Long-term Patterns of Excess Mortality among Endometrial Cancer Survivors

This article is featured in Highlights of This Issue, p. 1033

Endometrial cancer is the fourth most commonly diagnosed cancer among women in the United States, with more than 65,000 new cases estimated in the year 2020 (1). Fortunately, 5-year survival is high for patients with endometrial cancer overall, at more than 80% for all stages combined (1), and recent data project that the number of endometrial cancer survivors in the United States will grow from approximately 800,000 in 2019 to just more than 1 million by 2030 (2). With continued growth in the survivor population, there is a corresponding need for additional survivorship research to guide the long-term care of women with an endometrial cancer history.

Using standard survival curves, estimates of 5-year survival reflect a patient's probability of surviving for 5 years beyond the date of a cancer diagnosis. While useful for recently diagnosed patients, these estimates are clearly less relevant for patients who have already lived for several years after diagnosis because prognosis is generally expected to improve with each additional year survived. Thus, for mid- to long-term survivors, estimates of conditional survival, which account for the length of time already survived, are more useful measures, and provide indicators of prognosis relevant to specific stages of survivorship (3, 4). Examining conditional survival among cancer survivors relative to expected survival among similar groups in the general population (i.e., conditional relative survival) can reveal excess mortality remaining among survivors within specific time windows after diagnosis. Likewise, standardized mortality ratios (SMR), or ratios of observed mortality in a cancer cohort to expected mortality in the general population, are another measure that can be used to quantify excess mortality, from all causes, as well as from specific causes, among cancer survivors. A comprehensive examination of excess mortality according to time since an endometrial cancer diagnosis, using SMRs and conditional relative survival estimates, could inform planning for surveillance and follow-up care in the years after initial treatment, but to our knowledge, this has not been reported for U.S. endometrial cancer survivors.

The objective of this study was to estimate 5-year relative survival among U.S. women with endometrial cancer at the time of diagnosis and at each additional year survived (conditional relative survival) up to 12 years after diagnosis. We also estimated SMRs for all-cause and cause-specific mortality according to time since diagnosis to characterize how long-term patterns of mortality among endometrial cancer survivors compare with those among demographically similar women in general U.S. population. Analyses were performed overall and according to demographic and tumor characteristics.

Women with an endometrial cancer diagnosis were identified using data from the Surveillance Epidemiology and End Results (SEER) program (5, 6), a system of population-based cancer registries which collects and reports data on cancer incidence and survival and covers approximately 35% of the U.S. population (7). Information available in the SEER research database includes patient demographics, primary tumor site and morphology, stage, and vital status. For deceased patients, SEER recodes International Classification of Diseases (ICD) codes from the death certificate and reports cause of death in major groupings (8). Mortality data for the general U.S. population are accessible through the SEER database and come from the National Center for Health Statistics. This study was considered exempt by the University of North Carolina (Chapel Hill, NC) Institutional Review Board.

From the SEER 18 registries, we identified women with a first malignant primary endometrial cancer (sites C54.0–C54.9 and C55.9; ref. 9) between 2000 and 2017. We excluded death certificate or autopsy-only cases, those who were younger than 15 years at diagnosis, and those with missing information on race. In analyses of conditional 5-year relative survival, we also excluded those diagnosed after 2012, to allow a minimum of 5 years survival data through the end of follow-up for vital status on December 31, 2017. We used the following ICD-O-3 codes to define histologic subtypes as endometrioid: 8140, 8210, 8260, 8262, 8380–8384, 8440, 8480–8482, 8560, and 8570; serous: 8441, 8450, and 8460–8461; carcinosarcoma: 8950–8951 and 8980–8981; clear cell: 8310 and 8313; and mixed: 8255 and 8323 (10). All other codes were categorized together as other histologies.

We estimated 5-year relative survival among women with endometrial cancer at diagnosis and at each additional year survived up to 12 years after diagnosis, conditional on being alive at the beginning of that year. Relative survival was calculated as the ratio of observed survival among women with endometrial cancer to expected survival among women in the general U.S. population with a similar distribution of age, race, and calendar year. Survival was calculated using the actuarial method. Expected survival tables for the general population were generated using the Ederer II method. We considered the years at which conditional relative survival exceeded 90% and 95% to reflect little and minimal excess mortality, respectively, among endometrial cancer survivors compared with the general population (3, 4).

SMRs were estimated as the number of observed deaths among women with endometrial cancer divided by the number of expected deaths in the general population. The number of expected deaths was calculated as the product of the person-time at risk in the endometrial cancer cohort and the mortality rate for women in the general population with the same distribution of age, race (White, Black, and other), and calendar year. Confidence intervals (CI) for all SMRs were produced using exact methods. SMRs were estimated for all-cause mortality and for cause-specific mortality from endometrial cancer, other cancers, cardiovascular diseases (CVD: diseases of the heart; hypertension without heart disease; cerebrovascular diseases; atherosclerosis; aortic aneurysm and dissection; and other diseases of arteries, arterioles, and capillaries), and other causes (8). We also report absolute excess risks (AER), calculated as the difference between observed and expected deaths divided by the total person-years of observation, and expressed per 10,000 person-years. SMRs and AERs were estimated for the total study period and within the following time intervals: diagnosis–<1 year, 1–<5 years, 5–<10 years, and 10+ years post-diagnosis. Subgroup analyses were performed according to race, age at diagnosis, disease stage, histology, and grade. All analyses were performed using SEER*Stat, version 8.3.6.1.

A total of 121,273 women, diagnosed with endometrial cancer during 2000–2012, contributed to analyses of conditional relative survival. Overall, 5-year conditional relative survival was 81.6% (95% CI, 81.4–81.9) at diagnosis and increased consistently to 87.7% (95% CI, 87.7–88.0), 96.2% (95% CI, 95.9–96.5), and 97.1% (95% CI, 96.5–97.6), respectively, at 1, 5, and 10 years post-diagnosis (Table 1; Fig. 1). Conditional relative survival first exceeded 95%, reflecting minimal excess mortality compared with the general population, at 4 years after diagnosis.

The year at which minimal excess mortality was reached varied considerably according to patient demographic and cancer-related characteristics. Among White women, relative survival was >95% by 4 years after diagnosis, compared with 8 years among Black women and 6 years among women of other races. Survival estimates were consistently somewhat higher for women who were younger at diagnosis, exceeding 95% at 4 years among those ages 15–64 years, and 5 years among those ages 65 years and older. While women with localized stage disease had minimal excess mortality at diagnosis and consistently thereafter, those with more advanced-stage disease did not surpass 95% relative survival by 12 years post-diagnosis; at 10 years, estimates were 91.3% (95% CI, 89.7–92.8) and 87.3% (95% CI, 81.6–91.3) among those with regional and distant stage disease, respectively. Likewise, throughout follow-up, relative survival remained consistently higher, and >95% was achieved earlier, for those with lower grade disease.

Patterns of conditional relative survival also varied according to histology; minimal excess mortality was observed as early as year 3 for those with endometrioid histology, but was observed much later or not within the study period for those with serous, carcinosarcoma, clear cell, mixed, or other histologies (Table 1). In analyses according to race stratified by stage, histology, and grade, Black women tended to reach little or minimal excess mortality later than White or other race women with similar disease characteristics (Supplementary Table S1). For example, among women with localized stage disease, 95% relative survival was first exceeded at diagnosis and 2 years for White women and women of other races, respectively, but not until 6 years for Black women.

SMR and AER analyses included a total of 183,153 women diagnosed with endometrial cancer between 2000 and 2017. Overall, the SMR for all-cause mortality decreased over time, from 5.90 (95% CI, 5.81–5.99) in the first year after diagnosis to 2.76 (95% CI, 2.72–2.79) and 1.30 (95% CI, 1.28–1.33) at 1–<5 years and 5–<10 years, respectively, but remained significantly elevated at 10+ years post-diagnosis (SMR, 1.16; 95% CI, 1.13–1.19; Table 2). In general, SMRs declined over time within all subgroups, but tended to be higher for those who were Black or other race, younger at diagnosis, had higher stage disease, and had nonendometrioid histologies. However, even at 10+ years, those with localized stage disease, grade 1 disease, and endometrioid histology had a small, but significant increase in all-cause mortality compared with the general population. AERs followed similar patterns for cancer-related characteristics, but for demographic characteristics, AERs were higher for older women, rather than younger, and were much higher for Black women than either White women or those of other races. Patterns according to race observed in overall analyses, with higher SMRs for all-cause mortality among Black and other race women and the highest AERs among Black women, were also generally apparent within subgroups defined by stage, histology, and grade (Table 3).

Findings for cause-specific mortality, overall and according to race, are shown in Table 4. Overall, the SMR for endometrial cancer–specific mortality declined over time, but was still significantly elevated at 10+ years post-diagnosis (SMR, 10.37; 95% CI, 9.24–11.59; Table 4). SMRs for mortality from other cancers also declined consistently over time, from 3.93 (95% CI, 3.78–4.08) between diagnosis and <1 year, to 1.06 (95% CI, 1.00–1.13) at 10+ years. In contrast, mortality from CVDs and other causes (noncancer and non-CVD) was most elevated during the year after diagnosis, slightly elevated at 10+ years, and either significantly lower than or similar to the general population between 1 and <10 years. Although the number of deaths from CVDs and other causes exceeded the number of deaths from endometrial cancer at 5–<10 years and 10+ years, the AER for the full cohort was highest for endometrial cancer–related deaths within all time periods. In analyses according to race, SMRs for endometrial cancer were highest for women of other races during all time periods, but AERs were generally highest for Black women. For other cancers, CVDs, and other causes, SMRs were consistently higher for Black women and women of other races than White women, and AERs were nearly always highest for Black women. SMRs and AERs for cause-specific mortality according to age at diagnosis, stage, histology, and grade are shown in Supplementary Tables S2–S5. Although patterns varied somewhat according to cause of death and time since diagnosis, SMRs and AERs tended to be higher for women with more advanced-stage or higher grade disease and for those with nonendometrioid histologies.

In this registry-based study, we estimated conditional 5-year relative survival up to 12 years after an endometrial cancer diagnosis and examined long-term patterns of excess mortality among endometrial cancer survivors according to demographic and cancer-related characteristics. As expected, relative survival increased with each additional year survived, and overall, exceeded 95% by 4 years after diagnosis. However, among the full cohort and within all subgroups, relative survival was still significantly below 100%, indicating some remaining elevation in mortality compared with the general population, at 10 years post-diagnosis. SMR and AER analyses further demonstrated that excess mortality, from all causes and from specific causes, compared with demographically similar women in the general U.S. population, tended to be greater among non-White women and those with less favorable disease characteristics, even at 10+ years after endometrial cancer diagnosis.

Conditional survival estimates provide valuable information for cancer survivors who are well beyond their initial diagnosis and treatment period, but remained concerned about the impact of their cancer history on their future mortality risk. Our analyses suggested that >95% relative survival, which we considered to reflect minimal excess mortality, was achieved relatively quickly by endometrial cancer survivors overall, at 4 years after diagnosis. However, this was largely driven by women with more favorable disease characteristics, namely those with localized and grade 1 disease, whose relative survival exceeded 95% at diagnosis and consistently thereafter. In contrast, among women with regional or distant stage disease, undifferentiated disease, and nonendometrioid histologies, relative survival increased over time since diagnosis, but did not reach 95% within the study period of up to 12 years post-diagnosis. Understanding which subgroups of survivors, defined by demographic and cancer-related characteristics, continue to have lower than expected survival for many years after cancer treatment can help in predicting the type and intensity of care that will be needed across various phases of survivorship.

In addition to estimating conditional relative survival, we also used SMRs and AERs to quantify excess deaths from all causes and specific causes among women with endometrial cancer within specified post-diagnosis time windows. Our findings suggested that even at 5–<10 and 10+ years post-diagnosis, the greatest contributor to excess mortality relative to the general population was still death from endometrial cancer. However, it was notable that certain subgroups, particularly Black women and those with more advanced-stage or higher grade disease, had excess deaths attributable to other cancers, CVDs, and other causes within certain post-diagnosis time periods. These findings underscore the importance of long-term follow-up and monitoring of women with an endometrial cancer history, particularly for Black women and those whose initial prognosis was less favorable.

Associations between obesity and endometrial cancer incidence (11) suggest that endometrial cancer survivors may have elevated rates of CVD incidence and mortality relative to the general population. Women with more advanced-stage disease, although they comprise a minority of all patients with endometrial cancer, may also be treated with certain chemotherapeutic agents that may have cardiotoxic effects and could also contribute to future risk of adverse cardiovascular outcomes (12, 13). A previous study using SEER data reported that women diagnosed with endometrial cancer between 1988 and 2012 were 8.8 (95% CI, 8.7–9.0) times more likely to die from CVDs than women in the general population (14). Our analyses suggested a smaller, although still significant increase in CVD mortality, which was most apparent among endometrial cancer survivors who were non-White, younger, or had more advanced-stage disease. We also found that the magnitude of the SMR was not consistent across time periods, with greater elevations in CVD mortality within the first year after endometrial cancer diagnosis and at 10+ years post-diagnosis. It is unclear the extent to which excess mortality from CVDs within the year after diagnosis reflects a direct impact of endometrial cancer diagnosis and treatment on CVD-related deaths, or misattribution of cancer-related death to CVDs. Nevertheless, these results suggest the importance of monitoring cardiovascular health during the initial cancer diagnosis and treatment period. Excess CVD mortality among longer term endometrial cancer survivors in our study also suggests that CVD prevention efforts should begin early in follow-up care. Neither the earlier SEER report nor ours was able to account for CVD risk factors, such as obesity and diabetes, or specific cancer treatments, because this information is not available in SEER. Future studies may be warranted to investigate the impact of these factors on CVD outcomes among endometrial cancer survivors, and why risk relative to the general population may vary according to time since endometrial cancer diagnosis.

Prior reports have documented pronounced racial disparities in endometrial cancer outcomes, with lower 5-year survival among Black women than White women that is not fully explained by different distributions of stage, grade, or histologic subtype by race (15–17). Our findings add information on the extent to which these disparities persist among longer term survivors. Overall and in every subgroup defined by disease characteristics, conditional relative survival among Black women increased steadily over time since diagnosis, but remained slightly lower than that of White women at 12 years, and >95% relative survival was reached later among Black women than White women. Likewise, in all post-diagnosis time windows up to 10+ years, both SMRs and AERs for all-cause mortality were consistently higher for Black women than for White women, even among those with localized stage or lower grade disease, and they tended to be higher for cause-specific mortality as well. Calculation of relative survival and SMRs accounts for race and, when stratified by race, estimates, therefore, reflect excess mortality among endometrial cancer survivors compared with women of the same race in the general population. Persistently lower conditional relative survival and higher SMRs long after diagnosis suggest a greater and more lasting impact of an endometrial cancer diagnosis on mortality among Black women, relative to their cancer-free peers, than among White women, and the need for efforts to reduce disparities not just among recently diagnosed patients, but also among long-term survivors.

Our study has several strengths and limitations. Use of the SEER database allowed for a large sample size and examination of long-term patterns of mortality according to basic demographic and disease-related characteristics. However, SEER data lack information on cancer recurrence, and information on first course of cancer treatment is thought to be fairly incomplete (18), so we were unable to consider these factors in our analyses. For some cancer characteristics, such as grade, a relatively high proportion of patients had missing information. We also did not have information on factors such as comorbidities, income, or obesity, all of which may be associated with patterns of mortality after endometrial cancer diagnosis. In addition, because all of our analyses involved comparisons with the general U.S. population, we were limited in our stratified analyses to only those factors accounted for by the U.S. population mortality statistics (e.g., age, sex, and race) used in this study. Cause-specific mortality analyses are also subject to potential misclassification due to inaccurate coding of cause of death on death certificates. Finally, race information in the SEER registries comes from patient medical records, and misclassification could occur if the race indicated in the medical record does not match the woman's identity or experience. Our analyses by race also do not account for diversity within racial categories. Nevertheless, our findings provide insight into how excess mortality among endometrial cancer survivors varies according to patient characteristics and time since cancer diagnosis, and may inform planning for follow-up care throughout survivorship.

Results of this study suggest that overall, endometrial cancer survivors have only a small, although significant, survival deficit beyond 4 years post-diagnosis. However, excess mortality was greater in magnitude and persisted longer into survivorship for Black women and those with more advanced-stage or higher grade disease. Strategies to mitigate disparities in mortality after endometrial cancer will be needed as the number of endometrial cancer survivors in the United States continues to increase.

No disclosures were reported.

C. Anderson: Formal analysis, writing–original draft, writing–review and editing. V.L. Bae-Jump: Writing–review and editing. R.R. Broaddus: Writing–review and editing. A.F. Olshan: Writing–review and editing. H.B. Nichols: Writing–review and editing.

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.