Incidence of Ovarian, Peritoneal, and Fallopian Tube Carcinomas in the United States, 1995-2004

Fallopian tube and primary peritoneal cancers are uncommon malignancies with many similarities to epithelial ovarian carcinoma, including histologic, clinical, and etiologic factors (1). Similar to the ovary, the fallopian tube is derived from the Mullerian ducts (2). The epithelial layer of the ovary and the peritoneum also shares a common embryonal heritage, deriving from coelomic epithelium early in life (3). More than 90% of fallopian tube cancers are serous adenocarcinomas, which are histologically indistinguishable from papillary serous ovarian carcinoma (4, 5), so they must be defined morphologically by the invasion of the peritoneal surfaces with minimal or no involvement of the ovaries (6). Clinically, women diagnosed with fallopian tube cancer and primary peritoneal cancer are treated using the same surgical and chemotherapeutic approach as epithelial ovarian cancer because of the similarities in biological behavior (6, 7).

Both fallopian tube and peritoneal cancers are rare, so etiologic studies are difficult to conduct, requiring large collaborations among research centers. The biological and pathologic similarities between fallopian tube, peritoneal, and ovarian carcinomas support the likelihood of a common molecular pathogenesis (1). The absence of descriptive and analytic studies of these malignancies underscores the need for a comprehensive comparative analysis of incidence that might provide leads for further research. As in past descriptions of ovarian cancer incidence patterns (8), the North American Association of Central Cancer Registries (NAACCR) analytic files provide an extraordinary opportunity to conduct this comparative analysis in a large and well-defined cancer registry network. In the present analysis, we examine ethnic and racial variation in primary carcinomas of the ovary, peritoneum, and fallopian tube by age at diagnosis, and pathologic and clinical features, such as histology and stage.

Cancer registries in the United States that participate in the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) program or the Centers for Disease Control and Prevention's National Program of Cancer Registries (NPCR) contribute data annually to NAACCR. Data from each contributing registry are required to pass rigid quality criteria for completeness of reporting, nonduplication of records, internal consistency among data items, and low percentage of death certificate only cases (9). NAACCR provided two data sets that were used in this analysis. The first file, Cancer Incidence in North America Deluxe, was used in the analysis of race (White, Black, American Indian-Alaska Native, Asian-Pacific Islander; ref. 9). The second file, Cancer in U.S. Hispanics/Latinos, was used in the analysis of Hispanic ethnicity (10). Only registries with data available for all 10 y from 1995 to 2004 were included in this analysis. Qualifying states that participated through written consent were Arizona, California, Colorado, Connecticut, Florida, Hawaii, Idaho, Illinois, Iowa, Kentucky, Louisiana, Maine, Michigan, Nebraska, New Jersey, New Mexico, New York, Pennsylvania, Rhode Island, Texas, Utah, Washington, Wisconsin, and Wyoming. These states represented 64% of the U.S. population during the 10-y time period, including 62% of the White population, 54% of the Black population, 81% of the Asian-Pacific Islander population, 59% of the Alaska Native-American Indian population, and 88% of the U.S. Hispanic population. Although there were too few American Indian-Alaska Native cases (294 ovary, 13 peritoneum, and 9 fallopian tube) to do some of the subgroup analyses, they were included as part of the “total.”

Population counts, used as denominators for calculating the cancer rates included in this analysis, were derived from the 2000 U.S. Census (11). The Census Bureau provided estimates of the resident populations of the U.S. counties by 5-y age groups and newborns (0, 1-4, 5-9, …, 80-84, and 85+ y), race (White, Black, American Indian-Alaska Native, Asian-Pacific Islander), and Hispanic origin (Hispanic, non-Hispanic).

All women with invasive cancer meeting the NAACCR standards with International Classification of Diseases for Oncology, Third Edition (ICD-O-3) topography codes C48.1-8 (peritoneum), C56.9 (ovary), and C57.0 (fallopian tube) were included in the analytic file (12). We restricted this comparative analysis to microscopically confirmed cases with carcinoma. Carcinomas comprised 91.2% of ovarian malignancies, 77.0% of peritoneal malignancies, and 93.7% of fallopian tube malignancies. We excluded borderline (“low malignant potential”) tumors of the ovary because the behavior code for these tumors changed after diagnosis year 2000 from “malignant” under ICD-O-2 to “benign” under ICD-O-3. Data were available from 112,541 women with primary ovarian carcinoma, 6,458 women with primary peritoneal carcinoma, and 3,479 women with primary fallopian tube carcinoma.

A summary stage variable, developed by the SEER Program, was used to group cases for comparative analysis of stage of disease at diagnosis (13, 14). Cases diagnosed from 1995 to 2000 were coded according to guidelines established in 1977 (SSS1977; ref. 13), whereas cases diagnosed between 2001 and 2003 were coded according to revised guidelines established in 2000 (SSS2000; ref. 14). As a result of the changes made to staging schema, direct conversion between SSS1977 and SSS2000 was not possible for ovary and fallopian tube cases (15). The major shift in coding was for ovary and fallopian tube cases in distant stage according to SSS1977 moving to a regional stage in SSS2000. Accordingly, we grouped cases into one of three categories: localized, regional and distant, or unknown. Localized tumors were those confined entirely to the organ of origin; regional tumors and distant tumors were those that extended into surrounding organs and tissues, or had spread to remote organs or lymph nodes. Because another revision of the staging system was implemented in 2004, cases diagnosed in 2004 were not included in the analyses involving stage.

SEER*Stat (version 6.4.4) was used to calculate both annual and average annual age-adjusted incidence rates (AAIR) expressed per million population (16). All rates were age adjusted by 5-y age groups to compare racial and ethnic groups by demographic, pathologic, and clinical variables of interest. Confidence intervals (CI) for the rates were calculated by the method of Fey and Feuer (17). The 2000 U.S. population was used for age standardization. For all analyses, rates were suppressed when the category had fewer than 20 cases. Incidence rate ratios (IRR), 95% CI, and significance tests for the IRRs were calculated to assess patterns in cancer incidence (18, 19). To control for possible confounding of the difference in the age distribution among specific racial and ethnic groups on the comparison of, for example, stage or histology, a relative percentage of the column (C%) AAIR was calculated instead of a percentage based on case counts. This quantity was calculated as the AAIR for a particular category divided by the AAIR for the total group in the column. For example, the relative percentage of localized cancers for Whites was the age-adjusted localized rate divided by the overall AAIR for Whites. Additionally, to control for the differences in race distributions among geographic regions on the comparison of stage, the relative column percentages were race standardized using the 2000 U.S. population as the standard. We examined differences in the log age-specific incidence rates separately by race and Hispanic ethnicity using the Poisson model (20). To assess a trend in incidence rates, the annual percent change in rates between 1973 and 2005 was calculated. The statistical comparison of trends was based on the t test of the trend regression variable. Three-year moving averages were used in the trend plots to reduce the effect of random variation (“noise”) in the plots.

The AAIR for ovarian carcinoma (119.9 per million) was substantially higher than for peritoneal (6.78 per million) or fallopian tube (3.72 per million) carcinomas (Table 1). The highest rates for all three malignancies were among Whites, although rates for American Indian-Alaska Native women could only be calculated for ovarian carcinoma. Race-specific IRRs within each cancer site were significantly lower among all racial groups compared with White women. The most disparate IRRs were for carcinoma incidence among Black women compared with White women, which were 0.62 for ovary, 0.39 for peritoneum, and 0.72 for fallopian tube. By contrast, IRRs for carcinoma incidence among Asian-Pacific Islander women compared with White women had a narrower range, from 0.56 for peritoneum to 0.67 for fallopian tube.

Hispanic women had lower rates of malignancy for all three anatomic sites in comparison with non-Hispanic women. The smallest difference was for ovarian carcinoma (IRR, 0.78) and the largest difference was for fallopian tube carcinoma (IRR, 0.64).

Peritoneal carcinomas were diagnosed at a more advanced age (67 years) than fallopian tube carcinomas (64 years) and ovarian carcinomas (63 years). This pattern was generally consistent for all racial and ethnic groups, although Black, Asian-Pacific Islander, and Hispanic women tended to be diagnosed at younger mean ages compared with White women (data not shown).

The incidence of ovarian carcinoma rose monotonically with increasing age up to the age of 75 to 79 years before declining slightly among women ≥80 years (Fig. 1). Rates for peritoneal and fallopian tube carcinomas rose more gradually than for ovarian carcinoma and peaked earlier among women aged 70 to 74 years.

Histologic comparisons between anatomic sites were hampered by the high percentage of cases with a diagnosis of “adenocarcinoma, NOS” or “unspecified.” This was particularly true for fallopian tube carcinomas in which 45.8% of cases fell within these categories (Table 2). Serous carcinomas were the most commonly diagnosed histologic type for all anatomic sites. Nonserous carcinomas were extremely uncommon in the peritoneum, whereas endometrioid tumors were the next most commonly specified ovarian and fallopian tube histologic subtype. Clear cell and mucinous tumors were rarely diagnosed in the fallopian tube.

Peritoneal carcinomas were diagnosed at more advanced stages (85% regional/distant) than were ovarian (76% regional/distant) or fallopian tube (62% regional/distant) malignancies, and a comparatively greater percentage of peritoneal carcinomas were unstaged at diagnosis (Table 3). Few notable differences in stage at diagnosis for ovarian and peritoneal carcinomas were observed by race or ethnicity. However, White women with fallopian tube carcinoma tended to be diagnosed at later stages compared with Black or Asian-Pacific Islander women.

The incidence rate for all three carcinoma types in the South was lower than in any of the other three U.S. regions (Table 4). The rates for ovarian and fallopian tube cancers were highest in the Northeast, and the rate for peritoneal carcinoma was highest in the West. The largest regional variation was found for peritoneal carcinomas with AAIRs that were 83% higher among women in the West compared with women in the South.

White women had higher rates for carcinoma at each anatomic site in all regions compared with women of other races. The highest rates for ovarian and fallopian carcinomas were observed among White and Black women in the Northeast, whereas White and Black women in the West had significantly higher rates for peritoneal carcinoma compared with White and Black women in other regions. Regional patterns of cancer incidence were distinct for Asian-Pacific Islander women who experienced the highest incidence of ovarian and peritoneal cancers in the West, although there were too few fallopian tube cancers among Asian-Pacific Islanders to examine regional variation directly. The South was the only region where the AAIR for ovarian carcinoma among Black women exceeded that among Asian-Pacific Islander women.

Regional variation in the incidence for ovarian and fallopian tube carcinomas among Hispanic women was in strong contrast to that among non-Hispanic women. Ovarian cancer rates for Hispanic women were highest in the West and lowest in the Northeast, where the incidence among non-Hispanic women was 44% higher than the incidence among Hispanic women. In comparison, the incidence for fallopian tube carcinoma was highest in the South among Hispanic women but lowest in the South among non-Hispanic women. The greatest regional difference in rates for fallopian tube cancer was in the West, where the incidence rates were 91% higher among non-Hispanic compared with Hispanic women.

A higher proportion of women in the South compared with other regions were unstaged for ovarian carcinoma, and a higher proportion of women in the Northeast were unstaged for peritoneal carcinoma. However, few other consistencies by cancer site in stage at diagnosis by region were evident. A greater percentage of ovarian carcinomas were diagnosed at a localized stage in the Northeast and a lower percentage in the West compared with other regions. Localized fallopian tube cancer was most common in the Midwest and least common in the South. Little regional variation in stage at diagnosis was observed for peritoneal cancer.

The incidence rate for ovarian cancer decreased by 26.5% (P_trend < 0.0001) between 1973 and 2005, with an annual change in incidence of −0.84% (95% CI, −0.769% to −0.99%; see Fig. 2). This annual percent change was somewhat greater for Whites (−0.76%; 95% CI, −0.59% to −0.93%; P < 0.0001) than for Blacks (−0.72%; 95% CI, −0.43% to −1.01%; P < 0.0001). By contrast, the incidence of fallopian tube cancer increased by 79.3% (P_trend = 0.05). The increase in the average annual incidence rate for fallopian tube cancer among White women was 0.81% (95% CI, 0.29-1.32%; P = 0.003) but could not be calculated for other racial/ethnic groups. Trend statistics for peritoneal carcinoma could not be calculated because few cases were diagnosed before the mid-1980s.

Results of this study suggest some important parallels in the incidence patterns for peritoneal, fallopian tube, and ovarian carcinomas. The rates for all three malignancies were significantly higher among White women compared with Black or Asian-Pacific Islander women, and higher among non-Hispanic women compared with Hispanic women. These observations agree with a previous report that White and non-Hispanic women in the United States are at a substantially greater risk of ovarian cancer compared with women of other racial and ethnic groups (21).

In contrast to ovarian cancer, few descriptive studies of fallopian tube cancer have been conducted, presumably because of the difficulty in assembling a sufficient number of cases. A recent SEER and NPCR-based study conducted by Stewart et al. (22) reported a rate of 0.41 cases per 100,000. In agreement with our data, rates were higher among White women than among Black and Hispanic women. The present analysis also identified low rates for fallopian tube carcinomas among Asian-Pacific Islanders, a group that has not been previously examined. Studies published in Finland (23) and Denmark (24) in the past several decades suggest rates comparable with the United States, varying between 0.3 and 0.5 per 100,000, although direct comparisons are not possible because of differences in standard populations used for age adjustment.

The etiology of fallopian tube cancer is unknown, although some recent data are beginning to emerge for a hormonal etiology (25). A case-control study using Finnish Cancer Registry data reported a protective effect of pregnancy and sterilization against fallopian tube cancer, suggesting possible commonalities with ovarian cancer (26). Oral contraceptive use may also be protective (27). In addition to molecular evidence linking fallopian tube cancer to BRCA germ-line mutations (28), several investigators have detected a higher prevalence of BRCA1 and BRCA2 mutations among fallopian tube cancer patients than in the general population (29, 30).

Primary peritoneal cancer was first recognized in 1959 as a distinct clinical entity (31), although the present analysis identified only two cases diagnosed before the 1980s in the United States. Published reports have been limited to clinical presentations and case series, few exceeding 100 patients (1, 3, 6). An analysis of U.S. data between 1992 and 1997 reported primary peritoneal cancer rates of 0.03 per 100,000 (1), but this report was limited to mucoepidermoid carcinoma, cystadenocarcinoma, serous cystadenocarcinoma, or serous cystadenoma of borderline malignancy. Eltabbakh et al. (4) compared information from 50 women with primary peritoneal cancer and 503 women with epithelial ovarian cancer and found that women with primary peritoneal cancer were older, had later menarche, and were less likely to have used perineal talc. These differences were not confirmed in a larger comparison of 95 peritoneal cancer cases with 117 advanced ovarian cancer cases (6). Peritoneal cancer seems to be part of the hereditary breast-ovarian cancer syndrome as the frequency of BRCA mutations in peritoneal and ovarian cancer cases is similar (32, 33).

Geographic patterns of cancer incidence provide clues to environmental exposures or other factors that may influence risk. The Southern region of the United States had the lowest rates for all pelvic sites, suggesting common risk factors. As noted by Hall et al. (34), surgical procedures such as oophorectomy, tubal ligation, and tubal sterilization that may influence pelvic cancer rates also vary among regions of the United States and by racial and ethnic group. Some regional variation in rates may also reflect diagnostic and clinical practice. For example, the higher rates of peritoneal carcinoma among White and Black women in the West may have resulted from nationwide differences in the histopathologic classification of pelvic malignancies. Screening methods for the early detection of pelvic malignancies, such as CA125 and transvaginal ultrasound, have been ineffective, so they cannot account for regional variation by stage (35). However, regional differences in access to medical care and the extent of health insurance coverage may account for some of the observed geographic patterns by stage at diagnosis in this analysis (34).

Time trend analyses suggest that at least some of the decrease in ovarian cancer during the past 3 decades might be artifactual and attributable to increased diagnosis of peritoneal and fallopian tube malignancies. According to Chen et al. (36), Gynecologic Oncology Group criteria established in 1993 to define primary peritoneal cancer resulted in reclassification of 7% to 20% of ovarian carcinomas to the peritoneum. In the present analysis, SEER data showed a 243% increase in the incidence of peritoneal carcinoma and a 14% decrease in ovarian carcinoma between 1994 and 2005.

An important strength of this analysis was the quality and comprehensiveness of the NAACCR data, which allowed us to examine patterns in pelvic carcinoma incidence by demographic, clinical, and pathologic factors. Limitations of this study include the quality of race and Hispanic origin coding in registries that might lead to misreporting and underreporting of race and ethnicity (37, 38). A recent evaluation of the accuracy of race and Hispanic ethnicity designations in the SEER data found the quality of race information to be excellent (39). The assignment of Hispanic ethnicity was deemed moderately accurate in the SEER data and the Hispanic identification algorithm implemented by NAACCR has been shown to reliably assign Hispanic ethnicity across the NAACCR registries (10). Hispanics are presented throughout the monograph as a single ethnic group, but clearly, there are regional differences in cultural background and origin among Hispanics that could influence pelvic carcinoma rates. Additional concerns in the interpretation of the data include possible differences in the classification of carcinomas by the pathologists in the various hospitals contributing to the NAACCR registries, the large percentage of unclassified cases, especially for the fallopian tube, and the discrepancies in the two classification systems used to stage cases during the time period under examination. Due to the large number of unclassified cases, the incidence estimates reported for detailed histologic types and stage do not reflect their true magnitude, and conclusions based on histologic-specific rates should be interpreted with great caution.

In conclusion, we observed a variety of racial, ethnic, and geographic similarities in the incidence patterns for ovarian, peritoneal, and fallopian tube carcinomas that support the likelihood of a common molecular pathogenesis for these malignancies. The high rates for all three pelvic carcinomas among White women suggest a shared genetic etiology. Reasons for the significantly higher rates of these carcinomas among non-Hispanic White women compared with other racial and ethnic groups need to be explored in well-designed, analytic studies within large research consortia.

No potential conflicts of interest were disclosed.

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.