic type, tumor stage, tumor grade, age at diagnosis, year of diagnosis, month of diagnosis, survival time, vital status, and cause of death from a case listing session. Data were restricted to individuals with a single primary diagnosis of malignant cancer diagnosed during 1973 to 2006. Cox proportional hazards models were used to estimate the male, relative to female, hazard of cause-specific mortality, defined here as the cause of death being the specific cancer originally diagnosed and death being within 5 years of cancer diagnosis. (All-cause mortality analyses for the 5 years following cancer diagnosis were also conducted, the results of which are provided as Supplementary Material.) All analyses were adjusted for age at diagnosis (10-year age groups to 80+) and stratified (equal coefficients across strata but with baseline hazard unique to each stratum) by year of cancer diagnosis (1973–1979, 1980–1986, 1987–1993, 1994–2000, and 2001–2006). Subsequent analyses adjusted for additional variables of cancer stage and grade were restricted to cancers where such information was available for at least 60% of cases that were included in the age-adjusted model. When models adjusted for age and year of diagnosis were restricted to those with stage and/or stage and grade information, similar estimates were obtained, thus the maximal number of participants were retained in each of the models conducted. Log-log plots of survival against analysis time indicated that the proportional hazards assumption was upheld for each of the cancers assessed (data not shown). Data for the cancer peritoneum, omentum, and mesentery were not amenable to Cox proportional hazards models due to small numbers.

Data were analyzed using STATA version 10.1 (StataCorp LP). Graphs were generated using SigmaPlot version 11.0 (Systat Software, Inc.).

Sex-specific, age-adjusted cancer mortality rates per 100,000 person years and male-to-female MRRs are shown in Table 1 for the periods 1977–1986, 1987–1996, 1997–2006, and 1977–2006. For the vast majority of cancers studied, age-adjusted mortality rates were higher for males than females; over the entire period, the 10 cancers with the highest male-to-female MRR were lip (MRR = 5.51, 95% CI: 5.05–6.03), larynx (MRR = 5.37, 95% CI: 5.29–5.45), hypopharynx (MRR = 4.47, 95% CI: 4.30–4.65), esophagus (MRR = 4.08, 95% CI: 4.05–4.11), urinary bladder (MRR = 3.36, 95% CI: 3.34–3.39), tonsil (MRR = 3.28, 95% CI: 3.17–3.39), oropharynx (MRR = 3.05, 95% CI: 2.95–3.16), floor of mouth (MRR = 2.89, 95% CI: 2.77–3.03), tongue (MRR = 2.53, 95% CI: 2.49–2.58), and nasopharynx (MRR = 2.47, 95% CI: 2.40–2.55). Only 3 cancers had a higher mortality rate in women than men: peritoneum, omentum, and mesentery (MRR = 0.39, 95% CI: 0.38–0.41), gall bladder (MRR = 0.56, 95% CI: 0.55–0.57), and anus, anal canal and anorectum (MRR = 0.78, 95% CI: 0.75–0.81). Male-to-female MRRs have changed over time for certain cancers; for example, lung and bronchus, larynx, and pancreas cancers had relatively high male-to-female MRRs in the 1977 to 1986 period which consistently decreased with time over the 30-year period. In contrast, the male-to-female MRRs consistently increased for several sites between 1977 and 2006, including the esophagus, skin excluding basal and squamous, and liver and intrahepatic bile duct.

Next, we examined whether sex-specific mortality rates and male-to-female MRRs changed with age from 1977 to 2006 (Fig. 1A–E; Supplementary Figs. S1–35, can be accessed online). Several patterns emerged. For example, laryngeal cancer mortality (Fig. 1A) has been decreasing in both sexes for the majority of ages, but this trend has been more rapid in males. In addition, female laryngeal cancer mortality has been increasing in older age groups (70+ years). Both of these factors have contributed to the decline in male-to-female MRR of laryngeal cancer over the 30-year time period of analysis among those 60+ years of age. Esophageal cancer (Fig. 1B) has been trending in the opposite direction with an increasing MRR being observed, particularly between the ages of 40 to 69 years. This has been caused by a decrease in the female mortality rate, whereas the equivalent rates for males have remained fairly stable. In the oldest age groups (70+ years), esophageal cancer mortality rates have been increasing in both sexes, but at a faster rate in males, resulting in increased MRRs for these age groups. Lung and bronchus cancer (Fig. 1C) has been decreasing in males but has remained fairly stable and has even increased at older ages, in females during 1977 to 2006. This has caused the male-to-female MRR to dramatically decrease for all adult age groups. Mortality from skin cancer (Fig. 1D) has also been more incident in males than in females. Moreover, the MRR for these cancers has been increasing in the age groups 50+ years due to increases in male mortality yet stable female mortality. Mortality rates of liver and intrahepatic bile duct (Fig. 1E) follow a similar trend, although the observed increases have been substantially greater for males, relative to females, between the ages 30 and 59, which has resulted in a large increase in the MRR during the observed 30-year period.

Next, we examined cancer survival. Cox proportional hazards models, adjusted for age, stage, and grade, were used to test for sex differences in survival in the 5 years following cancer diagnosis (Table 2). A large number of cancers had higher hazards of death for men (i.e., worse survival) than women including, but not limited to, skin excluding basal and squamous (HR = 1.58, 95% CI: 1.52–1.64), endocrine system (HR = 1.32, 95% CI: 1.24–1.42), floor of mouth (HR = 1.32, 95% CI: 1.07–1.63), anus, anal canal, and anorectum (HR = 1.21, 95% CI: 1.02–1.43), lymphoma (HR = 1.20, 95% CI: 1.18–1.22), and lung and bronchus (HR = 1.17, 95% CI: 1.16–1.18). In contrast, 2 sites were notable for their decreased risk of cause-specific mortality in men relative to women: urinary bladder (HR = 0.83, 95% CI: 0.81–0.86) and tongue (HR = 0.89, 95% CI: 0.83–0.95). Adjustment for stage, and grade, when available, had moderate effects for some cancers (e.g., the HR for tongue cancer went from 1.07 to 0.89). The excess male hazards for the cancers salivary gland and skin excluding basal and squamous were attenuated when adjusted for stage. In contrast, the HRs for the 2 cancers ureter and anus, anal canal and anorectum were strengthened after adjustment for both stage and grade. Other cancers, such as urinary bladder had a more complicated pattern: higher risks in women were attenuated yet persisted after adjustment for stage and grade.

In this study, we show that cancer mortality was much higher in males relative to females for a majority of cancer types (Table 1 and Fig. 1). Cancer survival was generally similar between the sexes; even when differences were observed, these sex disparities were relatively modest.

Disparities of cancer mortality have largely paralleled those of cancer incidence, which we have described previously (2). For example, the incidence rate ratio (IRR) and MRRs were largely similar, differing by more than 20% for only 4 cancer sites (lip: MRR = 5.51, IRR = 7.16; salivary gland: MRR = 2.32, IRR = 1.59; skin excluding basal and squamous: MRR = 2.30, IRR = 1.43; and ureter: MRR = 1.86, IRR = 2.45). Qualitatively, the patterns of sex-specific mortality rates and male-to-female MRRs by age and stratified by decade (Fig. 1 and Supplementary Figs.), appeared to be nearly identical to the patterns observed in cancer incidence rates (2).

This supports the idea that sex disparities in cancer mortality arise from the sex differences in cancer incidence. Examples of risk factors that have been implicated in cancer sex disparities include tobacco smoking in lung and bronchus, oral human papilloma virus (HPV) infection in tongue and oropharyngeal (20–22), cosmetic and occupational UV radiation exposure in skin excluding basal and squamous (23, 24), and anal HPV infection in anus, anal canal, and anorectum (6, 25). Universal mechanisms that may be causal in observed sex differences in cancer incidence and, thus, mortality include antioxidative capacity (26, 27), sex chromosome complement/aneuploidy/aberrations (28, 29), gene expression (30–32), hormones (33, 34), and immunocompetence (35). These issues relating to sex disparities in cancer incidence are discussed in further detail in our previous article (2).

For cancer survival, the largest sex differences occurred for the cancers: skin excluding basal and squamous; endocrine system; floor of mouth; retroperitoneum; salivary gland; small intestine; trachea, mediastinum and other respiratory organs; anus, anal, canal and anorectum; lymphoma; nose, nasal cavity, and middle ear; lung and bronchus; urinary bladder; and tongue. For all but the last 2 of these cancers, males had a higher risk of death from cancer. It is feasible that differential environmental exposures and/or physiologic processes, such as sex hormones, could explain the observed sex disparities in survival (2, 36, 37).

Alternatively, observed sex differences in survival may be artifactual. In analyses unadjusted for extent of disease, lead time bias could give the false impression of sex disparities in 5-year survival rates. For cancers with the largest sex differences in survival, the sex with the poorer survival almost always presented with later stage and higher grade tumors. Additional adjustment for stage and grade substantially attenuated most observed sex differences. However, it is unlikely that categorical metrics of stage and grade fully account for variation in extent of disease, thus residual confounding remains a distinct possibility.

Overdiagnosis through screening could also disproportionately affect sex-specific cancer survival. For example, a large proportion of cancer is thought to be asymptomatic and undiagnosed (38). If asymptomatic cancers tended to be diagnosed more frequently in females, relative to males, females could appear to have better cancer survival (39, 40). In support of this hypothesis, females typically present with earlier stage, lower grade, less-aggressive, and unifocal cancers, compared with males (39), perhaps because women more readily and more often utilize health resources available to them (37, 41–44). More research on this subject is required to accurately ascertain to what degree this may play a role in sex discrepant cancer survival.

Sex differences in comorbidity at cancer diagnosis could also skew cancer survival in favor of 1 sex over the other. Some (45–49), but not all (50–52), studies have suggested that males have more comorbid conditions at the point of cancer diagnosis than do females. As comorbidities are independent prognostic indicators, preexisting chronic conditions may contribute to poorer cancer survival.

Of our results for specific cancers, better survival for lung and bronchus among women is noteworthy. This concurs with previous studies which have addressed this question (40, 53), the results of which have piqued the idea that estrogen receptor β, expressed in lung cancer cells, may be causal to this observation (54, 55), though various other hypotheses have also been suggested (56).

Urinary bladder cancer was unusual in that females had lower 5-year survival, compared with males, an observation also made by others (8, 14). Given that U.S. male urinary bladder cancer rates are much higher than female rates (2) and that females often present with later stage and higher grade lesions (8, 15, 57–59), the observed disparity in survival may partly be due to diagnostic misclassification. For example, when presenting with similar symptoms, males may be more readily referred for cystoscopy than women (60).

Cancers of the anus, anal canal, and anorectum show the opposite pattern to urinary bladder; these cancers are more common in females, yet males have lower rates of 5-year survival. One hypothesis is that men may be more likely than women to have anus, anal canal, and anorectum tumors caused by HIV infection (61), and that such tumors may be more aggressive (25).

Cancer of the tongue was unique, in that adjustment for stage and grade changed the HR estimate from indicating a higher risk of death in males to indicating a higher risk of death in females, in the 5 years following diagnosis. These unusual observations are difficult to explain, mainly because the etiopathogenesis of this cancer is poorly understood (62, 63).

Strengths of this study include the use of a large, population-based cancer registry database. In addition, SEER has extensive quality control procedures (64, 65). Limitations of this analysis include use of cause of death extracted from death certificates which is known to have problems and imperfections (66). However, inaccuracies are likely to be nondifferential by sex. Other limitations include lack of information on comorbidities and only having adjusted for age, year of diagnosis, stage and grade, which may be suboptimal for certain cancers. Finally, although we utilized the largest U.S. data set currently available for cancer survival analyses, our results are not perfectly generalizable to the total U.S. population due to the fact that the data are restricted to the 17 cancer registries currently in SEER (18).

In conclusion, this analysis shows that male cancer mortality rates were higher than equivalent female rates for a majority of cancers and these differences largely mirror sex differences in cancer incidence. This analysis also shows modestly, but appreciably, worse survival in men for a number of cancers. Future analytic studies should focus upon the etiologic factors responsible for the systematically higher cancer incidence rates among men.

No potential conflicts of interest were disclosed.

The authors thank Ms. Sabah Quraishi for her help in figure development.

This work was supported by grants from the Intramural Program of the National Cancer Institute, NIH, Department of Health and Human Services.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.