Introduction

Cancer incidence rates should reflect the rate of cancer among individuals at risk [1]. Removal of the uterus (hysterectomy) or ovaries (oophorectomy) eliminates a woman’s risk of uterine or ovarian cancer [2], while hysterectomy with removal of the cervix substantially lowers the risk of cervical cancer [3]. Failure to adjust the denominator of these cancer rates for women who have had a hysterectomy and/or oophorectomy may lead to underestimations in cancer rates [1, 4, 5]. These corrections are significant because hysterectomy is the second most common surgery performed in women [6].

Previous studies used a variety of data sources to estimate hysterectomy prevalence and cancer incidence rates in women of all races in the United States [7–10]. To date, no study has focused on these corrections in AI/AN women at a regional level, an important consideration because several recent reports have shown that wide regional variation in cancer burden is characteristic of AI/AN populations [11, 12]. Additionally, AI/AN gynecologic cancer rates are more severely impacted by these corrections because of their comparatively high hysterectomy prevalence. In a previous study, AI/AN women had a higher hysterectomy prevalence than other racial groups in 19 of 31 states with reportable numbers [13]; another found that AI/AN women had the highest prevalence of hysterectomy in women under 40 years old [14]. Our aim is to present, by geographic region, the rates of cervical, uterine, and ovarian cancer in AI/AN women adjusted for population-at-risk using updated estimates of cancer incidence and hysterectomy and oophorectomy prevalence in this population. Improved cancer incidence data may strengthen the evaluation of cancer control efforts, particularly for cervical cancer for which established prevention and early detection measures exist.

Materials and methods

Cancer incidence

In the United States, two federal programs fund central cancer registries: the National Program of Cancer Registries (NPCR) of the Centers for Disease Control and Prevention (CDC) and the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute (NCI). Together, these registries covered 100% of the US population. Primary cancer site and histology data are coded according to the International Classification of Diseases for Oncology (ICD-O) edition in use at the time of diagnosis and are converted to the Third Edition [15]. Incident cancer cases diagnosed from 1999 to 2004 from central cancer registries that met the United States Cancer Statistics standard for high-quality data were included in this analysis (see footnote to Table 2 for list of registries) [16].

Only invasive cancers of the cervix (ICD-O-3 C530–C539), uterus (ICD-O-3 C540–C549, C559), and ovaries (ICD-O-3 C569) are included in our analysis. Lymphomas, mesothelioma, and Kaposi’s sarcoma were excluded from the analysis.

To reduce the racial misclassification of AI/AN cases as non-native, all case records from the NPCR and SEER central cancer registries were linked with the Indian Health Service (IHS) patient registration database as described elsewhere [17]. The IHS provides medical services to AI/AN persons who are eligible members of federally recognized tribes.

To further improve on race classification, we restricted our analyses to counties designated by the IHS as “Contract Health Service Delivery Areas” (CHSDA) as described elsewhere [17]. Although less geographically representative, analyses restricted to CHSDA counties are presented in this report for the purpose of improving accuracy for cancer incidence for AI/AN women.

The analyses were completed for all regions combined and by individual IHS regions: Alaska, Pacific Coast, Northern Plains, Southern Plains, Southwest, and East (Fig. 1). The rationale for regional analyses for cancer, and other health outcomes, for AI/AN has been described elsewhere [17].

Fig. 1
figure 1

States and IHS Contract Health Service Delivery Areas (CHSDA) counties by Indian Health Service Region

Full size image

Estimating hysterectomy prevalence

The Behavioral Risk Factor Surveillance System (BRFSS) was the primary source of hysterectomy prevalence among self-identified AI/AN and non-Hispanic White (NHW) women [18]. Women who reported “an operation to remove the uterus/womb” were identified in BRFSS as having had a hysterectomy. No questions on the date of the surgery, the type of procedure, or the reason for hysterectomy are asked. We limited analysis of the BRFSS data to the years 1999, 2000, 2002, 2004, and 2006 when the question on hysterectomy status was included in the survey questionnaire for all states.

Since some hysterectomy procedures leave the cervix intact (subtotal hysterectomy), adjustment was needed to accurately estimate the proportion of women with a history of hysterectomy still at the risk of cervical cancer. Similarly, only those hysterectomies accompanied by bilateral oophorectomy, a procedure in which both ovaries are removed, were utilized to adjust the rate of ovarian cancer. No estimate for the procedure bilateral oophorectomy alone was available because no reliable and comparable data source for NHW and AI/AN women exists.

The National Hospital Discharge Survey (NHDS) collects data on inpatient utilization, including treatment and procedures received during hospitalization and discharge diagnoses, from non-federal short-stay hospitals [19]. The NHDS data from 1999 to 2005 were used to generate adjustment factors to estimate the proportion of hysterectomies that removed the cervix (ICD-9-CM code 68.3) and hysterectomies that were accompanied by bilateral oophorectomy (ICD-9-CM codes: 65.5–65.6 and 68.3–68.9) for NHW women.

Similar adjustments to hysterectomy prevalence data for AI/AN were necessary; however, NHDS had limited numbers of AI/AN women. We therefore queried years 2002–2005 from the IHS National Data Warehouse (NDW), which contains registration and encounter-based patient data from IHS/Tribal/Urban facilities, using methods similar to those described above for NHDS [20]. The hysterectomy prevalence for both NHW and AI/AN women was corrected by the appropriate hysterectomy subtype adjusted by age, as summarized in Table 1.

Table 1 Adjustment of hysterectomy prevalence by hysterectomy type to assess the impact of hysterectomy prevalence on the rates of cervical, uterine, and ovarian cancer among AI/AN and NHW women in the United States, 1999–2004
Full size table

Correcting population estimates for hysterectomy prevalence

Population estimates used to derive denominators in the rate calculations were obtained from the NCI’s publicly available web-based statistical resources [21] as described elsewhere [17].

To derive the appropriate denominator, by region and age group, to calculate the cervical cancer rates, the population estimates were adjusted downward in proportion to the prevalence of hysterectomy (further adjusted for subtotal hysterectomy as described above). Similarly, denominators to calculate uterine cancer rates were adjusted for the prevalence of hysterectomy alone, without further adjustment by the type of hysterectomy. Finally, denominators for the ovarian cancer rates were adjusted for BRFSS prevalence of hysterectomy further adjusted for the proportion of hysterectomy accompanied by bilateral oophorectomy. The denominators in the rate calculations were adjusted by age, region, and county-specific hysterectomy and oophorectomy prevalence.

Statistical analyses

All rates, expressed per 100,000 population per year, were directly age-adjusted, using SEER*Stat software [22], to the 2000 US standard population [23]. Using the corrected incidence rates, standardized rate ratios (RR) were calculated for AI/AN populations using NHW rates for comparison. Rate ratios are calculated in SEER*Stat prior to rounding of rates and may not equal RR calculated by the reader from rounded rates presented in the tables. Confidence intervals (CI) for age-adjusted rates and standardized rate ratios (RR) were calculated based on methods described by Tiwari et al. [24] using SEER*Stat 6.3.6. The CDC and IHS Institutional Review Boards determined the protocol to be public health practice, not research, and therefore not requiring IRB approval.

Results

The prevalence of hysterectomy in AI/AN versus NHW women by region is displayed in Fig. 2. Hysterectomy prevalence was highest among AI/AN and NHW women living in the Southern Plains region (26%) and lowest among women in Alaska (18%). AI/AN women had an overall higher prevalence of hysterectomy compared with NHW women in all regions except Alaska and the Southern Plains. Among AI/AN women, prevalence was lowest in the younger age groups of 18–29 years (1%) and 30–39 years (5%) and highest in the 60–69 year age group (52%). Prevalence of hysterectomy tended to be higher among AI/AN than NHW in all age categories except 30–39 and 70+ years (Fig. 3). Estimates for the proportion of all hysterectomies that were subtotal hysterectomy were similar in AI/AN (4%) and NHW women (6%). NHW women had a higher proportion of hysterectomy accompanied by bilateral oophorectomy (55%) compared with AI/AN women (35%).

Fig. 2
figure 2

Hysterectomy prevalence by Indian Health Service Region for the American Indians/Alaska Natives and non-Hispanic White populations—BRFSS, CHSDA counties, 1999–2006. AI/AN American Indians/Alaska Natives, BRFSS Behavioral Risk Factor Surveillance System, CHSDA Contract Health Service Delivery Areas, NHW non-Hispanic Whites

Full size image
Fig. 3
figure 3

Hysterectomy prevalence by age group for the American Indians/Alaska Natives and non-Hispanic White populations—BRFSS, CHSDA counties, 1999–2006. AI/AN American Indians/Alaska Natives, BRFSS Behavioral Risk Factor Surveillance System, CHSDA Contract Health Service Delivery Areas, NHW non-Hispanic Whites

Full size image

The uncorrected and corrected cervical cancer incidence rates among AI/AN women are shown in Fig. 4 and reveal a 43% increase in overall cervical cancer incidence rate after correction for hysterectomy where the cervix was removed for AI/AN women. The largest increases were found in the Southwest (72%, from 7.8 uncorrected to 13.4 corrected per 100,000) and East (58%, 7.1 to 11.2 per 100,000). For uterine cancer, the percent change resulting from the correction for hysterectomy status ranged from 51% in Alaska to 76% in the Southern Plains; the overall increase in uterine cancer rate was 67% among AI/AN women (Table 2). Corrected ovarian cancer rates were approximately one-third higher than uncorrected rates for AI/AN women across all regions (Table 2). The largest increases were noted in the Southwest and East.

Fig. 4
figure 4

Uncorrected and corrected rates per 100,000 persons of cervical cancer by Indian Health Service Region for American Indians/Alaska Natives, CHSDA counties, 1999–2004. CHSDA Contract Health Service Delivery Areas

Full size image
Table 2 Uncorrected and corrected gynecologic cancer incidence by Indian Health Service Region for American Indians/Alaska Nativesa and non-Hispanic Whites, CHSDA counties, 1999–2004
Full size table

AI/AN women had higher rates of cervical cancer than NHW women in all regions, and an overall 5% increase in the difference (assessed by an increase in rate ratio of AI/AN:NHW) of cervical cancer incidence between these racial/ethnic groups was found (Table 2). A 20% increase in disparities for cervical cancer was noted in the East and Southwest, while disparities decreased slightly in Alaska (−6%), the Southern Plains (−3%), and Pacific Coast (−3%). Uterine cancer rates in AI/AN were lower than NHW in all regions except the Southern Plains. Correction for hysterectomy status had no significant effect on the rate ratio of nationwide uterine cancer rates for AI/AN compared with NHW. Though rates of ovarian cancer are generally lower among AI/AN women when compared with NHW women, correction for oophorectomy status brought nationwide AI/AN rates of ovarian cancer 20% closer to NHW rates.

Discussion

In our study, risk-corrected cervical, uterine, and ovarian cancer incidence rates were substantially higher for AI/AN and NHW women than cancer rates traditionally reported that are uncorrected for hysterectomy and oophorectomy status. Given the importance of regional differences in cancer occurrence in AI/AN [12], including cervical cancer [25], we present our current findings by IHS region to further aid in their interpretation. The East and Southwest experienced the largest increases in AI/AN rates for the 3 gynecologic cancers examined after correction for estimated prevalence of hysterectomy and oopherectomy. Furthermore, we found that risk-correction affected the differences in measures of incidence of these cancers for AI/AN compared with NHW women, most notably with substantially larger cervical cancer disparities in the East and Southwest regions. The effects of risk-correction on cancer incidence rates underscore the need to more accurately describe the impact of these cancers on women truly at risk.

Consistent with previous studies, we found the hysterectomy prevalence for AI/AN women to be higher than for NHW women [13, 14]. Several explanations have been suggested for the higher prevalence of hysterectomy in AI/AN women. Greater use of hysterectomy for sterilization among AI/AN in the 1960s and 1970s may be a contributing factor [26]. Also, predictors for hysterectomy, such as limited education, low socioeconomic status, high parity, and history of miscarriage, are more commonly found in the AI/AN population [27–29]. Indications for hysterectomy differ by age; uterine leiomyoma is most common indication for women aged 35–54 years, whereas uterine prolapse and cancer are the most common indications for women aged 55 years or older [30]. Few studies have been published on indications specifically for AI/AN women, but the higher hysterectomy prevalence is not explained by higher rates of uterine fibroids among AI/AN (1.8%) compared with NHW (1.9%) [31].

Our analysis revealed an overall increase in cervical cancer incidence rates of 43% among AI/AN women after risk-correction for hysterectomy. In a previous study using US data from 1990 to 1992, risk-correction for the point prevalence of hysterectomy increased the cervical cancer incidence rate by 38% [1]. A more recent study on the impact of hysterectomy on race-specific gynecological cancer rates reported that the risk-corrected compared with uncorrected incidence rates were 45.6% higher for cervical cancer in American Indians/Alaska Native (AI/AN) women [4]. Though substantial progress has been achieved in cervical cancer control in AI/AN women, our analysis revealed an increased disparity in cervical cancer incidence between AI/AN and NHW women, especially in certain regions, than previously reported for this time period. In addition to this finding, AI/AN women are on average diagnosed at more advanced stages of cervical cancer than NHW women [25, 32] underscoring the need to maintain and strengthen efforts to screen and provide adequate follow-up for abnormal results. For those women who have not undergone hysterectomy, screening for prevention and early detection of cervical cancer is critical.

Estimating the actual population-at-risk is especially important in cervical cancer for evaluating the impact of cervical cancer screening, which has been conducted through Pap cytology [33, 34]. The majority of newly diagnosed cervical cancer cases can be attributed to a lack of, or infrequent, screening [35]. Recent BRFSS data suggest a lower prevalence of cervical cancer screening in AI/AN women compared with NHW [36]. The higher incidence of cervical cancer incidence rates among AI/AN after risk-correction underscores the continued need for culturally sensitive population-based screening programs as well as promotion of the HPV vaccines recommended in the United States [37].

Uterine cancer is the most commonly diagnosed malignancy of the female reproductive system across all races in the United States [38]. Few studies have examined uterine cancer rates among AI/AN women; most published studies are limited to subpopulations or older data [28, 39, 40]. In our study, correction for hysterectomy status increased the national AI/AN incidence rate of uterine cancer by 67% with the most substantial rises in those over 65 years old. This change is even more striking than those in an earlier study of impact of hysterectomy on uterine cancer incidence from 1960 to 1973, in which risk-correction increased the rates by 20–45% with the greatest increases in women over 60 years [41]. Also consistent with previous studies, we found AI/AN women to have lower rates of uterine cancer compared with NHW women, even after correction for hysterectomy status [28, 39, 40]. The reason for the lower burden of uterine cancer in AI/AN women is unclear. Some risk factors for uterine cancer such as obesity (OR ≥ 4.0) and diabetes (OR ≥ 1.2) are more prevalent among AI/AN women [42, 43]. This paradox is also observed for uterine cancer among African American women [44]. However, unlike African American women who tend to have more advanced disease than NHW women [45], AI/AN women are diagnosed with uterine cancers of similar grade and stage as NHW women [40]. Some factors that have been associated with lower risk—such as increased parity [46] and smoking [43, 47] are more common in AI/AN women and point to our limited knowledge about risk factors for uterine cancer among AI/AN women. Correction for hysterectomy status had no effect on the overall uterine cancer incidence rate ratio for AI/AN to NHW women.

Using uncorrected estimates, ovarian cancer is the seventh most common cancer and the leading cause of death from gynecologic cancers among AI/AN women [28]. While ovarian cancer incidence decreased in white and non-Hispanic women, rates remained stable in AI/AN women from 1995 to 2004 [28]. With correction for bilateral oophorectomy status in our study, ovarian cancer rates among AI/AN women were 37% higher. Similar to cervical cancer, AI/AN women are more often diagnosed at a more distant stage than NHW women [48]. The high mortality and relative frequency of ovarian cancer underscores the importance of continuing the search for better methods for early detection and treatment. Adequate monitoring of oophorectomy prevalence and, hence, more accurate measures of ovarian cancer incidence will be critical to evaluate future ovarian cancer prevention and early detection efforts.

Though we presented findings based on the most representative and accurate data available on AI/AN and NHW populations, our findings should be considered in light of several limitations. First, cancer incidence estimates are restricted to the 624 CHSDA counties, where 56% of the US AI/AN population reside [17]. Despite restriction of the analysis to CHSDA counties and registry linkages described elsewhere [17], race misclassification likely continues to influence cancer surveillance data, such as in this report. Second, incidence rates for these cancers are based on small case numbers, particularly when stratified by region, county, and age group, which diminish the stability of the estimates. Third, use of BRFSS for estimating hysterectomy prevalence is limited to respondents with landline telephones and to persons who speak English; both characteristics are limited in some rural AI/AN populations. BRFSS also relies on self-report of hysterectomy status and cross-sectional cumulative prevalence of hysterectomy, rather than age-specific hysterectomy rates. Furthermore, BRFSS led to some small sample cell sizes when stratified by age and region.

Other limitations relate to assumptions required to adjust gynecologic cancer rates for hysterectomy and oophorectomy status. We assumed that the proportion of hysterectomies and/or oophorectomies undertaken for gynecologic cancers was very small compared with those done for other indications and will therefore not substantially impact the denominator adjustment. In a previous study using NHDS from 1988 to 1997, the majority of hysterectomies (83.1%) were performed for benign conditions and only 9.7% were for the primary diagnosis of cervical neoplasia [49]. Also the distribution of hysterectomy types change over time. There is evidence that the rate of hysterectomy and the indications for hysterectomy have changed little over the last decade [50, 51], but actual proportions of prophylactic oophorectomy at the time of hysterectomy increased from 1979 to 2004 [52]. Additionally, the hysterectomy subtype estimates were derived from two different data sources, IHS NDW for AI/AN and NHDS for NHW, which may introduce bias if surveillance accuracy differs between the two sources. Though we utilized the longest time frame of data available and some historical procedures are captured in both systems, we used only more recent data from NDW (2002–2005) and NHDS (1999–2005) to determine the proportions of hysterectomies that were subtotal or accompanied by oophorectomy in AI/AN and NHW women. In spite of these limitations, our estimates of these procedures in NDW were similar to findings in previous studies, which included data for the general population of women [6, 52–54] and AI/AN women specifically [7]. Finally, NHDS did not have a “NHW” designation for race/ethnicity, so “white” was used as the best available estimate for the NHW hysterectomy corrections to enable comparisons between AI/AN- and NHW-adjusted cancer rates. “White” and NHW are terms that may describe epidemiologically different populations, and approximately 18% of “white” individuals also identified as Hispanic [55].

Current methods of calculating cervical, uterine, and ovarian cancer incidence underestimate the rates of these cancers in women at risk. Our study utilized more accurate and regionally specific hysterectomy and bilateral oophorectomy-corrected surveillance methods for AI/AN and NHW women. Wide regional variation is characteristic of AI/AN cancer incidence and gynecologic care practices, such as use of hysterectomy. Because of these differences and the anticipated increase in cancers diagnosed in the AI/AN population [56], region-specific data that are corrected for true population-at-risk are necessary to characterize the cancer burden among AI/AN women [28].

Though such corrections have not been systematically used in any country [5], we recommend continued monitoring of AI/AN gynecologic cancer incidence rates using these enhanced surveillance methods. Risk-correction should also be considered when assessing cancer disparities, as illustrated by the regional increases in cervical cancer disparity between AI/AN and NHW after hysterectomy correction. For all races, additional questions in national surveys, such as BRFSS, are needed to provide more accurate national data on prevalence of hysterectomy and bilateral oophorectomy and indications for these procedures. Improved surveillance data are needed to better inform disease control priorities for reducing AI/AN health disparities and to track progress toward that goal. In addition, more accurate rates will identify the real needs of these women for early detection, treatment, and survivorship cancer services.