Descriptive epidemiology provides a foundation for many disciplines within medical research. In cancer epidemiology, descriptive analyses of incidence trends often promote the formulation of new ideas and hypotheses. Previous descriptive studies have shown that cancer incidence and mortality rates are much higher in males than females at nearly all ages in the majority of countries (2-10). However, few studies have explicitly set out to focus on sex ratios of cancer incidence, and no study has done this using US data.

Data extraction was restricted to malignant tumors. Male-to-female IRRs were calculated for cancers which had at least 50 cases in each sex by using the male age-adjusted incidence rate as the numerator and the female age-adjusted incidence rate as the denominator. Confidence intervals for the male-to-female IRRs were generated in SEER³. Cancers with consistently increasing or decreasing male-to-female IRRs for the period 1975-2004 and an incidence rate of at least 0.2 per 100,000 for each sex were tabulated and sorted by relative change in male-to-female IRR. Further investigation was considered for cancers which had the highest incidence in males and/or females, had the highest or lowest male-to-female IRRs, or had the most extreme consistent changes in male-to-female IRR over time. Cancers which qualified for one or more of these groups were not further stratified by histologic group if they had low incidence rates (<0.2 per 100,000) for at least one of the sexes, if further stratification was deemed uninformative or if the cancer site recode grouping was non-specific. Selected cancers had their sex-specific incidence rates and male-to-female IRRs tabulated by subsite and/or predominant histologic group(s) for the years 1975-1984, 1985-1994, 1995-2004 and 1975-2004.

Table 2 shows the relative change in male-to-female IRRs and relative change in sex-specific incidence rates for the periods 1985-1994 and 1995-2004 compared with 1975-1984 for cancers showing a consistently increasing or decreasing IRR and incidence rates for each sex of at least 0.2 per 100,000 listed according to percent change in IRR. During 1975-2004, tonsil cancer had the largest increase in male-to-female IRR, and this was the result of an increasing male incidence and decreasing female incidence. Other cancers with rising IRRs were esophagus and mesothelioma, the former due to increases among males not apparent among females, and the latter a female decrease in incidence. Rates for the remaining cancers with rising male-to-female IRRs increased among both sexes, but more so among males than females.

Many cancers had a decline in the male-to-female IRR during 1975-2004. The IRRs for the cancer sites ureter, floor of mouth, retroperitoneum, and lip decreased due to faster declines in male incidence than in female incidence. Reductions in IRR for other cancers - pancreas, nose, nasal cavity and middle ear, gum and other mouth, Hodgkin lymphoma, and larynx - were solely caused by decreased male incidence, with female rates remaining fairly stable. More rapidly rising rates among females than males reduced the IRRs for cancers of the kidney and renal pelvis, cranial nerves and other nervous system, and thyroid. Perhaps the most notable declining IRR was that for lung and bronchus cancer, which is singular in that it was caused by a simultaneous decrease in male rates and increase in female rates.

Of the cancer sites in Table 2, those with the largest increases in male incidence between the periods 1975-1984 and 1995-2004 were skin excluding basal and squamous (111%), liver and intrahepatic bile duct (103%), anus, anal canal and anorectum (58%), and thyroid (44%). For females, the cancer sites were liver and intrahepatic bile duct (88%), thyroid (76%), skin excluding basal and squamous (69%), and lung and bronchus (59%). Over the same time period, the sites with the largest decreases in incidence have been lip (57%), floor of mouth (48%), larynx (30%), and gum and other mouth (30%) for males, and floor of mouth (41%), tonsil (29%), ureter (15%), and gum and other mouth (14%) for females.

Table 3 shows sex-specific incidence rates and male-to-female IRRs for specific histologic groups of selected cancers. The rates and IRRs for nearly all histologic groups follow the same trends as their parent cancer site. However, this is not true for the two histologic types of esophageal cancer, which trended in opposite directions. The IRR, by 10-year period, for all esophageal cancers combined increased from 3.14, to 3.43 to 3.82 (Table 1), however, this is a combination of increasing IRRs for esophageal adenocarcinoma (6.25, 7.96, and 7.69) and decreasing IRRs for esophageal squamous cell carcinoma (2.93, 2.57, and 2.25) (Table 3). Although the IRRs did not change greatly over time, the IRRs for hepatocellular carcinoma ranged 3.44-3.56, considerably larger than those for intrahepatic bile duct carcinoma which ranged 1.51-1.62, not apparent in the overall IRR for liver and intrahepatic bile duct cancer range of 2.51-2.71. In contrast to the overall lung and bronchus cancer IRR of 2.06, that declined from 3.00 to 1.61 over the decades, the overall IRR ranged from 3.50 for squamous cell carcinoma, to 2.19 for large cell, 1.60 for small cell, and 1.53 for adenocarcinoma; IRRs for each type declined from an early maximum of 5.29 for squamous cell to 1.26 for small cell carcinoma recently.

Incidence rates and male-to-female IRRs are plotted by age of diagnosis for nine cancer sites/histologies (Figs 1a-i) selected because they had the most extreme consistent changes in male-to-female IRR and an incidence of at least 0.4 per 100,000 which was increasing in at least one of the sexes (graphs for other cancers, Supplementary Figures 1-30, can be accessed online⁵). During the earliest decade the male-to-female IRR for tonsil cancer was about 2 at ages less than 70 years and rose to about 4 at older ages; by the recent decade, rising male rates and declining female rates resulted in a male-to-female IRR of more than 5 at ages 40-59 years (Fig. 1a). Rates for the cancer site skin excluding basal and squamous (Fig. 1b) predominated among females at younger ages (<40 years) and among males in the older age groups (≥40), and rates rose over time more rapidly among younger women and among older men, resulting in declines in the male-to-female IRR at younger ages and increases at older ages. The incidence of esophageal squamous cell carcinoma among males at all ages and in females less than 70 years of age declined, and the male-to-female IRR decreased at older ages from about 3 to 2 (Fig. 1c). The male-to-female IRR for esophageal adenocarcinoma exceeded 10 at ages 50-59 years and declined with age to 4-5 at the oldest ages (Fig. 1d). Rates for anus, anal canal and anorectal cancers increased comparably in both sexes in nearly all age groups, making the male-to-female IRR fairly stable and less than 1 at ages of more than 50 years (Fig. 1e). The increasing rates of hepatocellular carcinoma in younger age groups (<50 years) has been more rapid among males than females, and the male-to-female IRR rose from less than 3 to more than 5, in contrast to decreases in the male-to-female IRR at older ages (Fig. 1f). Kidney and renal pelvis cancer IRR were about one at very young ages (Wilms tumors), decreased over the decades at ages 10-39, and rose to a fairly constant two at older ages in spite of rising rates (Fig. 1g). Thyroid cancer male-to-female IRR were around 0.2 at ages 20-29 and rose to at least 0.7 at the oldest ages; rates rose over time more rapidly among younger women and older men, leading to progressively lower IRRs (Fig 1h). Male rates for lung and bronchus malignancies declined slightly at virtually all ages while female rates increased, especially in the age groups of 60 years and older (Fig. 1i), and the male-to-female IRR declined dramatically across all adult age groups, dropping below 1 in recent years among those less than 40 years of age.

Tonsil and oropharyngeal cancers increased in male predominance between 1975 and 2004. These two cancer sites have the strongest and most consistent associations with human papillomavirus (HPV) 16 infection of all the oral cancer sites (12). Although tobacco and alcohol are known to be important risk factors for head and neck malignancies (13, 14), historical trends indicate that these exposures are not diverging between the sexes (15)^6,7. This suggests that other risk factors, including oral HPV infection, could be driving the increases in male-to-female IRRs of these sites. Evidence to support this is that increased and earlier sexual activity in males, but perhaps not in females, increases the risk of HPV16-associated oral cancer (16). In addition, two studies have shown increased risks of oral HPV-16/18/33 (17) and head and neck cancer (18) in male partners of females diagnosed with various cervical abnormalities. In total, this evidence may support a speculative hypothesis of differential transmission risks between males and females when engaging in opposite sex, genital-oral sexual practices (17, 19).

At younger ages, skin excluding basal and squamous (Fig. 1b) exhibited a slow but progressive increased female incidence compared with the relatively stable rates of males. Intermittent exposure to high doses of ultra-violet radiation is the main risk factor for this tumor site (20), and the suggestion that females attempt to tan more than males may explain the increased incidence of melanoma in younger females (21). Alternative explanatory factors include hypotheses of a hormonal etiology (22) and an association with increasing body mass index/body surface area (23). In contrast to the female predominance at younger ages, male incidence rates predominated and increased at a greater pace beyond age 40 years, which caused the IRR to inflate. One possible explanation may be the greater lifetime sun exposure in men with outdoor occupations.

The two predominant histologic types of esophageal cancer (Table 3, Figs 1c & 1d) have diverged in both their incidence and male-to-female IRRs. The IRR for esophageal adenocarcinoma increased then stabilized, a pattern unlikely to be accounted for by alcohol and tobacco due to their historical trends and stronger associations with esophageal squamous cell carcinoma relative to esophageal adenocarcinoma (24). Pre-neoplastic states following severe reflux exhibit an increasing male predominance in the step-wise progression to adenocarcinoma (25), which may indicate that the causal exposure(s) of the sex disparity is important at all stages of disease. There is some evidence to suggest that erosive reflux disease may be more prevalent in males at earlier ages compared with females (26, 27). Other risk factors, including estrogen exposure (28, 29), body mass index (30) and Helicobacter pylori infection (31)⁸, have provided no evidence for being the causal agents of the sex ratio imbalance. Putative differences between the sexes that may explain the male predominance of esophageal adenocarcinoma include android obesity (32), production and concentration of gastric acid (33), hiatal hernia (34), defective lower-esophageal sphincter (34), and higher intra-abdominal and intra-gastric pressures (35).

Cancers of the anus, anal canal and anorectum is one of just a few cancers which were more common in females than males at nearly all ages (Fig. 1e). Cloacogenic tumors, a subtype of squamous cell carcinoma which contributes 15% of the malignancies of this site, are more common in females (IRR=0.49, 1975-2004), but this does not fully account for the female predominance of these tumors (IRR=0.77, 1975-2004, for squamous cell carcinoma excluding cloacogenic tumors). A potential reason for the increasing predominance of females may be evolving sexual practices, given the putative association of these cancers with anal HPV infections (36, 37).

Hepatocellular carcinoma (HCC) rates rose among both sexes at older ages, but especially among males younger than 60 years of age (Fig. 1f). Major risk factors for HCC in the United States are chronic infection with hepatitis C virus and hepatitis B virus, as well as alcoholic liver disease (38). Chronic hepatitis B and C infections are more common in males than in females (39) and this difference is likely to account for some of the disparity in IRRs. Historical trends of alcohol consumption ^{9, 10} imply this to be an unlikely cause of the diverging male-to-female IRRs at younger ages. The strongest hypothesis to date, which may account for the sex disparity in incidence, is the anti-carcinogenic effect of estrogen, which constrains transcription of the pro-inflammatory cytokine interleukin-6 gene, leading to a decrease in HCC (40). Increased body mass index (41) and diabetes (42) have also been associated with increased risk of HCC, although it is not clear that either exposure could account for the increased male-to-female IRR at younger ages (43)¹¹.

The male-to-female IRR for kidney and renal pelvis cancer has not significantly changed (Table 1). Moreover, although the incidence in males and females has been increasing, the IRR for adult malignancies has remained stable at around 2 (Fig. 1g). Therefore, it is unlikely that associated exposures of tobacco (44, 45) and obesity (45) are major factors in the male predominance of this cancer site, as these exposures have been changing over time. Reasons for the sex discrepancy of cancer incidence at this site remain unresolved.

Between 1975 and 2004, thyroid cancer incidence was higher in females than males at all ages (Fig. 1h). The etiology of this cancer is poorly understood (46), but increased female incidence during reproductive ages, suggests a possible hormonal pathogenesis (47). Evidence from autopsy studies indicates that the prevalence of thyroid microcarcinoma is equivalent between the sexes while females account for ~83% of clinical cases (48). This may indicate that females have an increased propensity for diagnostic investigation given their higher rates of thyroid disease, relative to males.

The male-to-female IRRs for lung and bronchus cancer reflect historical exposure to tobacco smoking; as smoking habits have converged between the sexes (15), so has the incidence (Fig. 1i). This pattern is also evident for other cancers for which tobacco smoking is a risk factor, including lip (Supp. Fig. 1) and, at older ages, esophageal squamous cell carcinoma (Fig. 1c), floor of mouth (Supp. Fig. 4) and larynx (Supp. Fig. 16). The dramatic decrease in the IRR in later time periods for cancers of the lung and bronchus exemplifies the effect that a single sex-discrepant exposure can have on cancer incidence.

The vast majority of cancers included in this review have higher incidence rates in males than in females, as accentuated in Figure 2, and this suggests the possibility of universal mechanisms that increase male susceptibility to cancer. Tentative explanatory hypotheses include sex differences in anti-oxidative capacity (49), metal toxicity (50), beliefs and behaviors (51), health care access and utilization (43), sex chromosome complement/aneuploidy/aberrations (52), gene expression (53), hormones (54), immunocompetence (55) and the disposable soma theory of aging (56).

The main strength of male-to-female IRRs, relative to absolute differences in incidence rates, is that they are less likely to be affected by changes in diagnostic techniques, preventative strategies, tumor definitions and coding practices (57); changes in such parameters may not be expected to affect the sexes disproportionately. Male-to-female IRRs, however, are still subject to the effects of sex differences in reporting behavior, illness behavior, health care access and utilization, and physician behavior (9, 58). In-spite of these unavoidable caveats, the use of a sex-ratio statistic is still advantageous and superior in representing sexual dimorphism in the pathogenesis of cancer. The main limitation of male-to-female IRRs is their dimensionless scale; thus, sex-specific incidence rates have been considered in concert with this measure to provide the reader with both relative and absolute scales. Although it may also be of interest to assess birth cohort effects of male-to-female IRRs, this was not a primary aim of the analyses described herein.

Legend for Supplementary Figures

Supplementary Figures