Volume 123, Issue 6 p. 1044-1050
Original Article
Free Access

Hysterectomy-corrected cervical cancer mortality rates reveal a larger racial disparity in the United States

Anna L. Beavis MD, MPH

Anna L. Beavis MD, MPH

Kelly Gynecologic Oncology Service, Department of Gynecology and Obstetrics, Johns Hopkins Medicine, Baltimore, Maryland

Search for more papers by this author
Patti E. Gravitt PhD

Patti E. Gravitt PhD

Milken Institute School of Public Health, Department of Global Health, George Washington University, Washington, DC

Search for more papers by this author
Anne F. Rositch PhD, MSPH

Corresponding Author

Anne F. Rositch PhD, MSPH

Department of Epidemiology, Bloomberg School of Health at Johns Hopkins, Baltimore, Maryland

Corresponding author: Anne F. Rositch, PhD, MSPH, Department of Epidemiology, Bloomberg School of Health at Johns Hopkins, 615 North Wolfe Street, Office E6150, Baltimore, MD 21205; [email protected]Search for more papers by this author
First published: 23 January 2017
Citations: 151

See editorial on pages 915-6, this issue.

Abstract

BACKGROUND

The objectives of this study were to determine the age-standardized and age-specific annual US cervical cancer mortality rates after correction for the prevalence of hysterectomy and to evaluate disparities by age and race.

METHODS

Estimates for deaths due to cervical cancer stratified by age, state, year, and race were derived from the National Center for Health Statistics county mortality data (2000-2012). Equivalently stratified data on the prevalence of hysterectomy for women 20 years old or older from the Behavioral Risk Factor Surveillance System survey were used to remove women who were not at risk from the denominator. Age-specific and age-standardized mortality rates were computed, and trends in mortality rates were analyzed with Joinpoint regression.

RESULTS

Age-standardized rates were higher for both races after correction. For black women, the corrected mortality rate was 10.1 per 100,000 (95% confidence interval [CI], 9.6-10.6), whereas the uncorrected rate was 5.7 per 100,000 (95% CI, 5.5-6.0). The corrected rate for white women was 4.7 per 100,000 (95% CI, 4.6-4.8), whereas the uncorrected rate was 3.2 per 100,000 (95% CI, 3.1-3.2). Without the correction, the disparity in mortality between races was underestimated by 44%. Black women who were 85 years old or older had the highest corrected rate: 37.2 deaths per 100,000. A trend analysis of corrected rates demonstrated that white women's rates decreased at 0.8% per year, whereas the annual decrease for black women was 3.6% (P < .05).

CONCLUSIONS

A correction for hysterectomy has revealed that cervical cancer mortality rates are underestimated, particularly in black women. The highest rates are seen in the oldest black women, and public health efforts should focus on appropriate screening and adequate treatment in this population. Cancer 2017;123:1044–50. © 2016 American Cancer Society.

INTRODUCTION

Each year in the United States, more than 12,000 women are diagnosed with cervical cancer, and more than 4,000 women die of it.1 Racial minorities, particularly black women, have significantly higher incidence and mortality rates for this disease.2-7 Prior studies have shown that a failure to account for the prevalence of hysterectomy has resulted in an underestimation of cervical cancer incidence rates because women who have had their cervix surgically removed are inappropriately retained in the population-at-risk denominator.2, 3, 8 This underestimation has the most profound effect in black women because they have the highest prevalence of hysterectomy: age specific cervical cancer incidence rates in black women increase by as much as 125% after correction, whereas there is an 83% increase in white women of the same age.2 As such, the uncorrected race-specific incidence rates represent an underestimation of the true racial disparity.2

Recent data suggest that the mortality gap between black and white women is closing because of a decline in cervical cancer mortality rates among black women.9 However, the effect of correction for hysterectomy prevalence on US cervical cancer mortality rates and trends has not been evaluated. Therefore, we sought to estimate the hysterectomy-corrected US annual age-specific and age-standardized cervical cancer mortality rates over the last decade to demonstrate the true magnitude of the racial disparity.

MATERIALS AND METHODS

To estimate hysterectomy-corrected cervical cancer mortality rates, we first estimated the age-, race-, year-, and state-stratified hysterectomy prevalence from national survey data, and we then used those estimates to correct the denominator of equivalently stratified cervical cancer mortality rate estimates.10, 11 Direct standardization with the US 2000 standard population was performed to estimate age-standardized rates.12 The institutional review board deemed the study exempt from review as only publicly available de-identified information was used.

Data Sources and Database Linkage

Behavioral Risk Factor Surveillance System (BRFSS)

Survey-weighted hysterectomy prevalence estimates were obtained from the BRFSS survey, a nationally representative random-digit dialing survey of health behaviors that reaches more than 400,000 persons in the United States annually.10 These results were stratified by age (≥20 years in 5-year age categories), race (black, white, or other without regard for Hispanic ethnicity), state, and year (2000-2012). Data were available for hysterectomy prevalence only during even-numbered years; therefore, stratified hysterectomy prevalence estimates were imputed for the odd-numbered years by the calculation of a weighted average of the flanking even-numbered years. A similar approach was used in the case of other missing data; for example, in Hawaii, no hysterectomy data were collected in 2004, so a weighted average was used from the flanking years.

Surveillance, Epidemiology, and End Results (SEER) and National Center for Health Statistics (NCHS)

The SEER mortality database was used to obtain estimates of cervical cancer mortality rates, including case counts and population-at-risk denominators, stratified in the same manner as the hysterectomy prevalence.11 This database obtains its data from death certificate information provided by the NCHS and combines these data with data from the SEER incidence registries. However, because of the suppression of data for any rate based on fewer than 10 cases produced by SEER*Stat software, we obtained permission from the NCHS to access the exact annual case counts, stratified by age, race, year, and state, for deaths coded as “malignant neoplasm of the cervix uteri.”13 The SEER mortality database was queried to determine the equivalently stratified population-at-risk estimates for the states present in the SEER-18 database.11

Statistical Methods

The estimates for hysterectomy prevalence from the BRFSS, the case counts of death due to cervical cancer (n) from the NCHS, and the population-at-risk denominator (p) from SEER were merged into 1 database with stratification by age category, race, state, and year. The hysterectomy-corrected population-at-risk denominator (pc) was then calculated with the following equation:
urn:x-wiley:0008543X:media:cncr30507:cncr30507-math-0001
where h is the hysterectomy prevalence.

Age-standardized annual mortality rates were calculated with direct standardization with weights from US Census 2000 population estimates for women ≥ 20 years old.12 Standard errors for the age-standardized mortality rates were calculated with standard methods published by the National Vital Statistics System of the Centers for Disease Control and Prevention.14 Mortality rate ratios and corresponding 95% confidence intervals (CIs) were calculated with methods published by the International Agency for Research on Cancer for the comparison of age-standardized rates.15 Because the heterogeneity of the other-race category made it difficult to draw meaningful conclusions about the implications of the results, we present only the overall, black, and white racial category results. The results from the other-race category can be found in the online Supporting Information in Figure 1, Table 1, and Table 2.

Overall and race-stratified age-specific mortality rates were calculated with the following formula:
urn:x-wiley:0008543X:media:cncr30507:cncr30507-math-0002
where MRc is the mortality rate corrected for the hysterectomy prevalence. Ninety-five percent CIs were calculated with Bernoulli's equation:
urn:x-wiley:0008543X:media:cncr30507:cncr30507-math-0003(1)

Data linkage and management were performed with Excel 13, and statistical analyses were performed with Stata (version 14).

Trends in hysterectomy prevalence and cervical cancer mortality rates over time and across age groups were evaluated with the National Cancer Institute's Joinpoint regression software (version 4.2.0), which uses log-linear regression to fit the simplest trend of the data and calculate percentage changes and to perform pairwise comparisons of trends.16 The Bayesian information criterion method for selecting the best model was chosen because it finds the model with the best fit by penalizing for the cost of extra parameters.17 An α value < .05 was considered significant.

RESULTS

Hysterectomy Estimates

Overall, the prevalence of hysterectomy was 20% (95% CI, 20%-21%) for women ≥ 20 years old (see Supporting Information Table 1) and was higher for black women than white women (23% vs 21%, P < 0.05). The age-specific hysterectomy prevalence was also higher for black women than white women for all ages between 45 and 69 years (P < .05). The hysterectomy prevalence peaked at the same age for white and black women (65-69 years), but the peak was higher for black women (58% vs 43% for white women, P < .05; Fig. 1). Despite minor changes in prevalence from 2000 to 2012, there were no significant trends across time.

Details are in the caption following the image

Locally weighted scatterplot smoothing plots of age-specific estimates of the prevalence of hysterectomy for black and white women. The prevalence for both black and white women peaks at 65 to 69 years, but black women have a significantly higher peak prevalence (58% vs 43%, P < .05).

Age-Standardized Cervical Cancer Mortality Rates

Correction for hysterectomy resulted in significantly higher rates of cervical cancer mortality. The uncorrected age-standardized cervical cancer mortality rate for all races combined from 2000 to 2012 was 3.4 deaths per 100,000 women (95% CI, 3.3-3.4), whereas the corrected rate was 5.0 deaths per 100,000 women (95% CI, 4.9-5.1; data not shown). The corrected age-standardized rate for black women was 10.1 per 100,000 (95% CI, 9.6-10.6), whereas the uncorrected rate was 5.7 per 100,000 (95% CI, 5.5-6.0; Table 1). For white women, the corrected age-standardized rate was 4.7 per 100,000 (95% CI, 4.6-4.8), whereas the uncorrected rate was 3.2 per 100,000 (95% CI, 3.1-3.2). The uncorrected mortality rate ratio for black women versus white women was 1.8 (95% CI, 1.7-1.9), and the ratio increased to 2.2 (95% CI, 2.0-2.3) after correction.

Table 1. Age-Standardized Cervical Cancer Mortality Rates and Mortality Rate Ratios, Uncorrected and Corrected for the Prevalence of Hysterectomy, for Black and White Women From 2000 to 2012
Year Uncorrected Rates (95% CI) Corrected Rates (95% CI)
White Women Black Women Uncorrected Mortality Rate Ratio (95% CI) White Women Black Women Corrected Mortality Rate Ratio (95% CI)
2000 3.5 (3.3-3.7) 6.6 (5.6-7.7) 1.9 (1.5-3.2) 5.4 (5.1-5.8) 12.6 (10.4-14.8) 2.3 (1.8-3.6)
2001 3.6 (3.3-3.8) 6.5 (5.4-7.5) 1.8 (1.5-3.2) 5.4 (5.0-5.7) 12.8 (10.6-15.0) 2.4 (1.8-3.7)
2002 3.1 (2.9-3.3) 6.3 (5.3-7.3) 2.0 (1.6-3.3) 4.7 (4.3-5.0) 13.2 (10.8-15.6) 2.8 (2.1-4.1)
2003 3.2 (3.0-3.4) 6.1 (5.1-7.1) 1.9 (1.5-3.3) 4.8 (4.4-5.1) 11.4 (9.4-13.4) 2.4 (1.8-3.7)
2004 3.0 (2.8-3.2) 6.2 (5.2-7.1) 2.1 (1.6-3.4) 4.5 (4.2-4.8) 10.7 (9.0-12.5) 2.4 (1.9-3.7)
2005 3.2 (2.9-3.4) 5.7 (4.8-6.6) 1.8 (1.5-3.2) 4.7 (4.3-5.0) 9.7 (8.0-11.3) 2.1 (1.6-3.4)
2006 3.2 (3.0-3.4) 5.8 (4.9-6.7) 1.8 (1.4-3.2) 4.8 (4.4-5.1) 10.5 (8.7-12.4) 2.2 (1.7-3.5)
2007 3.1 (2.9-3.4) 5.6 (4.7-6.5) 1.8 (1.4-3.2) 4.6 (4.3-4.9) 9.8 (8.2-11.5) 2.1 (1.7-3.5)
2008 3.0 (2.8-3.3) 5.0 (4.2-5.8) 1.7 (1.3-3.1) 4.4 (4.1-4.7) 8.4 (7.0-9.8) 1.9 (1.5-3.3)
2009 3.1 (2.9-3.3) 5.8 (5.0-6.7) 1.9 (1.5-3.2) 4.4 (4.1-4.7) 9.9 (8.3-11.5) 2.2 (1.8-3.5)
2010 3.1 (2.9-3.3) 5.4 (4.5-6.2) 1.7 (1.3-3.1) 4.5 (4.2-4.9) 9.6 (8.0-11.2) 2.1 (1.7-3.4)
2011 3.0 (2.8-3.2) 5.8 (5.0-6.7) 1.9 (1.6-3.3) 4.4 (4.0-4.7) 9.8 (8.2-11.4) 2.3 (1.8-3.5)
2012 3.1 (2.9-3.3) 4.3 (3.5-5.0) 1.4 (1.1-3.0) 4.5 (4.2-4.8) 7.2 (5.8-8.5) 1.6 (1.3-3.1)
2000-2012 3.2 (3.1-3.2) 5.7 (5.5-6.0) 1.8 (1.7-1.9) 4.7 (4.6-4.8) 10.1 (9.6-10.6) 2.2 (2.0-2.3)
  • Abbreviation: CI, confidence interval.

A trend analysis of changes in age-standardized rates over time demonstrated similar patterns for uncorrected and corrected rates in white women: there was a nonsignificant (NS) decrease from 2000 to 2002 (uncorrected annual percentage change [APC], −6.05% [P = NS]; corrected APC, −7.45% [P = NS]), and there was a slower rate of decline from 2002 to 2012 that was significant only in the corrected rates (uncorrected APC, −0.42% [P = NS]; corrected APC, −0.76% [P < .05]; Fig. 2A). In black women, uncorrected and corrected rates also showed a similar pattern, with a significant decline demonstrated over the entire decade (uncorrected APC, −2.24% [P < .05]; corrected APC, −3.57% [P < .05]); the corrected APC was significantly greater than the uncorrected one (P for the pairwise comparison < .01; Fig. 2B). The corrected APC trend in black women was greater than the corrected APC trend in white women (−3.57% vs −0.76%, P for the pairwise comparison < .05).

Details are in the caption following the image

Trends in age-standardized cervical cancer mortality rates, uncorrected and corrected for the prevalence of hysterectomy, from 2000 to 2012 for (A) white and (B) black women. *The APC P value was significant at α < .05. APC indicates annual percentage change.

Age-Specific Cervical Cancer Mortality Rates

Age-specific cervical cancer mortality rates increased with age for all races, and corrected and uncorrected rates did not plateau at any point (Fig. 3A,B). Both corrected and uncorrected age-specific cervical cancer mortality rates were significantly higher for black women than white women in all age categories except 20 to 29 and 35 to 39 years. Corrected rates were highest for black women who were 85 years and older: before correction, the rate was 18.6 per 100,000 (95% CI, 15.4-21.7), and after correction, it was 37.2 per 100,000 (95% CI, 31.0-43.5). This demonstrates 101% increase (Table 2). The highest percentage increase was seen in black women aged 65 to 69 years, for whom rates increased by more than 126%. The percentage increase for white women of the same age was 75%.

Details are in the caption following the image

Age-specific cervical cancer mortality rates, uncorrected and corrected for the prevalence of hysterectomy, in (A) white and (B) black women.

Table 2. Uncorrected and Corrected Age-Specific Cervical Cancer Mortality Rates and Percentage Increases After Correction in White and Black Women
White Women Black Women
Age, y Uncorrected Rate Corrected Rate % Increase Uncorrected Rate Corrected Rate % Increase
20-24 0.1 0.1 1 0.2 0.2 1
25-29 0.5 0.5 1 0.7 0.7 1
30-34 1.3 1.3 3.1 2.1 2.1 2
35-39 2.3 2.4 6.1 2.9 3.1 7
40-44 2.9 3.3 13 4.6 5.3 15
45-49 3.7 4.5 22 5.0 6.7 34
50-54 4.1 5.5 34 6.6 10.9 65
55-59 4.6 6.7 47 7.1 12.2 72
60-64 4.8 7.7 59 9.1 17.7 95
65-69 4.9 8.6 75 10.5 23.8 126
70-74 5.2 9.5 81 9.3 19.2 106
75-79 5.7 11.0 96 13.4 29.7 122
80-84 6.1 11.8 93 15.9 33.4 110
≥85 6.5 11.6 77 18.6 37.2 101
All agesa 3.2 4.7 48 5.7 10.01 76
  • Rates are per 100,000 women.
  • a The age-standardized rates are based on direct standardization with the US 2000 census population for women 20 years old or older.12

DISCUSSION

National cervical cancer mortality rates have previously been reported without consideration of the proportion of women who are no longer at risk for cervical cancer by virtue of having had their cervix removed during a complete hysterectomy; this has resulted in an underestimation of the overall mortality rate and the race-specific disparity. Using a large, nationally representative sample to estimate the hysterectomy prevalence, we have demonstrated that age-standardized and age-specific cervical cancer mortality rates are significantly higher in all races after correction for the hysterectomy prevalence. The effect of hysterectomy correction is most notable for black women, in whom the corrected age-standardized mortality rates are more than double those seen in white women. In fact, this study demonstrates that the racial disparity is underestimated by 44% if the correction is not performed. In addition, the disparity between black and white women starts at an early age and increases with age, so the oldest black women are the most at risk to die of this disease. However, a trend analysis of the corrected cervical cancer mortality rate does suggest that the mortality gap may in fact be closing because black women have a larger annual percentage decrease in mortality in comparison with white women after correction.

Comparing the corrected age-standardized morality rates of black and white women living in the United States with the global mortality rates of women helps to put the significance of the racial disparity into perspective. The corrected overall mortality rate for white women of 3.4 per 100,000 is equivalent to the 2012 GLOBOCAN estimate of the cervical cancer mortality rate for women living in developed nations (ie, Europe, Northern America, Australia/New Zealand, and Japan) of 3.3 per 100,000. However, the corrected estimate for black women living in the United States of 10.1 per 100,000 is on par with the GLOBOCAN estimate of 9.8 per 100,000 for less developed nations (ie, all of Africa, Asia [excluding Japan], Latin America, and the Caribbean). In fact, the corrected mortality rates for black women in the current study are as high as those seen in sub-Saharan Africa.18 The current study adds to the body of literature supporting the periodic recalculation of national cervical cancer incidence and mortality rates to better understand the true racial disparity and changes over time.2, 3, 8 These data should also be taken into consideration when one is calibrating mathematical models of the natural history of cervical cancer.

The current study complements previously reported hysterectomy-corrected age-specific cervical cancer incidence rates, which were calculated with the same data sources and methodology. Incidence rates for black women were disproportionately increased by correction for hysterectomy in comparison with white women. The highest corrected incidence rates were also found in the oldest black women.2 Together, these analyses of incidence and mortality provide a complete and contemporary picture of cervical cancer in at-risk women in the United States. Similar analyses have demonstrated increased uterine cancer incidence rates after correction for the prevalence of hysterectomy, and they suggest that the correction for hysterectomy is necessary when one is reporting racial disparities in US cancer rates.19 The disparity in cervical cancer incidence has been hypothesized to be related to a lack of access to screening; however, several studies have demonstrated equivalent screening rates in black and white women.20, 21 A differential loss to follow-up after abnormal Papanicolaou test results and differences in tumor biology may also contribute to the disparities in cervical cancer incidence and mortality.22-24

There are likely additional factors that contribute specifically to the mortality disparity noted in the current study. A differential distribution of histologic subtypes by age and race may explain some of the difference in mortality: rates of adenocarcinoma in black women increase with increasing age, whereas they plateau at the age of 35 years in white women, and adenocarcinoma has a higher rate of death from disease in comparison with squamous cell carcinoma in both early- and late-stage cervical cancer.25, 26 Differences in treatment likely also play a role: black women are more likely to present at a later stage of diagnosis and may receive different treatment after diagnosis in comparison with white women.27-29 An evaluation of Maryland Cancer Registry data found that black women had 50% lower odds of undergoing surgery and 50% higher odds of receiving radiation in comparison with white women with the same stage and insurance.30 In addition, in an evaluation of SEER data, an age greater than 70 years and black race were both independently associated with a lower likelihood of receiving complete surgical staging and appropriate adjuvant radiation (if the primary treatment was surgery) and with a lower likelihood of receiving complete curative-intent radiation (if the primary treatment was radiation).31 This is concordant with the current study's finding that the oldest black women have the highest cervical cancer mortality rates and supports research to gain an understanding of why this unique population receives differential treatment for the same disease.

There are several considerations when the BRFSS is being used to estimate the prevalence of hysterectomy. First, the estimate of the hysterectomy prevalence is based on self-reporting. Although prior studies have demonstrated self-reporting as a valid method of assessing a woman's hysterectomy status,32, 33 the BRFSS does not distinguish between a complete hysterectomy (ie, removal of the uterus and cervix) and a supracervical hysterectomy (in which the cervix is left in situ). This would likely minimally overestimate the corrected mortality rates because several studies before 2004 estimated the prevalence of supracervical hysterectomy at <2%, whereas recent data suggest that the prevalence is 4% to 7.5%.3, 34, 35 Second, the methodology of BRFSS data collection changed in the 2011 and 2012 data sets: cellphone-only household data were included, and a more robust weighting method (or raking) was applied. This may have resulted in higher state-level hysterectomy prevalence estimates than those in 2002-2010, but the hysterectomy prevalence estimates may also be more accurate for minority populations because of the increased reach of the survey in the latter years.36 However, Joinpoint regression failed to show significant trends in hysterectomy prevalence by race from 2002 to 2012, so the overall trends are likely still accurate.

In addition, the current study provides estimates from the states represented in SEER-18 (excluding Louisiana) and does not include data from all 50 states. However, a major strength of the study was the ability to obtain individual-level data for these states to provide the most accurate composite estimates. As emphasized in Temkin et al,19 the hysterectomy prevalence differs by time, age, race, and geographic location; our estimates were stratified by these factors to minimize errors in our mortality estimates. In contrast, solely using registry data or using less well-stratified data would have likely resulted in inaccurate rates. Another important consideration is that an analysis by histologic subtype could not be performed because these data were not available in the NCHS mortality database.

In conclusion, our study illustrates the impact of correction for hysterectomy prevalence on cervical cancer mortality rates in black and white women in the United States. This correction reveals a larger racial disparity, with the most profound effect seen in the rates of the oldest black women.2 Given that both older age and black race have been shown to be independent predictors of inadequate treatment after diagnosis, clinicians must be aware of the disparity to help to ensure equal treatment in the future.31 Although corrected rates demonstrate a decline in mortality rates in black women over time, overall, the rate of death is still twice that for white women; this is a public health disparity that cannot be ignored. Future research should focus on overcoming the underlying factors that contribute to this mortality gap (differential follow-up for abnormal Papanicolaou test results and barriers to care after diagnosis) and on determining why black women and older women receive different and inadequate treatment for the same disease.

FUNDING SUPPORT

This work was made possible in part by a research grant from the Cigarette Restitution Fund (principal investigator Anne F. Rositch) to the Johns Hopkins Medical Institutions (fiscal year 2016).

CONFLICT OF INTEREST DISCLOSURES

Patti E. Gravitt reports reagent and travel support from Hologic for research not related to the current study.

AUTHOR CONTRIBUTIONS

Anna L. Beavis: Conceptualization, methodology, software, formal analysis, investigation, resources, data curation, writing–original draft, writing–review and editing, and visualization. Patti E. Gravitt: Conceptualization, methodology, software, investigation, resources, data curation, writing–review and editing, visualization, and supervision. Anne F. Rositch: Conceptualization, methodology, software, formal analysis, investigation, resources, data curation, writing–original draft, writing–review and editing, visualization, supervision, project administration, and funding acquisition.