- Email: [email protected]
Declining trends of epithelial ovarian cancer in California
Available online at www.sciencedirect.com
Gynecologic Oncology 108 (2008) 207 – 213 www.elsevier.com/locate/ygyno
Declining trends of epithelial ovarian cancer in California Cyllene R. Morris a,⁎, Anne O. Rodriguez b , Joan Epstein c , Rosemary D. Cress a,d a
b
d
California Cancer Registry, Public Health Institute, Sacramento, California, USA Division of Gynecologic Oncology, University of California Davis Medical Center, Sacramento, California, USA c Survey Research Group, California Department of Health Services, Sacramento, California, USA Department of Public Health Sciences, Division of Epidemiology, University of California Davis, California, USA Received 25 April 2007 Available online 29 October 2007
Abstract Objective. The purpose of this study was to examine trends of ovarian cancer in California. Known risk factors for ovarian cancer are unequally distributed across different cultures. California, the most ethnically diverse State in the U.S., is uniquely positioned to detect trends in different race/ethnic groups. Methods. This study was based on ovarian cancer cases and deaths occurring between 1990 and 2003 and reported to the California Cancer Registry (CCR). Age-adjusted rates were calculated by race/ethnicity, age at diagnosis, tumor behavior, and morphology. Trends were evaluated using the annual percent change, estimated through joinpoint regression. Results. The study population included 30,669 invasive ovarian cancers, 5925 tumors of borderline malignancy, and 19,719 ovarian cancer deaths. Age-adjusted incidence rates of invasive ovarian cancer per 100,000 women decreased from 15.6 in 1990 to 12.9 in 2003, a statistically significant decline of 1.3% per year. Rates of epithelial malignancies declined in all age groups examined, with the largest decreases noticed among women 65–74 years old. Substantial differences in trends were detected by race/ethnicity. Epithelial malignancies declined significantly among non-Hispanic black and white women by 2.5% and 1.2% per year, respectively. Changes among Asian/Pacific Islander and Hispanic women were smaller and non-significant. Ovarian cancer mortality rates did not change significantly during the study period. Conclusions. The incidence of epithelial ovarian cancer in California declined significantly among all age groups examined and among nonHispanic white and black women. Among Hispanic and Asian/Pacific Islander women, declines were small and non-significant © 2007 Elsevier Inc. All rights reserved. Keywords: Ovarian cancer; Incidence; Race; Ethnicity
Introduction Ovarian cancer is the sixth most commonly diagnosed cancer in women in the United States. In the absence of effective screening methods, the great majority of malignant ovarian neoplasms in the U.S. are detected at advanced stages of disease, when patients present with more obvious symptoms and face poorer prognosis [1]. Data from the National Cancer Institute's Surveillance, Epidemiology and End Results (SEER) program show that, overall, ovarian cancer incidence rates in the U.S. have declined by 0.9% per year since the late 1980s [1], although little is known about incidence trends in different race/ethnic groups. Declines also have ⁎ Corresponding author. 1700 Tribute Rd, Suite 100, Sacramento, CA 958154402, USA. Fax: +1 916 779 2608. E-mail address: [email protected] (C.R. Morris). 0090-8258/$ - see front matter © 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.ygyno.2007.09.026
been observed in northern European countries [2]. The purpose of the present study was to examine trends of ovarian cancer incidence and mortality among California women. A culturally diverse and multiethnic state, California has the country's second largest Hispanic and Asian/Pacific Islander populations (http:// www.census.gov/population/cen2000/phc-t6). Because the distribution of risk factors for ovarian cancer is most likely heterogeneous across different cultures, California is uniquely positioned to detect differences in temporal trends by race/ ethnicity. Materials and methods Data sources Ovarian cancer incidence data were obtained from the CCR, California's statewide population-based cancer surveillance system. Since 1988, all
208
C.R. Morris et al. / Gynecologic Oncology 108 (2008) 207–213
hospitals, physicians and laboratories in California have been mandated by law to report every cancer diagnosis, with the exception of basal and squamous cell carcinomas of the skin and carcinoma in situ of the cervix. For every newly diagnosed cancer, the CCR gathers information on tumor characteristics, diagnosis, treatment, and follow-up (plus patient's demographic information). Data are collected through a network of ten regional registries, which also participate in the National Cancer Institute's SEER Program. As of April 2006 statewide case ascertainment for 2003 was estimated to be complete. Data on ovarian cancer-related deaths were obtained from the death certificate master files, Center for Health Statistics, California Department of Health Services, while population estimates by age and race/ethnicity were obtained from the U.S. Census Bureau.
Ovarian cancer cases and deaths This study included 30,669 invasive ovarian cancers diagnosed in California from January 1990 through December 2003 and reported to the CCR as of April 2006 (Table 1). Incidence trends of 5925 ovarian tumors of borderline
Table 1 Frequency distribution of ovarian cancers diagnosed in California from 1990 through 2003 by demographic and tumor-related characteristics Characteristic
Number of cases
5925 30,669
16.2 83.8
Invasive tumors (n = 30,669) Tumor morphology Epithelial Germ cell Sex cord–stromal Other
28,365 975 356 973
92.4 3.2 1.2 3.2
a
Race/Ethnicity
%
All tumors (N = 36,594) Tumor behavior Borderline malignancy Invasive
Invasive epithelial tumors (N = 28,365) Age at diagnosis b30 30–49 50–64 65–74 75+ Race/Ethnicity a White Black Hispanic Asian/Pacific Islander Unknown Tumor grade b G1 G2 G3 G4 Unknown Stage at diagnosis (N = 20,425) c IA–IB IC–II III–IV Unstaged
malignancy are also presented. CCR data on anatomic site and tumor morphology were abstracted from the patient's medical records and pathology reports. Tumors with International Classification of Diseases for Oncology, Third Edition (ICD-O-3) morphology codes 8010-8570 (excluding 8240-8245), 9014-9015, or 9110 were grouped as epithelial tumors [3]. Tumors for which no morphologic descriptors are found when the information is abstracted are coded as not otherwise specified malignancies (NOS, ICD-O-3 codes 8000-8004). Because the vast majority of ovarian malignancies are epithelial in origin, NOS tumors were included in the epithelial category. Stage at diagnosis was based on extent of disease codes (available for cases diagnosed in 1994 forward) converted to the American Joint Committee on Cancer (AJCC) stage categories [4]. AJCC stages were subsequently grouped as IA–IB, IC–II, III–IV, and unstaged (Table 1). Tumor grade, or degree of differentiation of neoplastic tissue, was coded as G1 (well differentiated), G2 (moderately differentiated), G3 (poorly differentiated), or G4 (undifferentiated or anaplastic). Mortality data (Death Master Files) were provided by the California Department of Health Services Center for Health Statistics. A total of 19,719 deaths where ovarian cancer was identified as the underlying cause of death in the death certificate were included in the study. Deaths occurring between 1990 and 1998 were coded by the International Classification of Diseases (ICD), Ninth Edition, while deaths occurring from 1999 forward were coded by the ICD, Tenth Edition.
369 5296 8503 7009 7188
1.3 18.7 30.0 24.7 25.3
20,919 1165 3922 2223 136
73.7 4.1 13.8 7.8 0.5
2264 5122 10,223 1933 8823
8.0 18.1 36.0 6.8 31.1
2559 2941 13,287 1638
12.5 14.4 65.1 8.0
Race/Ethnic groups are mutually exclusive; women of Hispanic origin may be of any race. b Tumor grade: G1 (well differentiated), G2 (moderately differentiated), G3 (poorly differentiated), and G4 (undifferentiated). c Stage at diagnosis based on extent of disease codes (available for cases diagnosed in 1994 forward) converted to American Joint Committee on Cancer (AJCC) categories.
Race/Ethnicity for cases reported to the CCR was determined from information in the medical record or death certificate. Incidence and mortality figures were compiled for four major race/ethnicity groups: non-Hispanic white (white), Hispanic, non-Hispanic Asian/Pacific Islander (Asian/PI), and non-Hispanic black (black). To determine Hispanic ethnicity more accurately, information from the medical record was supplemented by linking the subject's last name or maiden name to the 1980 US Census list of Hispanic surnames [5]. Patients identified either by their medical record or their surnames as Hispanic were classified as Hispanic.
Adjustment for completeness of case reporting Incidence rates are calculated by the CCR when reporting of new cases is estimated to be at least 95% complete, which usually happens within 18 to 24 months after the end of a diagnosis year. However, cancer surveillance is a dynamic process and cases diagnosed at earlier years may be reported long after incidence data are considered “complete”. An internal review of ovarian cancer incidence rates reported in CCR annual publications revealed that rates for a given year of diagnosis increased annually up to 8 years after the initial release and stabilized after that (data not shown). For example, the 1991 incidence rate of ovarian cancer reported by the CCR in early 1994 was 14.5 cases per 100,000 women. By 2000, the rate for 1991 had increased to 15.0 per 100,000 due to additional cases being reported over time. In this internal review of former CCR reports, the average increase in ovarian cancer rates 1 year after the first release was 1.5%. Compared to the initial estimate, rates for the same year were about 4% higher after 8 years of surveillance. In order to minimize the effect of incomplete reporting, incidence rates were multiplied by a factor corresponding to the percent increase observed in CCR rates during the 8 years lapsed until estimates were stable. As of April 2006, rates from 1990 through 1995 were considered stable, while rates from 1996 through 2003 were inflated by 0.2%, 0.4%, 0.5%, 0.7%, 1.2%, 1.6%, 2.5%, and 4.0%, respectively. Multiplication factors were applied regardless of tumor behavior (invasive or borderline malignancy), morphology, race/ethnicity, or age at diagnosis.
Statistical analysis Frequency distributions of demographic (age at diagnosis and race/ethnicity) and tumor-related characteristics (tumor behavior, morphology, and stage at diagnosis) were obtained for all incident ovarian cancer cases. Incidence and mortality rates per 100,000 women were calculated for groups with at least 15 cases (or deaths) and were age-adjusted to the 2000 U.S. standard population. The Surveillance Research Program, National Cancer Institute SEER⁎Stat software version 6.2.3 (http://srab.cancer.gov/seerstat) was used to calculate all rates and frequency distributions. Joinpoint linear regression [6] was used to fit the log of
C.R. Morris et al. / Gynecologic Oncology 108 (2008) 207–213
209
age-adjusted rates. In this analysis, a statistical algorithm detects joinpoints, i.e., points in time where the slope of the regression line significantly changes. Thus, the model describes trends during different time segments. At each segment, trends in rates are measured using the estimated annual percent change (EAPC), which assumes that rates change by a constant percentage each year. The EAPC was calculated using the weighted least squares method, where a linear regression to the natural logarithm of the age-adjusted annual rates was fitted with year of diagnosis as the regressor variable. The estimated slope of the regression line m was then used to calculate the EAPC as equal to 100 ⁎ (em − 1). The SEER JoinPoint regression software version 3.0 (http://srab.cancer.gov/joinpoint) was used for all trend analyses in this study. Regression models were fit using weighted least squares, with default program parameters set. The significance of trends was evaluated with a threshold of 0.05. Incidence trends were evaluated by tumor behavior (invasive or borderline malignancy) and morphology (epithelial, germ cell, sex cord–stromal, and other). Trends for invasive epithelial tumors were further evaluated by tumor grade, age at diagnosis, and race/ethnicity. Two sets of estimates were obtained, with and without adjusting for completeness of reporting as described above. Rates for stage at diagnosis were not calculated, but chi-square tests for trend were performed to detect potential shifts over time in the distribution of cases diagnosed at earlier and late stages, or which were not staged. Trends in ovarian cancer mortality were evaluated by race/ethnicity and age at death only.
Results Characteristics of the study population Demographic and tumor-related characteristics of ovarian cancer patients in this study are shown in Table 1. Malignant neoplasms accounted for 83.7% of all ovarian cancers diagnosed in California from 1990 through 2003, while 16.1% of tumors were of borderline malignancy. Of the 30,669 invasive tumors, 92.4% were of epithelial origin, with germ cell, sex cord–stromal, and tumors of other histologies being diagnosed in 3.2%, 1.2%, and 3.2% of cases, respectively. Tumor grade was considered poorly differentiated in 36% of cases followed by moderately differentiated (18.1%), well differentiated (8%), and undifferentiated tumors (6.8%). Information on tumor grade was absent from patient records in 31.1% of cases. The majority of patients (65.1%) were diagnosed with a late-stage tumor (AJCC III or IV). A chi-square analysis of trend showed that, between 1994 and 2003, there were no significant changes in the proportion of cases diagnosed at localized (IA–IB) or regional (IC–II) stages. There was, however, a slight but significant (p = 0.04) increase in the proportion of cases diagnosed with late (III–IV) stage tumors, from 64.2% in 1994 to 66.8% in 2003. A significant decrease in the proportion of unstaged cases, from 9.3% to 7.1%, accounted for most of the observed shift. These patterns were similar in all age groups examined (data not shown). The proportion of tumors classified from well to poorly differentiated decreased significantly during the period. Such a decrease was associated with a very marked and significant increase in the proportion of undifferentiated tumors, from 5.4% of cases in 1990 to 8.6% in 2003 (p b 0.000). The proportion of tumors with unknown grade also increased, from 29.7% in 1990 to 36.9% in 2003. Trends in ovarian cancer incidence During the 14-year period evaluated in this study, invasive ovarian cancer age-adjusted rates per 100,000 women in California decreased from 15.6 in 1990 to 12.9 in 2003 (Fig. 1),
Fig. 1. Ovarian cancer age-adjusted incidence rates by tumor behavior, California, 1990–2003: observed and completeness-adjusted rates. Rates in the figure are per 100,000 women and age-adjusted to the 2000 U.S. standard population. To further adjust for completeness of reporting, observed age-adjusted rates were multiplied by factors representing average increases detected over time in ovarian cancer rates.
which corresponds to a statistically significant decline of 1.3% per year (95% CI for EAPC: −1.5, −1.1) (Table 2). Rates per 100,000 women in California for tumors of borderline malignancy, however, increased from 2.2 in 1990 to 2.9 in 2003, an also statistically significant change of 1.8% per year (95% CI for EAPC: 0.9, 2.7). The great majority of invasive ovarian tumors (92.4%) were of epithelial origin (Table 1). Accordingly, trends for epithelial tumors (EAPC = −1.4%, 95% CI: −1.6, −1.2) were similar to those for all invasive tumors, with rates per 100,000 women decreasing from 14.5 in 1990 to 11.9 in 2003. Germ cell tumors accounted for 3.2% of all invasive ovarian cancers, and their incidence trends were better described in two time periods. During the first period, from 1990 through 1998, rates declined significantly by 3.5% per year (95% CI for EAPC: −5.8, −1.2), while the second period was characterized by a significant and steep increase of 5.8% per year (95% CI for EAPC: 0.8, 11.1). A non-significant decline of 2.8% per year (95% CI for EAPC = −8.7, 3.4) was also detected among tumors of the sex cord– stromal tissue, a less common morphology accounting for 1.2% of malignant ovarian tumors. The incidence of tumors of other histologic types, such as malignant Brenner tumors, carcinosarcomas, and müellerian mixed tumors, remained unchanged during the 1990–2003 time period. Because they represented the vast majority of tumors, subsequent data analysis was limited to epithelial ovarian malignancies. The incidence of epithelial ovarian cancer in California declined
210
C.R. Morris et al. / Gynecologic Oncology 108 (2008) 207–213
Table 2 Estimated annual percent change (EAPC) and 95% confidence intervals (CI) for ovarian cancer incidence, California, 1990–2003: estimates based on actual rates and on rates adjusted a for completeness of reporting Characteristic
All tumors (N = 36,659) Tumor behavior Invasive Borderline
EAPC 95% CI
Adjusteda
95% CI
− 1.3 ⁎ 1.8⁎
− 1.5, −1.1 0.9, 2.7
− 1.1⁎ 1.6⁎
− 1.3, − 0.9 0.6, 2.6
− 1.6, −1.2
− 1.2⁎
− 1.4, − 0.9
− 5.8, −1.2 0.8, 11.1 − 8.7, 3.4 − 1.3, 2.2
− 3.5⁎ 5.8⁎ − 2.8 0.4
− 5.8, − 1.2 0.8, 11.1 − 8.7, 3.4 − 1.3, 2.2
Invasive tumors (N = 30,669) Tumor morphology Epithelial − 1.4 Germ cell 1990–1998 − 3.5⁎ 1998–2003 5.8⁎ Sex cord–stromal − 2.8 Other 0.4 Invasive epithelial tumors Age at diagnosis All ages b30 30–49 50–64 65–74 75+ Race/Ethnicity b All race/ethnicities c White Black Hispanic Asian/Pacific Islander
EAPC
Actual
(N = 28,365) − 1.4⁎ − 1.7 − 1.6⁎ − 1.1⁎ − 1.7⁎ − 1.3⁎
− 1.6, −1.2 − 4.0, 0.8 − 2.4, −0.8 − 1.5, −0.7 − 2.4, −1.1 − 2.1, −0.4
− 1.2⁎ − 1.7 − 1.4⁎ − 0.9⁎ − 1.5⁎ − 1.1⁎
− 1.4, − 0.9 − 4.0, 0.8 − 2.2, − 0.6 − 1.3, − 0.5 − 2.2, − 0.8 − 1.9, − 0.2
− 1.4⁎ − 1.2⁎ − 2.5⁎ − 0.4 − 0.7
− 1.6, −1.2 − 1.5, −0.9 − 4.3, −0.8 − 1.3, 0.5 − 1.5, 0.1
− 1.2⁎ − 1.0⁎ − 2.3⁎ − 0.2 − 0.5
− 1.4, − 0.9 − 1.3, − 0.7 − 4.0, − 0.6 − 1.1, 0.7 − 1.3, 0.4
actual CCR rates. Likewise, all other EAPCs were reduced by about 0.2% when adjusting for incomplete reporting. These corrections, however, did not result in loss of statistical significance in any of the groups examined. Trends in ovarian cancer mortality Changes in mortality rates from 1990 through 2003 were not statistically significant (EAPC = −0.5, 95% CI = −1.0, 0.0), from 9.5 deaths per 100,000 in 1990 to 9.1 deaths per 100,000 in 2003. Of all the race/ethnic groups and age categories evaluated, ovarian cancer-related deaths declined significantly only among women who were 50–64 years old at death (EAPC = −1.4, 95% CI = −2.3, −0.5) (Table 3). Discussion Our results show that incidence rates of epithelial ovarian malignancies in California women declined an average of 1.4% a year over the 14-year period 1990 through 2003. These results are consistent with an average 0.9% annual decline in ovarian cancer incidence observed in the U.S. between 1975 and 2003 [1], although rates declined faster in California. Our estimates are not directly comparable to those published by the SEER Program, due to different time periods examined and reporting
a To adjust for completeness of reporting, age-adjusted rates from 1996 forward were multiplied by factors corresponding to the average increase in ovarian cancer rates observed over time by the California Cancer Registry. b Race/Ethnic groups are mutually exclusive; women of Hispanic origin may be of any race. c “All race/ethnicities” category includes women of unknown race or ethnicity. ⁎ The estimated annual percent change is significantly different from zero (p b 0.05).
almost evenly in all age groups examined (Fig. 2), with statistically significant EAPCs varying from −1.1 to −1.7 (Table 2). The largest decrease in rates was noticed among women 65–74 years old, for whom incidence declined by about 22% (EAPC = −1.7, 95% CI= −2.4, −1.1) during the study's 14-year period. When trends were evaluated by race/ethnicity, substantial differences were uncovered (Fig. 3). Epithelial ovarian cancer incidence significantly declined among non-Hispanic black and white women. Rates among black women declined by a steep 52% from 1990 through 2003 (EAPC = −2.5, 95% CI = −4.3, −0.8). Among white women, rates decreased by about 18%, with a significant EAPC of −1.2 (95% CI: −1.5, −0.9). Changes in ovarian cancer incidence rates for Asian/Pacific Islander and Hispanic women were smaller and not statistically significant (Table 2). When rates were adjusted for completeness of reporting, trends were somewhat weakened (Table 2). For example, rates for invasive epithelial tumors still declined significantly over the period, but by 1.2% per year instead of the 1.4% estimated with
Fig. 2. Ovarian cancer age-adjusted incidence rates by age at diagnosis, California, 1990–2003. Rates in the figure are per 100,000 women and ageadjusted to the 2000 U.S. standard population.
C.R. Morris et al. / Gynecologic Oncology 108 (2008) 207–213
Fig. 3. Ovarian cancer age-adjusted incidence rates by race/ethnicity, California, 1990–2003. Rates in the figure are per 100,000 women and age-adjusted to the 2000 U.S. standard population. Race/ethnic groups are mutually exclusive; women of Hispanic origin may be of any race.
delay adjustment. However, the empirical adjustment for completeness employed in this study produced changes in EAPCs of about 0.2%, a magnitude of change similar to SEER trend estimates [1]. In addition to reporting delays, inadequate case finding or underreporting may also contribute to an artificial decline in cancer rates. Although these are recognized problems with cancers diagnosed outside of a hospital, it is
211
highly unlikely that this would be the case with ovarian cancer. Because most ovarian cancers are epithelial in origin, it is expected that epithelial tumors drive overall incidence trends. For other histologic types of ovarian tumors, trends were calculated based on a small number of cases diagnosed per year. For example, rates for germ cell tumors were based on an average of 70 cases diagnosed each year in California, and statistical significance does not necessarily equate to clinical significance. Declines in ovarian cancer incidence also have been documented in the UK and Northern European countries in recent years, particularly among young women, and have been attributed to increased use of oral contraceptives since the 1960s [2,7,8]. Oral contraceptive use has been shown to reduce risk of ovarian cancer [9,10]. Gnagy and colleagues demonstrated a correlation between observed ovarian cancer incidence rates in U.S. women between 1970 and 1995 with those predicted by changes in oral contraceptive use among younger women (aged 30 to 49) [11]. However, the results of their analysis suggested that oral contraceptives did not contribute to a decline in ovarian cancer incidence in older women (aged 50–64) [11]. In contrast to these earlier studies, in the current analysis, the largest declines were observed in women 65 years and older, those with the highest incidence rates. These women, in particular those aged 75 and older, would have had more limited exposure to oral contraceptives as they would have been in their 40s in the late 1960s as oral contraceptives emerged. Tubal ligation and/or hysterectomy also decrease risk [10] and may have played a role in the reduction of incidence among older California women. Hysterectomy rates declined in North America in the 1980s and 1990s, but in the 1950s–1970s, when older women in our study would have been in their reproductive years, hysterectomy rates were still quite high [12,13]. One study of California elderly women showed 25% to have undergone hysterectomy with bilateral oophorectomy, and an additional 11% to have had hysterectomy with ovarian conservation during the 1950s and 1960s [14]. The high rates of hysterectomy in those decades could account for a lowered incidence of ovarian cancer in the older population seen in our study, but it is unclear how much of a trend
Table 3 Estimated annual percent change (EAPC) and 95% confidence intervals (CI) for ovarian cancer-related mortality, California, 1990–2003
Age at death All ages b30 30–49 50–64 65–74 75+ Race/Ethnicity a All race/ethnicities b White Black Hispanic Asian/Pacific Islander a
Number of deaths
%
EAPC
95% CI
19,719 102 1843 4979 5585 7219
100 0.5 9.3 25.3 28.3 36.6
−0.5 – −1.2 −1.4 ⁎ −0.5 0.5
− 1.0, 0.0 – − 2.8, 0.4 − 2.3, −0.5 − 1.3, 0.3 − 0.4, 1.5
19,719 15,333 926 2291 1120
100 77.8 4.7 11.6 5.7
−0.5 −0.3 −0.3 0.6 0.3
− 1.0, 0.0 − 0.9, 0.3 − 2.4, 1.8 − 0.7, 1.9 − 1.1, 1.7
Race/Ethnic groups are mutually exclusive; women of Hispanic origin may be of any race. “All race/ethnicities” category includes women of unknown race or ethnicity. ⁎ The estimated annual percent change is significantly different from zero (p b 0.05). b
212
C.R. Morris et al. / Gynecologic Oncology 108 (2008) 207–213
existed in those rates during those earlier decades. Tubal ligation rates increased from 1970 to 1980 and it is possible that this increase impacted on the lowered incidence of ovarian cancer, particularly in our 65–74 age group [15]. We observed a decline of 2.5% per year in ovarian cancer incidence among black women, approximately twice the decline reported in the recent Annual Report to the Nation on the Status of Cancer, which reported decreases in ovarian cancer incidence during the 1995–2003 period of 1.2% per year among black women [16]. However, a decline of 1.3% per year among Hispanic women in the U.S. [16] was not observed in the California Hispanic population, where the decline in rates (−0.4% per year) was small and not statistically significant. Since the Hispanic population in California is largely composed of first- and secondgeneration immigrants who retain much of the social behavior of their country of origin, it is possible that this population was less likely to use oral contraceptives. Romo et al. showed that Spanishspeaking women were less likely to use contraceptives with a shorter duration of residence in the U.S. [17]. The California Women's Health Survey, an ongoing effort that interviews approximately 4000 randomly selected women per year, reported a much lower frequency of oral contraceptive use by women among minority race/ethnic groups, compared with non-Hispanic white women [18]. Additional analyses of trends using weighted percentages of respondents from that same survey (data not shown) revealed a significant decline in use of oral contraceptives among California Hispanic women from 1997 to 2004 [19]. California Hispanics also initially had lower rates of ovarian cancer than white and black women, due probably to higher parity and other lifestyle differences, and may increase their risk as they become more acculturated. Increasing incidence has been observed in Southern European populations with initially lower rates and has been attributed to a reduction in parity [2]. A study on the risk of epithelial ovarian cancer by invasiveness and cell type found that Hispanic women were more likely than non-Hispanic white women to be diagnosed with borderline or low malignant potential tumors [20]. We observed an increase of 1.8% per year in borderline tumors, although trends in this category of tumors were not analyzed by race/ethnicity. A 1987 population-based study in western Washington found an upward trend in the incidence of borderline tumors starting in the late 1970s [21], an increase which was also detected in Norway [22]. The distinction between borderline and invasive ovarian tumors is determined by histopathologic features, and it is possible that such assessment varies among pathologists, although we did not find references to any temporal shift in the recognition of borderline tumors. Some studies have found similar epidemiologic features among invasive and borderline ovarian tumors [23,24], while other studies suggest that they represent two different entities [25], with different genetic predisposition and molecular mechanisms [25,26]. Nonetheless, it has been shown that, compared with that of invasive tumors, the risk of borderline tumors is more clearly elevated among women with a history of infertility [24] and among those treated with fertility drugs [25]. While intriguing, these hormonal connections were not verified in this study. Overall, rates of death due to ovarian cancer in California did not decline over the period of our study, except among women 50–
64 years old at the time of death. There are several reasons why trends in cancer mortality may not parallel trends in incidence. Changes in treatment and shifts in distribution of stage at diagnosis have a profound impact in cancer mortality. For example, when screening is successful in detecting tumors at earlier and more favorable stages, or when more effective treatments are introduced, survival is prolonged and mortality decreases at a faster rate than incidence. Conversely, when fewer tumors are diagnosed but at more advanced stages, mortality will decrease over time, but at a slower rate. In the absence of effective screening methods, the distribution of stage at diagnosis for ovarian cancer in California has not substantially changed during the period covered in this study. However, for reasons not clearly understood, there was a significant increase in the proportion of undifferentiated tumors during the period. Although undifferentiated tumors account for only a small fraction of invasive tumors, their aggressiveness might at least in part explain the lack of a significant decline in ovarian cancer mortality rates. Notwithstanding the fact that ovarian cancer is still usually discovered at late stages of disease, five-year survival nationally improved from 37% in the mid-1970s to 44.9% for cases diagnosed between 1992 and 2002 [1]. This improvement is likely related to the introduction of more effective chemotherapeutic agents in the late 1970s and beyond. In our investigation, ovarian cancer cases diagnosed from 1990 to 2003 were included. Mortality data, collected from the Center for Health Statistics for the same time period, do not directly correspond to the same patient population as the incident cases being reported. It is possible, therefore, that with additional time the mortality rates will also decline to parallel the decline in incidence. If mortality rates do not decline in the near future despite lower incidence rates and improved treatment regimens, with no change in stage at presentation, it would be a strong indicator for further study to elucidate potential reasons. References [1] Ries LAGHD, Krapcho M, Mariotto A, Miller BA, Feuer EJ, Clegg L, Eisner MP, Horner MJ, Howlader N, Hayat M, Hankey BF, Edwards BK, editors. SEER Cancer Statistics Review, 1975–2003; 2006. cited; Available from: http://seer.cancer.gov/csr/1975_2003/. [2] Bray F, Loos AH, Tognazzo S, La Vecchia C. Ovarian cancer in Europe: cross-sectional trends in incidence and mortality in 28 countries, 1953–2000. Int J Cancer 2005;113(6):977–90. [3] Fritz APC, Jack A, Shanmugaratnam K, Sobin L, Parkin DM, Whelan S, editors. ICD-O: International Classification of Diseases for Oncology. Third Edition. Geneva: World Health Organization; 2000. [4] SEER Program: Comparative Staging Guide for Cancer, Version 1.1: National Cancer Institute, National Institutes of Health Publication No. 933640 June, 1993. [5] Stewart SL, Swallen KC, Glaser SL, Horn-Ross PL, West DW. Comparison of methods for classifying Hispanic ethnicity in a population-based cancer registry. Am J Epidemiol 1999;149(11):1063–71. [6] Kim HJ, Fay MP, Feuer EJ, Midthune DN. Permutation tests for joinpoint regression with applications to cancer rates. Stat Med 2000;19(3):335–51. [7] Villard-Mackintosh L, Vessey MP, Jones L. The effects of oral contraceptives and parity on ovarian cancer trends in women under 55 years of age. Br J Obstet Gynaecol 1989;96(7):783–8. [8] dos Santos Silva I, Swerdlow AJ. Recent trends in incidence of and mortality from breast, ovarian and endometrial cancers in England and Wales and their relation to changing fertility and oral contraceptive use. Br J Cancer 1995;72(2):485–92.
C.R. Morris et al. / Gynecologic Oncology 108 (2008) 207–213 [9] Siskind V, Green A, Bain C, Purdie D. Beyond ovulation: oral contraceptives and epithelial ovarian cancer. Epidemiology 2000;11(2): 106–10. [10] Ness RB, Grisso JA, Cottreau C, Klapper J, Vergona R, Wheeler JE, et al. Factors related to inflammation of the ovarian epithelium and risk of ovarian cancer. Epidemiology 2000;11(2):111–7. [11] Gnagy S, Ming EE, Devesa SS, Hartge P, Whittemore AS. Declining ovarian cancer rates in U.S. women in relation to parity and oral contraceptive use. Epidemiology 2000;11(2):102–5. [12] Oriel KA, Hartenbach EM, Remington PL. Trends in United States ovarian cancer mortality, 1979–1995. Obstet Gynecol 1999;93(1):30–3. [13] Millar WJ. Hysterectomy, 1981/82 to 1996/97. Health Rep 2001;12(2):9–22. [14] Kritz-Silverstein D, Von Muhlen DG, Ganiats TG, Barrett-Connor E. Hysterectomy status, estrogen use and quality of life in older women: the Rancho Bernardo study. Qual Life Res 2004;13(1):55–62. [15] MacKay AP, Kieke Jr BA, Koonin LM, Beattie K. Tubal sterilization in the United States, 1994–1996. Fam Plann Perspect 2001;33(4):161–5. [16] Howe HL, Wu X, Ries LA, Cokkinides V, Ahmed F, Jemal A, et al. Annual report to the nation on the status of cancer, 1975–2003, featuring cancer among U.S. Hispanic/Latino populations. Cancer 2006;107(8):1711–42. [17] Romo LF, Berenson AB, Segars A. Sociocultural and religious influences on the normative contraceptive practices of Latino women in the United States. Contraception 2004;69(3):219–25. [18] Foster D, Mikanda J, Godecker A, Steinberg S. Contraceptive use and risk for unintended pregnancy among women in California: results from the 1998–2001 California Women's Health Surveys. In: Women's Health: Findings from the California Women's Health Survey, 1997–2003. Office
[19]
[20]
[21] [22]
[23]
[24]
[25]
[26]
213
of Women's Health, California Department of Health Services, Sacramento, CA; 2004. Wayland S, Induni M, Davis B. California Women's Health Survey SAS Dataset Documentation and Technical Report: 1997–2006. Sacramento, CA: Survey Research Group, California Department of Health Services; 2007. Mills PK, Riordan DG, Cress RD. Epithelial ovarian cancer risk by invasiveness and cell type in the Central Valley of California. Gynecol Oncol 2004;95(1):215–25. Harlow BL, Weiss NS, Lofton S. Epidemiology of borderline ovarian tumors. J Natl Cancer Inst 1987;78(1):71–4. Bjorge T, Engeland A, Hansen S, Trope CG. Trends in the incidence of ovarian cancer and borderline tumours in Norway, 1954–1993. Int J Cancer 1997;71(5):780–6. Eltabbakh GH, Natarajan N, Piver MS, Mettlin CJ. Epidemiologic differences between women with borderline ovarian tumors and women with epithelial ovarian cancer. Gynecol Oncol 1999;74(1):103–7. Harris R, Whittemore AS, Itnyre J. Characteristics relating to ovarian cancer risk: collaborative analysis of 12 US case–control studies. III. Epithelial tumors of low malignant potential in white women. Collaborative Ovarian Cancer Group. Am J Epidemiol 1992;136(10):1204–11. Shushan A, Paltiel O, Schenker JG. Induction of ovulation and borderline ovarian cancer—the hormonal connection? Eur J Obstet Gynecol Reprod Biol 1999;85(1):71–4. Gotlieb WH, Chetrit A, Menczer J, Hirsh-Yechezkel G, Lubin F, Friedman E, et al. Demographic and genetic characteristics of patients with borderline ovarian tumors as compared to early stage invasive ovarian cancer. Gynecol Oncol 2005;97(3):780–3.
Gynecologic Oncology 108 (2008) 207 – 213 www.elsevier.com/locate/ygyno
Declining trends of epithelial ovarian cancer in California Cyllene R. Morris a,⁎, Anne O. Rodriguez b , Joan Epstein c , Rosemary D. Cress a,d a
b
d
California Cancer Registry, Public Health Institute, Sacramento, California, USA Division of Gynecologic Oncology, University of California Davis Medical Center, Sacramento, California, USA c Survey Research Group, California Department of Health Services, Sacramento, California, USA Department of Public Health Sciences, Division of Epidemiology, University of California Davis, California, USA Received 25 April 2007 Available online 29 October 2007
Abstract Objective. The purpose of this study was to examine trends of ovarian cancer in California. Known risk factors for ovarian cancer are unequally distributed across different cultures. California, the most ethnically diverse State in the U.S., is uniquely positioned to detect trends in different race/ethnic groups. Methods. This study was based on ovarian cancer cases and deaths occurring between 1990 and 2003 and reported to the California Cancer Registry (CCR). Age-adjusted rates were calculated by race/ethnicity, age at diagnosis, tumor behavior, and morphology. Trends were evaluated using the annual percent change, estimated through joinpoint regression. Results. The study population included 30,669 invasive ovarian cancers, 5925 tumors of borderline malignancy, and 19,719 ovarian cancer deaths. Age-adjusted incidence rates of invasive ovarian cancer per 100,000 women decreased from 15.6 in 1990 to 12.9 in 2003, a statistically significant decline of 1.3% per year. Rates of epithelial malignancies declined in all age groups examined, with the largest decreases noticed among women 65–74 years old. Substantial differences in trends were detected by race/ethnicity. Epithelial malignancies declined significantly among non-Hispanic black and white women by 2.5% and 1.2% per year, respectively. Changes among Asian/Pacific Islander and Hispanic women were smaller and non-significant. Ovarian cancer mortality rates did not change significantly during the study period. Conclusions. The incidence of epithelial ovarian cancer in California declined significantly among all age groups examined and among nonHispanic white and black women. Among Hispanic and Asian/Pacific Islander women, declines were small and non-significant © 2007 Elsevier Inc. All rights reserved. Keywords: Ovarian cancer; Incidence; Race; Ethnicity
Introduction Ovarian cancer is the sixth most commonly diagnosed cancer in women in the United States. In the absence of effective screening methods, the great majority of malignant ovarian neoplasms in the U.S. are detected at advanced stages of disease, when patients present with more obvious symptoms and face poorer prognosis [1]. Data from the National Cancer Institute's Surveillance, Epidemiology and End Results (SEER) program show that, overall, ovarian cancer incidence rates in the U.S. have declined by 0.9% per year since the late 1980s [1], although little is known about incidence trends in different race/ethnic groups. Declines also have ⁎ Corresponding author. 1700 Tribute Rd, Suite 100, Sacramento, CA 958154402, USA. Fax: +1 916 779 2608. E-mail address: [email protected] (C.R. Morris). 0090-8258/$ - see front matter © 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.ygyno.2007.09.026
been observed in northern European countries [2]. The purpose of the present study was to examine trends of ovarian cancer incidence and mortality among California women. A culturally diverse and multiethnic state, California has the country's second largest Hispanic and Asian/Pacific Islander populations (http:// www.census.gov/population/cen2000/phc-t6). Because the distribution of risk factors for ovarian cancer is most likely heterogeneous across different cultures, California is uniquely positioned to detect differences in temporal trends by race/ ethnicity. Materials and methods Data sources Ovarian cancer incidence data were obtained from the CCR, California's statewide population-based cancer surveillance system. Since 1988, all
208
C.R. Morris et al. / Gynecologic Oncology 108 (2008) 207–213
hospitals, physicians and laboratories in California have been mandated by law to report every cancer diagnosis, with the exception of basal and squamous cell carcinomas of the skin and carcinoma in situ of the cervix. For every newly diagnosed cancer, the CCR gathers information on tumor characteristics, diagnosis, treatment, and follow-up (plus patient's demographic information). Data are collected through a network of ten regional registries, which also participate in the National Cancer Institute's SEER Program. As of April 2006 statewide case ascertainment for 2003 was estimated to be complete. Data on ovarian cancer-related deaths were obtained from the death certificate master files, Center for Health Statistics, California Department of Health Services, while population estimates by age and race/ethnicity were obtained from the U.S. Census Bureau.
Ovarian cancer cases and deaths This study included 30,669 invasive ovarian cancers diagnosed in California from January 1990 through December 2003 and reported to the CCR as of April 2006 (Table 1). Incidence trends of 5925 ovarian tumors of borderline
Table 1 Frequency distribution of ovarian cancers diagnosed in California from 1990 through 2003 by demographic and tumor-related characteristics Characteristic
Number of cases
5925 30,669
16.2 83.8
Invasive tumors (n = 30,669) Tumor morphology Epithelial Germ cell Sex cord–stromal Other
28,365 975 356 973
92.4 3.2 1.2 3.2
a
Race/Ethnicity
%
All tumors (N = 36,594) Tumor behavior Borderline malignancy Invasive
Invasive epithelial tumors (N = 28,365) Age at diagnosis b30 30–49 50–64 65–74 75+ Race/Ethnicity a White Black Hispanic Asian/Pacific Islander Unknown Tumor grade b G1 G2 G3 G4 Unknown Stage at diagnosis (N = 20,425) c IA–IB IC–II III–IV Unstaged
malignancy are also presented. CCR data on anatomic site and tumor morphology were abstracted from the patient's medical records and pathology reports. Tumors with International Classification of Diseases for Oncology, Third Edition (ICD-O-3) morphology codes 8010-8570 (excluding 8240-8245), 9014-9015, or 9110 were grouped as epithelial tumors [3]. Tumors for which no morphologic descriptors are found when the information is abstracted are coded as not otherwise specified malignancies (NOS, ICD-O-3 codes 8000-8004). Because the vast majority of ovarian malignancies are epithelial in origin, NOS tumors were included in the epithelial category. Stage at diagnosis was based on extent of disease codes (available for cases diagnosed in 1994 forward) converted to the American Joint Committee on Cancer (AJCC) stage categories [4]. AJCC stages were subsequently grouped as IA–IB, IC–II, III–IV, and unstaged (Table 1). Tumor grade, or degree of differentiation of neoplastic tissue, was coded as G1 (well differentiated), G2 (moderately differentiated), G3 (poorly differentiated), or G4 (undifferentiated or anaplastic). Mortality data (Death Master Files) were provided by the California Department of Health Services Center for Health Statistics. A total of 19,719 deaths where ovarian cancer was identified as the underlying cause of death in the death certificate were included in the study. Deaths occurring between 1990 and 1998 were coded by the International Classification of Diseases (ICD), Ninth Edition, while deaths occurring from 1999 forward were coded by the ICD, Tenth Edition.
369 5296 8503 7009 7188
1.3 18.7 30.0 24.7 25.3
20,919 1165 3922 2223 136
73.7 4.1 13.8 7.8 0.5
2264 5122 10,223 1933 8823
8.0 18.1 36.0 6.8 31.1
2559 2941 13,287 1638
12.5 14.4 65.1 8.0
Race/Ethnic groups are mutually exclusive; women of Hispanic origin may be of any race. b Tumor grade: G1 (well differentiated), G2 (moderately differentiated), G3 (poorly differentiated), and G4 (undifferentiated). c Stage at diagnosis based on extent of disease codes (available for cases diagnosed in 1994 forward) converted to American Joint Committee on Cancer (AJCC) categories.
Race/Ethnicity for cases reported to the CCR was determined from information in the medical record or death certificate. Incidence and mortality figures were compiled for four major race/ethnicity groups: non-Hispanic white (white), Hispanic, non-Hispanic Asian/Pacific Islander (Asian/PI), and non-Hispanic black (black). To determine Hispanic ethnicity more accurately, information from the medical record was supplemented by linking the subject's last name or maiden name to the 1980 US Census list of Hispanic surnames [5]. Patients identified either by their medical record or their surnames as Hispanic were classified as Hispanic.
Adjustment for completeness of case reporting Incidence rates are calculated by the CCR when reporting of new cases is estimated to be at least 95% complete, which usually happens within 18 to 24 months after the end of a diagnosis year. However, cancer surveillance is a dynamic process and cases diagnosed at earlier years may be reported long after incidence data are considered “complete”. An internal review of ovarian cancer incidence rates reported in CCR annual publications revealed that rates for a given year of diagnosis increased annually up to 8 years after the initial release and stabilized after that (data not shown). For example, the 1991 incidence rate of ovarian cancer reported by the CCR in early 1994 was 14.5 cases per 100,000 women. By 2000, the rate for 1991 had increased to 15.0 per 100,000 due to additional cases being reported over time. In this internal review of former CCR reports, the average increase in ovarian cancer rates 1 year after the first release was 1.5%. Compared to the initial estimate, rates for the same year were about 4% higher after 8 years of surveillance. In order to minimize the effect of incomplete reporting, incidence rates were multiplied by a factor corresponding to the percent increase observed in CCR rates during the 8 years lapsed until estimates were stable. As of April 2006, rates from 1990 through 1995 were considered stable, while rates from 1996 through 2003 were inflated by 0.2%, 0.4%, 0.5%, 0.7%, 1.2%, 1.6%, 2.5%, and 4.0%, respectively. Multiplication factors were applied regardless of tumor behavior (invasive or borderline malignancy), morphology, race/ethnicity, or age at diagnosis.
Statistical analysis Frequency distributions of demographic (age at diagnosis and race/ethnicity) and tumor-related characteristics (tumor behavior, morphology, and stage at diagnosis) were obtained for all incident ovarian cancer cases. Incidence and mortality rates per 100,000 women were calculated for groups with at least 15 cases (or deaths) and were age-adjusted to the 2000 U.S. standard population. The Surveillance Research Program, National Cancer Institute SEER⁎Stat software version 6.2.3 (http://srab.cancer.gov/seerstat) was used to calculate all rates and frequency distributions. Joinpoint linear regression [6] was used to fit the log of
C.R. Morris et al. / Gynecologic Oncology 108 (2008) 207–213
209
age-adjusted rates. In this analysis, a statistical algorithm detects joinpoints, i.e., points in time where the slope of the regression line significantly changes. Thus, the model describes trends during different time segments. At each segment, trends in rates are measured using the estimated annual percent change (EAPC), which assumes that rates change by a constant percentage each year. The EAPC was calculated using the weighted least squares method, where a linear regression to the natural logarithm of the age-adjusted annual rates was fitted with year of diagnosis as the regressor variable. The estimated slope of the regression line m was then used to calculate the EAPC as equal to 100 ⁎ (em − 1). The SEER JoinPoint regression software version 3.0 (http://srab.cancer.gov/joinpoint) was used for all trend analyses in this study. Regression models were fit using weighted least squares, with default program parameters set. The significance of trends was evaluated with a threshold of 0.05. Incidence trends were evaluated by tumor behavior (invasive or borderline malignancy) and morphology (epithelial, germ cell, sex cord–stromal, and other). Trends for invasive epithelial tumors were further evaluated by tumor grade, age at diagnosis, and race/ethnicity. Two sets of estimates were obtained, with and without adjusting for completeness of reporting as described above. Rates for stage at diagnosis were not calculated, but chi-square tests for trend were performed to detect potential shifts over time in the distribution of cases diagnosed at earlier and late stages, or which were not staged. Trends in ovarian cancer mortality were evaluated by race/ethnicity and age at death only.
Results Characteristics of the study population Demographic and tumor-related characteristics of ovarian cancer patients in this study are shown in Table 1. Malignant neoplasms accounted for 83.7% of all ovarian cancers diagnosed in California from 1990 through 2003, while 16.1% of tumors were of borderline malignancy. Of the 30,669 invasive tumors, 92.4% were of epithelial origin, with germ cell, sex cord–stromal, and tumors of other histologies being diagnosed in 3.2%, 1.2%, and 3.2% of cases, respectively. Tumor grade was considered poorly differentiated in 36% of cases followed by moderately differentiated (18.1%), well differentiated (8%), and undifferentiated tumors (6.8%). Information on tumor grade was absent from patient records in 31.1% of cases. The majority of patients (65.1%) were diagnosed with a late-stage tumor (AJCC III or IV). A chi-square analysis of trend showed that, between 1994 and 2003, there were no significant changes in the proportion of cases diagnosed at localized (IA–IB) or regional (IC–II) stages. There was, however, a slight but significant (p = 0.04) increase in the proportion of cases diagnosed with late (III–IV) stage tumors, from 64.2% in 1994 to 66.8% in 2003. A significant decrease in the proportion of unstaged cases, from 9.3% to 7.1%, accounted for most of the observed shift. These patterns were similar in all age groups examined (data not shown). The proportion of tumors classified from well to poorly differentiated decreased significantly during the period. Such a decrease was associated with a very marked and significant increase in the proportion of undifferentiated tumors, from 5.4% of cases in 1990 to 8.6% in 2003 (p b 0.000). The proportion of tumors with unknown grade also increased, from 29.7% in 1990 to 36.9% in 2003. Trends in ovarian cancer incidence During the 14-year period evaluated in this study, invasive ovarian cancer age-adjusted rates per 100,000 women in California decreased from 15.6 in 1990 to 12.9 in 2003 (Fig. 1),
Fig. 1. Ovarian cancer age-adjusted incidence rates by tumor behavior, California, 1990–2003: observed and completeness-adjusted rates. Rates in the figure are per 100,000 women and age-adjusted to the 2000 U.S. standard population. To further adjust for completeness of reporting, observed age-adjusted rates were multiplied by factors representing average increases detected over time in ovarian cancer rates.
which corresponds to a statistically significant decline of 1.3% per year (95% CI for EAPC: −1.5, −1.1) (Table 2). Rates per 100,000 women in California for tumors of borderline malignancy, however, increased from 2.2 in 1990 to 2.9 in 2003, an also statistically significant change of 1.8% per year (95% CI for EAPC: 0.9, 2.7). The great majority of invasive ovarian tumors (92.4%) were of epithelial origin (Table 1). Accordingly, trends for epithelial tumors (EAPC = −1.4%, 95% CI: −1.6, −1.2) were similar to those for all invasive tumors, with rates per 100,000 women decreasing from 14.5 in 1990 to 11.9 in 2003. Germ cell tumors accounted for 3.2% of all invasive ovarian cancers, and their incidence trends were better described in two time periods. During the first period, from 1990 through 1998, rates declined significantly by 3.5% per year (95% CI for EAPC: −5.8, −1.2), while the second period was characterized by a significant and steep increase of 5.8% per year (95% CI for EAPC: 0.8, 11.1). A non-significant decline of 2.8% per year (95% CI for EAPC = −8.7, 3.4) was also detected among tumors of the sex cord– stromal tissue, a less common morphology accounting for 1.2% of malignant ovarian tumors. The incidence of tumors of other histologic types, such as malignant Brenner tumors, carcinosarcomas, and müellerian mixed tumors, remained unchanged during the 1990–2003 time period. Because they represented the vast majority of tumors, subsequent data analysis was limited to epithelial ovarian malignancies. The incidence of epithelial ovarian cancer in California declined
210
C.R. Morris et al. / Gynecologic Oncology 108 (2008) 207–213
Table 2 Estimated annual percent change (EAPC) and 95% confidence intervals (CI) for ovarian cancer incidence, California, 1990–2003: estimates based on actual rates and on rates adjusted a for completeness of reporting Characteristic
All tumors (N = 36,659) Tumor behavior Invasive Borderline
EAPC 95% CI
Adjusteda
95% CI
− 1.3 ⁎ 1.8⁎
− 1.5, −1.1 0.9, 2.7
− 1.1⁎ 1.6⁎
− 1.3, − 0.9 0.6, 2.6
− 1.6, −1.2
− 1.2⁎
− 1.4, − 0.9
− 5.8, −1.2 0.8, 11.1 − 8.7, 3.4 − 1.3, 2.2
− 3.5⁎ 5.8⁎ − 2.8 0.4
− 5.8, − 1.2 0.8, 11.1 − 8.7, 3.4 − 1.3, 2.2
Invasive tumors (N = 30,669) Tumor morphology Epithelial − 1.4 Germ cell 1990–1998 − 3.5⁎ 1998–2003 5.8⁎ Sex cord–stromal − 2.8 Other 0.4 Invasive epithelial tumors Age at diagnosis All ages b30 30–49 50–64 65–74 75+ Race/Ethnicity b All race/ethnicities c White Black Hispanic Asian/Pacific Islander
EAPC
Actual
(N = 28,365) − 1.4⁎ − 1.7 − 1.6⁎ − 1.1⁎ − 1.7⁎ − 1.3⁎
− 1.6, −1.2 − 4.0, 0.8 − 2.4, −0.8 − 1.5, −0.7 − 2.4, −1.1 − 2.1, −0.4
− 1.2⁎ − 1.7 − 1.4⁎ − 0.9⁎ − 1.5⁎ − 1.1⁎
− 1.4, − 0.9 − 4.0, 0.8 − 2.2, − 0.6 − 1.3, − 0.5 − 2.2, − 0.8 − 1.9, − 0.2
− 1.4⁎ − 1.2⁎ − 2.5⁎ − 0.4 − 0.7
− 1.6, −1.2 − 1.5, −0.9 − 4.3, −0.8 − 1.3, 0.5 − 1.5, 0.1
− 1.2⁎ − 1.0⁎ − 2.3⁎ − 0.2 − 0.5
− 1.4, − 0.9 − 1.3, − 0.7 − 4.0, − 0.6 − 1.1, 0.7 − 1.3, 0.4
actual CCR rates. Likewise, all other EAPCs were reduced by about 0.2% when adjusting for incomplete reporting. These corrections, however, did not result in loss of statistical significance in any of the groups examined. Trends in ovarian cancer mortality Changes in mortality rates from 1990 through 2003 were not statistically significant (EAPC = −0.5, 95% CI = −1.0, 0.0), from 9.5 deaths per 100,000 in 1990 to 9.1 deaths per 100,000 in 2003. Of all the race/ethnic groups and age categories evaluated, ovarian cancer-related deaths declined significantly only among women who were 50–64 years old at death (EAPC = −1.4, 95% CI = −2.3, −0.5) (Table 3). Discussion Our results show that incidence rates of epithelial ovarian malignancies in California women declined an average of 1.4% a year over the 14-year period 1990 through 2003. These results are consistent with an average 0.9% annual decline in ovarian cancer incidence observed in the U.S. between 1975 and 2003 [1], although rates declined faster in California. Our estimates are not directly comparable to those published by the SEER Program, due to different time periods examined and reporting
a To adjust for completeness of reporting, age-adjusted rates from 1996 forward were multiplied by factors corresponding to the average increase in ovarian cancer rates observed over time by the California Cancer Registry. b Race/Ethnic groups are mutually exclusive; women of Hispanic origin may be of any race. c “All race/ethnicities” category includes women of unknown race or ethnicity. ⁎ The estimated annual percent change is significantly different from zero (p b 0.05).
almost evenly in all age groups examined (Fig. 2), with statistically significant EAPCs varying from −1.1 to −1.7 (Table 2). The largest decrease in rates was noticed among women 65–74 years old, for whom incidence declined by about 22% (EAPC = −1.7, 95% CI= −2.4, −1.1) during the study's 14-year period. When trends were evaluated by race/ethnicity, substantial differences were uncovered (Fig. 3). Epithelial ovarian cancer incidence significantly declined among non-Hispanic black and white women. Rates among black women declined by a steep 52% from 1990 through 2003 (EAPC = −2.5, 95% CI = −4.3, −0.8). Among white women, rates decreased by about 18%, with a significant EAPC of −1.2 (95% CI: −1.5, −0.9). Changes in ovarian cancer incidence rates for Asian/Pacific Islander and Hispanic women were smaller and not statistically significant (Table 2). When rates were adjusted for completeness of reporting, trends were somewhat weakened (Table 2). For example, rates for invasive epithelial tumors still declined significantly over the period, but by 1.2% per year instead of the 1.4% estimated with
Fig. 2. Ovarian cancer age-adjusted incidence rates by age at diagnosis, California, 1990–2003. Rates in the figure are per 100,000 women and ageadjusted to the 2000 U.S. standard population.
C.R. Morris et al. / Gynecologic Oncology 108 (2008) 207–213
Fig. 3. Ovarian cancer age-adjusted incidence rates by race/ethnicity, California, 1990–2003. Rates in the figure are per 100,000 women and age-adjusted to the 2000 U.S. standard population. Race/ethnic groups are mutually exclusive; women of Hispanic origin may be of any race.
delay adjustment. However, the empirical adjustment for completeness employed in this study produced changes in EAPCs of about 0.2%, a magnitude of change similar to SEER trend estimates [1]. In addition to reporting delays, inadequate case finding or underreporting may also contribute to an artificial decline in cancer rates. Although these are recognized problems with cancers diagnosed outside of a hospital, it is
211
highly unlikely that this would be the case with ovarian cancer. Because most ovarian cancers are epithelial in origin, it is expected that epithelial tumors drive overall incidence trends. For other histologic types of ovarian tumors, trends were calculated based on a small number of cases diagnosed per year. For example, rates for germ cell tumors were based on an average of 70 cases diagnosed each year in California, and statistical significance does not necessarily equate to clinical significance. Declines in ovarian cancer incidence also have been documented in the UK and Northern European countries in recent years, particularly among young women, and have been attributed to increased use of oral contraceptives since the 1960s [2,7,8]. Oral contraceptive use has been shown to reduce risk of ovarian cancer [9,10]. Gnagy and colleagues demonstrated a correlation between observed ovarian cancer incidence rates in U.S. women between 1970 and 1995 with those predicted by changes in oral contraceptive use among younger women (aged 30 to 49) [11]. However, the results of their analysis suggested that oral contraceptives did not contribute to a decline in ovarian cancer incidence in older women (aged 50–64) [11]. In contrast to these earlier studies, in the current analysis, the largest declines were observed in women 65 years and older, those with the highest incidence rates. These women, in particular those aged 75 and older, would have had more limited exposure to oral contraceptives as they would have been in their 40s in the late 1960s as oral contraceptives emerged. Tubal ligation and/or hysterectomy also decrease risk [10] and may have played a role in the reduction of incidence among older California women. Hysterectomy rates declined in North America in the 1980s and 1990s, but in the 1950s–1970s, when older women in our study would have been in their reproductive years, hysterectomy rates were still quite high [12,13]. One study of California elderly women showed 25% to have undergone hysterectomy with bilateral oophorectomy, and an additional 11% to have had hysterectomy with ovarian conservation during the 1950s and 1960s [14]. The high rates of hysterectomy in those decades could account for a lowered incidence of ovarian cancer in the older population seen in our study, but it is unclear how much of a trend
Table 3 Estimated annual percent change (EAPC) and 95% confidence intervals (CI) for ovarian cancer-related mortality, California, 1990–2003
Age at death All ages b30 30–49 50–64 65–74 75+ Race/Ethnicity a All race/ethnicities b White Black Hispanic Asian/Pacific Islander a
Number of deaths
%
EAPC
95% CI
19,719 102 1843 4979 5585 7219
100 0.5 9.3 25.3 28.3 36.6
−0.5 – −1.2 −1.4 ⁎ −0.5 0.5
− 1.0, 0.0 – − 2.8, 0.4 − 2.3, −0.5 − 1.3, 0.3 − 0.4, 1.5
19,719 15,333 926 2291 1120
100 77.8 4.7 11.6 5.7
−0.5 −0.3 −0.3 0.6 0.3
− 1.0, 0.0 − 0.9, 0.3 − 2.4, 1.8 − 0.7, 1.9 − 1.1, 1.7
Race/Ethnic groups are mutually exclusive; women of Hispanic origin may be of any race. “All race/ethnicities” category includes women of unknown race or ethnicity. ⁎ The estimated annual percent change is significantly different from zero (p b 0.05). b
212
C.R. Morris et al. / Gynecologic Oncology 108 (2008) 207–213
existed in those rates during those earlier decades. Tubal ligation rates increased from 1970 to 1980 and it is possible that this increase impacted on the lowered incidence of ovarian cancer, particularly in our 65–74 age group [15]. We observed a decline of 2.5% per year in ovarian cancer incidence among black women, approximately twice the decline reported in the recent Annual Report to the Nation on the Status of Cancer, which reported decreases in ovarian cancer incidence during the 1995–2003 period of 1.2% per year among black women [16]. However, a decline of 1.3% per year among Hispanic women in the U.S. [16] was not observed in the California Hispanic population, where the decline in rates (−0.4% per year) was small and not statistically significant. Since the Hispanic population in California is largely composed of first- and secondgeneration immigrants who retain much of the social behavior of their country of origin, it is possible that this population was less likely to use oral contraceptives. Romo et al. showed that Spanishspeaking women were less likely to use contraceptives with a shorter duration of residence in the U.S. [17]. The California Women's Health Survey, an ongoing effort that interviews approximately 4000 randomly selected women per year, reported a much lower frequency of oral contraceptive use by women among minority race/ethnic groups, compared with non-Hispanic white women [18]. Additional analyses of trends using weighted percentages of respondents from that same survey (data not shown) revealed a significant decline in use of oral contraceptives among California Hispanic women from 1997 to 2004 [19]. California Hispanics also initially had lower rates of ovarian cancer than white and black women, due probably to higher parity and other lifestyle differences, and may increase their risk as they become more acculturated. Increasing incidence has been observed in Southern European populations with initially lower rates and has been attributed to a reduction in parity [2]. A study on the risk of epithelial ovarian cancer by invasiveness and cell type found that Hispanic women were more likely than non-Hispanic white women to be diagnosed with borderline or low malignant potential tumors [20]. We observed an increase of 1.8% per year in borderline tumors, although trends in this category of tumors were not analyzed by race/ethnicity. A 1987 population-based study in western Washington found an upward trend in the incidence of borderline tumors starting in the late 1970s [21], an increase which was also detected in Norway [22]. The distinction between borderline and invasive ovarian tumors is determined by histopathologic features, and it is possible that such assessment varies among pathologists, although we did not find references to any temporal shift in the recognition of borderline tumors. Some studies have found similar epidemiologic features among invasive and borderline ovarian tumors [23,24], while other studies suggest that they represent two different entities [25], with different genetic predisposition and molecular mechanisms [25,26]. Nonetheless, it has been shown that, compared with that of invasive tumors, the risk of borderline tumors is more clearly elevated among women with a history of infertility [24] and among those treated with fertility drugs [25]. While intriguing, these hormonal connections were not verified in this study. Overall, rates of death due to ovarian cancer in California did not decline over the period of our study, except among women 50–
64 years old at the time of death. There are several reasons why trends in cancer mortality may not parallel trends in incidence. Changes in treatment and shifts in distribution of stage at diagnosis have a profound impact in cancer mortality. For example, when screening is successful in detecting tumors at earlier and more favorable stages, or when more effective treatments are introduced, survival is prolonged and mortality decreases at a faster rate than incidence. Conversely, when fewer tumors are diagnosed but at more advanced stages, mortality will decrease over time, but at a slower rate. In the absence of effective screening methods, the distribution of stage at diagnosis for ovarian cancer in California has not substantially changed during the period covered in this study. However, for reasons not clearly understood, there was a significant increase in the proportion of undifferentiated tumors during the period. Although undifferentiated tumors account for only a small fraction of invasive tumors, their aggressiveness might at least in part explain the lack of a significant decline in ovarian cancer mortality rates. Notwithstanding the fact that ovarian cancer is still usually discovered at late stages of disease, five-year survival nationally improved from 37% in the mid-1970s to 44.9% for cases diagnosed between 1992 and 2002 [1]. This improvement is likely related to the introduction of more effective chemotherapeutic agents in the late 1970s and beyond. In our investigation, ovarian cancer cases diagnosed from 1990 to 2003 were included. Mortality data, collected from the Center for Health Statistics for the same time period, do not directly correspond to the same patient population as the incident cases being reported. It is possible, therefore, that with additional time the mortality rates will also decline to parallel the decline in incidence. If mortality rates do not decline in the near future despite lower incidence rates and improved treatment regimens, with no change in stage at presentation, it would be a strong indicator for further study to elucidate potential reasons. References [1] Ries LAGHD, Krapcho M, Mariotto A, Miller BA, Feuer EJ, Clegg L, Eisner MP, Horner MJ, Howlader N, Hayat M, Hankey BF, Edwards BK, editors. SEER Cancer Statistics Review, 1975–2003; 2006. cited; Available from: http://seer.cancer.gov/csr/1975_2003/. [2] Bray F, Loos AH, Tognazzo S, La Vecchia C. Ovarian cancer in Europe: cross-sectional trends in incidence and mortality in 28 countries, 1953–2000. Int J Cancer 2005;113(6):977–90. [3] Fritz APC, Jack A, Shanmugaratnam K, Sobin L, Parkin DM, Whelan S, editors. ICD-O: International Classification of Diseases for Oncology. Third Edition. Geneva: World Health Organization; 2000. [4] SEER Program: Comparative Staging Guide for Cancer, Version 1.1: National Cancer Institute, National Institutes of Health Publication No. 933640 June, 1993. [5] Stewart SL, Swallen KC, Glaser SL, Horn-Ross PL, West DW. Comparison of methods for classifying Hispanic ethnicity in a population-based cancer registry. Am J Epidemiol 1999;149(11):1063–71. [6] Kim HJ, Fay MP, Feuer EJ, Midthune DN. Permutation tests for joinpoint regression with applications to cancer rates. Stat Med 2000;19(3):335–51. [7] Villard-Mackintosh L, Vessey MP, Jones L. The effects of oral contraceptives and parity on ovarian cancer trends in women under 55 years of age. Br J Obstet Gynaecol 1989;96(7):783–8. [8] dos Santos Silva I, Swerdlow AJ. Recent trends in incidence of and mortality from breast, ovarian and endometrial cancers in England and Wales and their relation to changing fertility and oral contraceptive use. Br J Cancer 1995;72(2):485–92.
C.R. Morris et al. / Gynecologic Oncology 108 (2008) 207–213 [9] Siskind V, Green A, Bain C, Purdie D. Beyond ovulation: oral contraceptives and epithelial ovarian cancer. Epidemiology 2000;11(2): 106–10. [10] Ness RB, Grisso JA, Cottreau C, Klapper J, Vergona R, Wheeler JE, et al. Factors related to inflammation of the ovarian epithelium and risk of ovarian cancer. Epidemiology 2000;11(2):111–7. [11] Gnagy S, Ming EE, Devesa SS, Hartge P, Whittemore AS. Declining ovarian cancer rates in U.S. women in relation to parity and oral contraceptive use. Epidemiology 2000;11(2):102–5. [12] Oriel KA, Hartenbach EM, Remington PL. Trends in United States ovarian cancer mortality, 1979–1995. Obstet Gynecol 1999;93(1):30–3. [13] Millar WJ. Hysterectomy, 1981/82 to 1996/97. Health Rep 2001;12(2):9–22. [14] Kritz-Silverstein D, Von Muhlen DG, Ganiats TG, Barrett-Connor E. Hysterectomy status, estrogen use and quality of life in older women: the Rancho Bernardo study. Qual Life Res 2004;13(1):55–62. [15] MacKay AP, Kieke Jr BA, Koonin LM, Beattie K. Tubal sterilization in the United States, 1994–1996. Fam Plann Perspect 2001;33(4):161–5. [16] Howe HL, Wu X, Ries LA, Cokkinides V, Ahmed F, Jemal A, et al. Annual report to the nation on the status of cancer, 1975–2003, featuring cancer among U.S. Hispanic/Latino populations. Cancer 2006;107(8):1711–42. [17] Romo LF, Berenson AB, Segars A. Sociocultural and religious influences on the normative contraceptive practices of Latino women in the United States. Contraception 2004;69(3):219–25. [18] Foster D, Mikanda J, Godecker A, Steinberg S. Contraceptive use and risk for unintended pregnancy among women in California: results from the 1998–2001 California Women's Health Surveys. In: Women's Health: Findings from the California Women's Health Survey, 1997–2003. Office
[19]
[20]
[21] [22]
[23]
[24]
[25]
[26]
213
of Women's Health, California Department of Health Services, Sacramento, CA; 2004. Wayland S, Induni M, Davis B. California Women's Health Survey SAS Dataset Documentation and Technical Report: 1997–2006. Sacramento, CA: Survey Research Group, California Department of Health Services; 2007. Mills PK, Riordan DG, Cress RD. Epithelial ovarian cancer risk by invasiveness and cell type in the Central Valley of California. Gynecol Oncol 2004;95(1):215–25. Harlow BL, Weiss NS, Lofton S. Epidemiology of borderline ovarian tumors. J Natl Cancer Inst 1987;78(1):71–4. Bjorge T, Engeland A, Hansen S, Trope CG. Trends in the incidence of ovarian cancer and borderline tumours in Norway, 1954–1993. Int J Cancer 1997;71(5):780–6. Eltabbakh GH, Natarajan N, Piver MS, Mettlin CJ. Epidemiologic differences between women with borderline ovarian tumors and women with epithelial ovarian cancer. Gynecol Oncol 1999;74(1):103–7. Harris R, Whittemore AS, Itnyre J. Characteristics relating to ovarian cancer risk: collaborative analysis of 12 US case–control studies. III. Epithelial tumors of low malignant potential in white women. Collaborative Ovarian Cancer Group. Am J Epidemiol 1992;136(10):1204–11. Shushan A, Paltiel O, Schenker JG. Induction of ovulation and borderline ovarian cancer—the hormonal connection? Eur J Obstet Gynecol Reprod Biol 1999;85(1):71–4. Gotlieb WH, Chetrit A, Menczer J, Hirsh-Yechezkel G, Lubin F, Friedman E, et al. Demographic and genetic characteristics of patients with borderline ovarian tumors as compared to early stage invasive ovarian cancer. Gynecol Oncol 2005;97(3):780–3.