- Email: [email protected]
Trends in gestational choriocarcinoma: a 27-year perspective
Trends in Gestational Choriocarcinoma: A 27-Year Perspective Harriet O. Smith, MD, Clifford R. Qualls, PhD, Beth A. Prairie, MD, Luis A. Padilla, MD, William F. Rayburn, MD, and Charles R. Key, MD, PhD OBJECTIVE: To evaluate trends in incidence and survival rates for gestational choriocarcinoma with the use of population-based data. METHODS: Overall and 5-year average age-adjusted incidence rates were computed with the Surveillance, Epidemiology, and End Results program public-use database. Differences by age at diagnosis, race, stage, registry, and over calendar time were compared by Poisson regression, and survival censored for deaths other than choriocarcinoma by log-rank tests and Cox’s proportional hazard ratios. RESULTS: Between 1973 and 1999, 450 cases were recorded. The annualized age-adjusted incidence rate for choriocarcinoma was 0.133 per 100,000 woman-years and decreased by 49.7% (2.8% per year). By race (whites, blacks, and others), incidence rates declined by 62.3%, 27.2%, and 54.3%, respectively. In the Poisson model evaluating incidence rates, age, race, registry, and stage were significant main effects. Compared with whites, the relative risk was higher for blacks (2.14, 95% confidence interval [CI] 1.60, 2.86) and others (2.30, 95% CI 1.67, 3.18). Rates were highest in Utah and lowest in Iowa. By age at diagnosis, rates were higher in 20 –39-year-olds. The 5-year relative survival rate was 89.5%. Censored survival was significantly lower among blacks (whites 92.4%, blacks 84.9%, others 87.1%, P ⴝ .045), for advanced disease (localized 94.5%, regional 92.9%, distant 87.1%, P ⴝ .02), and with increasing age at diagnosis (P ⴝ .017). Age and calendar time significantly influenced censored survival independent of stage and registry. CONCLUSION: Gestational choriocarcinoma incidence rates have declined and survivals have improved, but blacks continue to have higher incidence and lower survival rates. (Obstet Gynecol 2003;102:978 – 87. © 2003 by The American College of Obstetricians and Gynecologists.) From the Departments of Obstetrics and Gynecology, Mathematics and Statistics, and Pathology, University of New Mexico Health Sciences Center, Albuquerque, New Mexico; and Department of Obstetrics and Gynecology, Dartmouth-Hitchcock Medical Center, Lebanon, New Hampshire. Partial funding provided by National Cancer Institute Surveillance, Epidemiology and End Results Contract NO1-67007.
978
Gestational choriocarcinoma is a relatively rare and highly malignant variant of gestational trophoblastic disease.1– 4 Although choriocarcinoma might be preceded by any gestational event, hydatidiform mole is the most common precursor. Throughout the world, incidence rates for choriocarcinoma differ widely.5 For example, in Europe and North America, choriocarcinomas are reported to affect one in every 30,000 – 40,000 pregnancies, and one in 40 molar pregnancies,6 whereas in Southeast Asia, rates as high as one in every 500 –3000 pregnancies have been reported.7 Comparisons of gestational trophoblastic disease incidence rates among different regions of the world are limited by the methods used to compute rates. The denominators most often used to compute rates (live births, pregnancies, deliveries) are hospital-based, whereas the numerator, incident cases, tends to reflect referral patterns within a given continent or region. Although of greater clinical relevance, it is even more difficult to compare choriocarcinoma incidence rates, because pathologic confirmation is not always feasible, and the disease is rare.5 Population-based data that can be applied consistently between regions is necessary to consistently determine rates and to accurately compare differences. The best available published estimate of choriocarcinoma incidence rates in the United States was derived from data obtained from the Surveillance, Epidemiology and End Results program (1973–1982).5 Differences in age-adjusted incidence rates by age at diagnosis and race (black, white, other) were reported, but there were insufficient data to analyze trends in incidence or survival rates. The objective of this study was to analyze trends in incidence and survival rates for gestational choriocarcinoma with the most recent version of the Surveillance, Epidemiology and End Results program’s cancer incidence publicuse database (SEERSTAT 4.2; Surveillance, Epidemiology and End Results 9 Registries public-use, November 2001 submission), which includes all cases recorded between 1973 and 1999. Although the original nine Surveillance, Epidemiology and End Results registries
VOL. 102, NO. 5, PART 1, NOVEMBER 2003 © 2003 by The American College of Obstetricians and Gynecologists. Published by Elsevier.
0029-7844/03/$30.00 doi:10.1016/S0029-7844(03)00669-0
record carcinoma cases for only approximately 10% of the US population, data derived from the Surveillance, Epidemiology and End Results registry is regarded as the best estimate of cancer incidence and mortality trends in the United States. MATERIAL AND METHODS The Surveillance, Epidemiology and End Results program is a set of geographically defined, population-based central registries for the United States, operated by local nonprofit organizations under contract to the National Cancer Institute. The original nine Surveillance, Epidemiology and End Results registries have been in operation since 1973. Each year, certified abstractors review the medical records and pathology reports of all hospitals and laboratories and identify all incident invasive carcinoma cases. Only cases affecting residents within the individual Surveillance, Epidemiology and End Results registry are included. Ethnicity is assigned according to the best available information, usually by the racial preference stated on the hospital charts, on the basis of identifiers in the record, or by surname. Because there is no mechanism for centralized pathology review, the potential for variability in experience and expertise of pathologists and clinicians involved is substantial—a known limitation of this data set. Registry data are submitted electronically without personal identifiers to the National Cancer Institute on a biannual basis. Surveillance, Epidemiology and End Results public-use data and software are provided at no charge to individuals interested in conducting scientific research (http://seer. cancer.gov/seerstat). Approval to perform this study was obtained from the internal review board of the University of New Mexico Health Sciences Center. We identified all gestational choriocarcinoma cases in the Surveillance, Epidemiology and End Results 9 registries public-use database by selecting for International Classification of Diseases for Oncology (ICD) morphology codes specific for hydatidiform mole (9100/0), invasive mole (9100/1), partial mole (9103/0), choriocarcinoma (9100/3), and placental site trophoblastic disease (9104).8 Because the Surveillance, Epidemiology and End Results national registry includes only malignant cases (terminal digit /3), there were no partial moles. Cases identified with the typography code for placenta and female genital tract were included, except for 24 ovarian primaries (569), which were more likely to be of germ cell origin,8 and 13 placental site trophoblastic tumors. Choriocarcinomas arising in all other sites were not included. Over the 27-year period, there have been additions and modifications to Surveillance, Epidemiology and
VOL. 102, NO. 5, PART 1, NOVEMBER 2003
End Results coding, but variables exist for stage and race that have not been modified since the program’s inception, and these were selected for this analysis. Stage of disease was therefore defined as localized (confined to the organ of origin), regional (spread to adjacent organs or regional lymph node involvement), distant (discontinuous organ spread, distant lymph nodes), and unstaged (stage of disease unknown).9 Race was defined as white, black, other nonwhite populations, and unknown. The diagnostic method was determined with the Surveillance, Epidemiology and End Results diagnostic confirmation variable. The proportion of cases by year of diagnosis, method of diagnosis, and stage were analyzed by logistic regression. Overall, and for each sequential 5-year period, age-specific and age-adjusted incidence rates per 100,000 woman-years standardized to both the United States 2000 and 1970 census were calculated.10 Of note, regardless of the standard, the rates obtained were the same to within two decimal places, and age-adjusted standardization had no impact on P values for trends. Therefore, only rates with the United States 2000 standard are reported. Estimated annual percent change in incidence rates by race, stage, Surveillance, Epidemiology and End Results registry, and age at diagnosis was calculated by fitting a least-squares regression line to the natural logarithm of the rates, with calendar years as a regressor variable.11 Differences between trends between two time periods were compared by 2 test.11 Because multivariate analyses are not feasible with the use of the statistical program in the Surveillance, Epidemiology and End Results database, the “list option” in the Surveillance, Epidemiology and End Results database was used to abstract case-specific data, including age at diagnosis, race, year of diagnosis, geographic region, and histology, as well as survival in months and causes of death. Poisson regression (SAS/STAT User’s Guide version 8, SAS Institute, Cary NC, 1999) was then used to construct multivariate models of the effect of calendar time, age, race, stage, and registry and their potential interactions on incidence rates.12 Overall and 5-year relative survivals and 95% confidence intervals (CI) within two standard errors were calculated with the Surveillance, Epidemiology and End Results program software. Data regarding disease recurrence is not available. Cause of death in the Surveillance, Epidemiology and End Results data is based on death certificate data. The Surveillance, Epidemiology and End Results program can be used to analyze overall survival (alive versus dead), including or excluding cases without microscopic confirmation and those with no survival time (ie, the date of diagnosis was either the date of the last recorded information within Surveillance,
Smith et al
Gestational Choriocarcinoma Trends
979
Table 1. Gestational Choriocarcinoma Incidence Rates by Race (1973–1999) Race Blacks
No. of cases
% of all races
Woman-years
AAIR*
Standard error
Relative risk† (95% CI)
86
19.1
33,696,739
0.217
0.024
2.14 (1.60, 2.86)
Others
62
13.8
25,387,535
0.225
0.029
2.30 (1.67, 3.18)
Whites
301
66.9
253,334,776
0.110
0.006
1.00
1
0.2 312,419,050
0.133
0.006
Unknown All races
450
100
AAIR ⫽ age-adjusted incidence rates; CI ⫽ confidence interval. * Per 100,000 woman-years, adjusted to the US 2000 census using the SEERStat 4.23 statistical program. † Adjusted for age and for calendar time.
Epidemiology and End Results or the date of death), but multivariate analyses by age, race, and stage are not feasible. Therefore, the data were transferred into SAS, and the Life test procedure was used to calculate censored survival curves by race, stage, registry, and over calendar time, counting deaths from choriocarcinoma as events, and censoring for deaths from unknown causes or causes other than choriocarcinoma. Censored survival differences were compared by log-rank tests.13 Multivariate survival analyses of the effects of race, stage, age, registry, and calendar time were performed with Cox’s proportional hazards model; only those variables found to be statistically significant main effects were included in analyses of potential interaction effects.14 The Surveillance, Epidemiology and End Results registry includes five state and four city or county registries, and annual live birth data are available only for the five state registries. Therefore, the effects of live births and pregnancy denominators were not included in the Poisson model. An overall estimate of live birth and pregnancy denominators was made with the use of published estimates of the annual number by year and race in the United States between 1981 and 1991.15 Because the Surveillance, Epidemiology and End Results program represents population-based data that encompasses 9.6% of the US population, an assumption was made that women within the Surveillance, Epidemiology and End Results program accounted for 9.6% of all live births and pregnancies. The numerator for this estimate was therefore the absolute count of choriocarcinoma cases and woman-years within the Surveillance, Epidemiology and End Results program between 1981 and 1991. RESULTS After excluding ovarian and placental site tumors, a total of 450 cases registered between 1973 and 1999 consistent with gestational choriocarcinoma were identified, 401
980
Smith et al
Gestational Choriocarcinoma Trends
(89.1%) of which were pathologically confirmed. Pathologically confirmed cases included 399 cases with positive histology, and one case each with microscopic confirmation, method not specified, and positive exfoliative cytology. By morphology, 446 were choriocarcinomas (ICD code 9100). The four cases identified with the ICD code for choriocarcinomas with mixed germ cell elements (9101) were included as gestational choriocarcinoma cases because of their topography sites (placenta, three cases, uterus, one case). By sites of origin, 383 (85.1%) were identified with the topography code for placenta, and 67 (14.9%) with codes for female genital tract organs except ovary: cervix (530 –539) (two cases), corpus (540 –549) (40 cases), uterus, not otherwise specified (559) (12 cases), broad ligament (one case), fallopian tube (570) (nine cases), adnexa (574) (one case), and female genital tract not otherwise specified (579) (two cases). The average annual age-adjusted incidence rates per 100,000 woman-years were 0.133 per 100,000 womanyears for all races, and 0.110 for whites, 0.217 for blacks, and 0.225 for others (Table 1). Overall, the estimated annual percentage change in age-adjusted incidence rates over 29 years was ⫺49.7%, an estimated annual percentage decrease of 2.8% per year. Trends in the 5-year average age-adjusted incidence rates per 100,000 woman-years by race (1973–1997) are depicted in Figure 1A. Among whites, incidence rates declined by 62.3% over 27 years; among blacks and others, by 27.2% and 54.3%, respectively. Although incidence data are available for other racial groups, the denominator of woman-years is available only for the original Surveillance, Epidemiology and End Results code for race. When analyzed according to Surveillance, Epidemiology and End Results stage, the age-adjusted incidence rates per 100,000 woman-years were 0.042, 0.0013, 0.059, and 0.0181, respectively, for localized, regional, distant, and unstaged disease. Localized disease rates decreased by 54.5%, with an estimated annual percent-
OBSTETRICS & GYNECOLOGY
Figure 1. Trends in 5-year average gestational choriocarcinoma incidence rates per 100,000 by race (A) and stage (B) according to Surveillance, Epidemiology and End Results (1973–1997) program data. Smith. Gestational Choriocarcinoma Trends. Obstet Gynecol 2003.
age change of 2.82% (P ⫽ .002), regional disease rates increased by 300.7% (P ⬎ .05), distant disease rates decreased by 54.7%, with an estimated annual percentage change of 3.25% (P ⫽ .006), and unstaged disease rates decreased by 100% (P ⬍ .05). No estimated annual percentage change was calculated for regional and unstaged disease, because there were no significant trends. Trends in 5-year age-adjusted incidence rates by stage are depicted in Figure 1B. Differences in the distribution of choriocarcinoma incidence rates by stage and race were compared by Fisher exact tests. The difference was not significant (P ⫽ .09), and in subset analyses, a higher proportion of unstaged cases among whites (P ⫽ .02) were the only significant findings. The 27-year annualized incidence rates and relative risks (95% CI) by geographic registry and by age at diagnosis are depicted in Table 2. The lowest rates were found in Iowa and the highest in Utah. Relative risk
VOL. 102, NO. 5, PART 1, NOVEMBER 2003
(95% CI) was calculated with respect to Iowa (the relative risk for Iowa is 1.00). Significantly higher rates were found in Detroit, Seattle–Puget Sound, San Francisco– Oakland, Hawaii, and Utah. Rates in Iowa were not significantly different from those found in Connecticut, Atlanta, or New Mexico. Similarly, age-specific incidence rates in 5-year increments were calculated and compared. Because women 20 –24 years old have the highest fertility rate, the relative risk (95% CI) in other age groups were calculated with respect to women 20 –24 years old. While rates were highest in women 25–29 years old, they were not significantly different from those of women 20 –24, or 30 –34 years old (all P ⱖ .40). There were significantly lower rates in women 10 –14, and in all age groups over 40 (all P ⬎ .001), with marginally significant differences in women 15–19 (P ⫽ .03) and 35–39 (P ⫽ .04) compared with 20 –24 year olds.
Smith et al
Gestational Choriocarcinoma Trends
981
Table 2. Incidence Rates by SEER Registry and by Age at Diagnosis (1973–1999)
Registry
Count
AAIR*
Relative risk, (95% CI)†
Utah Hawaii San Francisco–Oakland Seattle–Puget Sound Detroit New Mexico Atlanta Connecticut Iowa
46 26 92 76 94 21 28 36 31
0.191 0.176 0.176 0.166 0.152 0.098 0.083 0.080 0.079
3.46 (1.94, 6.18) 1.04 (0.48, 2.27) 2.38 (1.39, 4.08) 2.11 (1.20, 3.71) 1.86 (1.07, 3.22) 1.32 (0.64, 2.72) 1.07 (0.55, 2.11) 1.43 (0.79, 2.58) 1.00
450
0.133
Totals
Age at diagnosis (y)
Count
ASIR
0–9 10–14 15–19 20–24 25–29 30–34 35–39 40–44 45–49 50–54 55–59 60⫹ Totals
0 5 62 91 105 79 56 24 15 5 3 5 450
0.000 0.022 0.263 0.371 0.407 0.309 0.240 0.115 0.084 0.031 0.021 0.009 0.144
Relative Risk, (95% CI)‡ 0 0.06 (0.02, 0.18) 0.71 (0.47, 1.07) 1.00 1.23 (0.88, 1.75) 0.99 (0.68, 1.43) 0.78 (0.52, 1.17) 0.41 (0.24, 0.70) 0.31 (0.17, 0.59) 0.09 (0.03, 0.28) 0.10 (0.03, 0.30) 0.01 (0.00, 0.07)
SEER ⫽ Surveillance, Epidemiology, and End Results; ASIR ⫽ age-specific incidence rate; other abbreviations as in Table 1. * Per 100,000 woman-years, adjusted to the US 2000 census. † Relative risks calculated, adjusting for a relative risk of 1.0 for Iowa. ‡ Relative risks calculated adjusting force for a relative risk of 1.0 for women 20 –24 years of age.
In the Poisson model evaluating incidence rates by age at diagnosis, race, stage, registry, and calendar time, all of these factors were found to be significant main effects (all P ⬍ .001). However, there were no statistically significant interaction effects identified between calendar time and age at diagnosis, race, stage, or registry (all P ⬎ .15). The effect of calendar time was found to be linear, and in the Poisson model adjusting for age at diagnosis and calendar time, the relative risk (95% CI) for choriocarcinoma in blacks and others compared with whites was 2.14 (1.60, 2.86) and 2.30 (1.67, 3.18), respectively (Table 1). Although there was a significant difference in trends in incidence rates over calendar time between registries (P ⫽ .04), no significant interaction effects by age at diagnosis (P ⫽ .76) or by race (P ⫽ .91) were identified. With the Poisson model, the estimated annual decrease in incidence rates (percent per year plus or minus standard error) was 3.3% (⫾ 0.9%) among whites (P ⫽ .004), 4.6% (⫾1.7%) among blacks (P ⫽ .006), and 3.4% (⫾ 1.9%) among other nonwhites (P ⫽ .08), respectively. The proportion of cases identified by positive laboratory or marker (nine cases, P ⫽ .004) but not those identified by radiographic studies (27 cases, P ⫽ .71), relative to the 399 histologically confirmed cases, significantly increased over the 27-year period. However, the increase in cases without (49 cases) compared with those with (401 cases) pathologic confirmation (histologic, cytologic, or microscopic, method not specified) was not significant (P ⫽ .46). Analyzed by stage, 139 (30.9%) of the pathologically confirmed cases had localized, 207 (46.0%) had regional or distant, and 55 (12.2%) had unstaged disease. Six (12.2%) cases without pathologic
982
Smith et al
Gestational Choriocarcinoma Trends
confirmation had localized disease, 38 (77.6%) had regional or distant metastases, and five (10.2%) were unstaged. The proportion of cases with distant (P ⫽ .49) or localized (P ⫽ .82) disease relative to the total cases did not change over calendar time. Gestational choriocarcinoma must be preceded by a gestational event, and ratios have almost always been reported with live birth and pregnancy denominators. With the use of the best published estimates for live birth and pregnancy denominators, the ratio of gestational choriocarcinoma cases per live births and clinically recognized pregnancies among all races was 1:26,000 live births and 1:41,000 pregnancies; 1:29,000 and 1:45,000 for whites, and 1:27,000 and 1:50,000 for blacks plus others (Table 3). The relative risk with live birth denominators was significantly higher for blacks and others than for whites (relative risk 1.10, 95% CI 1.02, 1.22, P ⫽ .01). However, with pregnancy denominators, the relative risks were not significantly different (relative risk for blacks and others, 1.06, 95% CI 0.98, 1.17, P ⫽ .18). We used the Surveillance, Epidemiology and End Results program to calculate 5-year observed and relative survivals by selecting for actively followed, malignant behavior and excluding second and later primaries, death certificate and autopsy cases, and adjusting for cumulative rates over one and increases in cumulative rates from a prior interval. The 27-year annualized 5-year relative survival (plus or minus two standard errors) was 89.5% (⫾ 3.15%), including those without microscopic confirmation (60 cases), and excluding those identified by second or later primary codes (25 cases), or based on death certificate or autopsy data (five cases). Stratified by race, the 5-year average relative
OBSTETRICS & GYNECOLOGY
Table 3. Estimated Choriocarcinoma Incidence Rates per Live Births and Pregnancy Within SEER (1981–1991)*
Race Whites Blacks and others Totals
Estimated Estimated mean mean Case to Case to live births in pregnancies in live birth pregnancy Count Woman-years SEER (SE) SEER (SE) ratio (SE) ratio (SE) 128
103,465,000
49
24,107,000
177
127,572,000
294,000 (3,700) 74,000 (2,000) 368,000 (5,600)
451,000 (5,400) 138,000 (3,800) 589,000 (9,000)
29,000 (3,400) 27,000 (6,800) 26,000 (3,000)
45,000 (5,300) 50,000 (12,600) 41,000 (5,000)
Relative Relative risk (95% risk (95% CI) per CI) per live births pregnancies 1.00
1.00
1.10 1.06 (1.02, 1.22) (0.98, 1.17)
SE ⫽ standard error; other abbreviations as in Tables 1 and 2. * Estimates based on annual number of live births and pregnancies in the US (1981–1991) (14). These estimates assume that 9.6% of total US live births and registries occurred within regions covered by the nine SEER registries.
survival was 92.1% (⫾ 3.36%) for whites, 81.0% (⫾ 9.23%) for blacks, and 87.6% (⫾ 9.28%) for other nonwhites (P ⫽ .02). The differences in odds ratios for survival among blacks versus whites was significant (2.69; 95% CI 1.23, 5.74, P ⫽ .006), but not for blacks versus others (1.73; 95% CI 0.61, 5.42, P ⫽ .37) or for others versus whites (1.54; 95% CI 0.53, 3.98, P ⫽ .46). According to case-specific data abstracted by stage, race, registry, age, survival in months, and cause of death based on death certificate information, 378 patients (84.0%) were listed as alive, and there were 72 deaths— 44 (9.8%) from choriocarcinoma, 20 (4.4%) from known causes other than choriocarcinoma, and eight (1.8%) for which the cause was unknown. Twenty (45.5%) deaths from choriocarcinoma occurred before 12 months of follow-up, and in five cases the survival time was less than 1 month. Survival by race (Figure 2), censored for deaths from intercurrent illnesses other than choriocarcinoma (unknown, and causes of death other than choriocarcinoma), showed significant differences (whites 92.4%, blacks 84.9%, others 87.1%, P ⫽ .045). Censored survival decreased with advancing stage of disease (localized 94.5%, regional 92.9%, distant 87.1%, P ⫽ .02) and with increasing age at diagnosis
(10 –20 years 97.0%; 20 –29 years, 87.2%; 30 –39 years, 93.3%; 40 – 49 years, 87.2%; 50 –59 years, 87.5%, and 60 or more years, 60.0%; P ⫽ .017). Censored survival was highest in Iowa (96.7%) and lowest in Hawaii (84.6%), but differences in survival between registries were not significant (P ⫽ .61). The median censored survival was 115 months (⫾ 96 months [standard error], range 0 –205 months). We analyzed the effect of calendar time on survival, subdividing the cases into three 9-year periods and censoring the analysis with the median follow-up for the last time period, 42 months, so that there were comparable follow-up intervals for each group. The difference (1973–1981, 174 cases, 23 deaths, survival rate 86.8%; 1982–1991, 145 cases, 12 deaths, survival rate 91.7%; 1992–1999, 130 cases, four deaths, survival rate 96.9%, log-rank P ⫽ .01) was statistically significant. In the Cox’s proportional hazards model (hazards ratio, 95% CI), calendar time, age at diagnosis, and race were significant main effects on survival. There was no interaction effect with calendar time and stage (P ⫽ .38). The effect of calendar time by age of diagnosis was significant (P ⫽ .01), but only for the 1973–1981 period (hazards ratio 0.51, 95% CI 0.33, 0.80, P ⫽ .004), and was not
Figure 2. Censored survival by race according to Surveillance, Epidemiology and End Results (1973–1999) program data. Smith. Gestational Choriocarcinoma Trends. Obstet Gynecol 2003.
VOL. 102, NO. 5, PART 1, NOVEMBER 2003
Smith et al
Gestational Choriocarcinoma Trends
983
significant by race (P ⫽ .90) or stage (P ⫽ .38). The effect of age at diagnosis was significant among whites (hazard ratio 1.06, 95% CI 1.03, 1.10), which indicates that mortality increased by 6.2% for each additional 10 years of age. However, age at diagnosis did not impact survival for blacks (hazard ratio 1.0, 95% CI 0.94, 1.06) or others (hazard ratio 0.97, 95% CI 0.90, 1.05). The effect of age at diagnosis did not change when we adjusted for either registry or stage of disease. DISCUSSION According to population-based data recorded over 27 years (Surveillance, Epidemiology and End Results, 1973–1999) and the best available published estimates of live births and pregnancies in the United States, the age-adjusted incidence rate for gestational choriocarcinoma per 100,000 woman-years for all races was 0.133, one in 25,674 live births, and one in 41,094 pregnancies. Over all, there was a 49.7% decrease, with an annual percentage decrease of 2.8% per year. In the Poisson model, calendar time, age at diagnosis, stage, registry, and race were all significant main effects, but no significant interaction effects between these variables and calendar time were identified. Although trends by registry were significant, because these data could not be controlled for live birth or pregnancy denominators, no conclusions can be drawn regarding the effects of fertility trends by race, age, or registry. Age-adjusted incidence rates were higher among blacks and other nonwhites compared with whites, but only blacks were found to have significantly poorer survival rates. Brinton et al,5 using Surveillance, Epidemiology and End Results data, have previously reported that the rates of choriocarcinoma, after adjusting for race and estimated birth rates, were 2.1- and 1.8-fold higher among blacks and other races compared with whites. Among women of Asian descent, these findings were as expected, because these women are known to have a higher incidence of hydatidiform mole,4 and most choriocarcinoma cases are preceded by a molar pregnancy.6 However, although the risk of hydatidiform mole in blacks is less well characterized, most studies conducted within the United States have not found significantly higher rates in blacks.16 –18 In Rhode Island, molar pregnancy rates from 1956 through 1965 were higher among blacks than for whites, but the difference was not statistically significant, possibly because of the small number of cases in blacks.17 Similarly, a case– control study that compared molar pregnancy rates among black, white, and Latin American women in Los Angeles also found no racial differences, but again, the sample size was small.18 Of interest, Hayashi et al16 estimated that the ratio of
984
Smith et al
Gestational Choriocarcinoma Trends
gestational trophoblastic disease for US blacks (58.2 per 100,000 pregnancies), was approximately half that of whites and others (118 and 117.9 per 100,000 pregnancies, respectively). This is especially intriguing, because it implies that the higher incidence of choriocarcinoma among blacks might be a reflection of greater risk for choriocarcinoma after a nonmolar pregnancy. This could potentially account for lower survival among blacks, because the prognosis for choriocarcinoma after a normal gestation is poorer.1 Reduced access to surveillance might also account for higher incidence and mortality rates among blacks.19 Although persistent gestational trophoblastic disease has an excellent prognosis, noncompliance with follow-up can lead to a delay in diagnosis, and inconsistent compliance with treatment can lead to the development of chemotherapy-resistant disease.1,19 Among blacks, survival rates are poorer for cervical, breast, endometrial, and colorectal carcinoma, and in some cases these disparities persist after correcting for stage of disease.20 –25 It is now recognized that differences in survival by race are a reflection of socioeconomic status and class, because blacks are disproportionately represented among the poor and disadvantaged.25,26 Indigent women, especially those belonging to ethnic and racial minority groups, are less likely to fully participate in breast and cervical carcinoma screening programs,27 and postmolar surveillance, including routine human chorionic gonadotropin determinations and timely delivery of chemotherapy are also poorer.19 With respect to other racial groups, data is also limited regarding trends in gestational trophoblastic disease incidence rates. In the United Kingdom, where centralized registration of all gestational trophoblastic disease has been standard practice since 1973, the molar pregnancy ratio is 1.54 per 1000 live births.28 In Sweden, which also maintains a population-based registry, there were no significant trends in incidence rates between 1975 and 1988.29 Population-based data indicate that there are higher gestational trophoblastic disease rates among nonwhite Hispanic,30,31 American Indian or Eskimo,32,33 and Asian34 –37 women. Alaskan natives32 and Southwestern American Indians33 have rates three to four times greater than United States whites. Japanese and Philippino women in Hawaii also have higher rates than do Hawaiian whites.38,39 The significance of Hispanic descent is uncertain. In New Mexico and Los Angeles,18,33 Hispanic women were not found to be at increased risk, whereas hospital-based studies from Mexico and Central America have consistently reported higher rates.30,31 In Asian populations, both populationbased and hospital-based studies indicate that rates are declining.4,34 –37 In Korea, with live births as the denom-
OBSTETRICS & GYNECOLOGY
inator, a 17.5-fold decline in all gestational trophoblastic disease incidence rates was recently reported (from 40.2 [1971–1975] to 2.3 [1991–1995]).34 The decline in incidence rates found in the Surveillance, Epidemiology and End Results data might be a reflection of more wide-spread use of standardized protocols for surveillance, clinical evaluation, and therapy for low-risk gestational trophoblastic disease, or might be a reflection of lower conception rates, especially in older women. Differences in rates by registry could not be accounted for by the effects of age at diagnosis, race, or stage. These unexpected findings might indicate that other factors not evaluated in this study, including registry-specific conception rates, environmental influences (climate, diet, other exposures), and possibly differences in access to quality medical care, account for differences by registry. Ascertainment bias might also contribute to the perceived decline in incidence rates. The proportion of cases without microscopic confirmation (49 of 450 cases, 10.9%), was the same as that of the previous Surveillance, Epidemiology and End Results study (20 of 203 cases, 9.7%).5 Although there was an increase in the number of cases diagnosed based on laboratory markers, the proportion of histologically confirmed cases did not change over time. Analysis of these data within the context of other studies is limited, because most population-based studies include all gestational trophoblastic disease variants, and choriocarcinoma cases account for only a small percentage. The vast majority of gestational trophoblastic disease cases are managed based on the number and location of metastases and the human chorionic gonadotropin titer instead of a tissue diagnosis. Therefore, an analysis limited primarily to histologically confirmed and metastatic cases potentially underestimates the total number of cases. Because the disease is rare, any variation in criteria for case selection could influence rate determinations. However, because the proportion of cases identified with regional/metastatic disease and with microscopic confirmation did not change over calendar time, it is probable that this analysis accurately reflects Surveillance, Epidemiology and End Results registry trends. Other limitations to this study deserve acknowledgment. A comparison of trends for all gestational trophoblastic disease cases is not possible with Surveillance, Epidemiology and End Results data because partial or invasive mole variants are not included in the public-use database. Most registries do not maintain data for molar pregnancies, and these cases are categorically excluded from the Surveillance, Epidemiology and End Results public-use database by coding definitions. Nevertheless, we have attempted to estimate the ratio of choriocarci-
VOL. 102, NO. 5, PART 1, NOVEMBER 2003
noma cases relative to gestational trophoblastic disease within the Surveillance, Epidemiology and End Results registry, extrapolating from an analysis of all gestational trophoblastic disease cases in one state registry that were systematically collected over 25 years. Given 939 incident gestational trophoblastic disease cases including 22 choriocarcinomas, the ratio of choriocarcinoma to gestational trophoblastic disease (excluding placental site trophoblastic tumors) is 1:42.7.33 In the same period, there were 422 choriocarcinoma cases in the Surveillance, Epidemiology and End Results study. If this ratio is reasonably representative of national registry incidence rates, hypothetically there were 18,019 cases of gestational trophoblastic disease. This estimation is imprecise, and overlooks the influence of geographic region, race, and number of cases arising from a normal antecedent pregnancy. However, this is the best estimate available, which emphasizes the potential for ascertainment bias. Of note, the three unpublished choriocarcinoma cases comprising the New Mexico numerator were found with the SEERStat program when topography codes for uterus (two case) and fallopian tube (one case) were used. The choice of cases selected for inclusion could influence incidence and survival rates. Unlike the previous Surveillance, Epidemiology and End Results study, all ovarian cases were excluded, and it is possible that some of these were gestational choriocarcinomas. The two cases identified as cervical in origin might have been primary cervical choriocarcinomas.40 However, because this condition is exceedingly rare, it is more likely that they were gestational choriocarcinomas that were coded as cervical in origin because the cervix was the site of biopsy. Similarly, the one case identified as arising in the adnexa may have been a germ cell neoplasm. All three of these cases were histologically confirmed and were unstaged (two cases) or had distant disease (one case). There were five, eight, and 15 cases, respectively, in women over 60, 50 –59, and 45– 49 years of age—women unlikely to have recently been pregnant. It is feasible that cases found in some or all postmenopausal women might be a reflection of delayed-onset choriocarcinoma, misinterpretation of pathologic specimens, or misclassification of other neoplasms as choriocarcinoma. The staging system available in the Surveillance, Epidemiology and End Results database does not take into account the number and location of metastases, which limits its usefulness, because women with pulmonary metastases alone have an excellent prognosis, but survival is poorer in those with liver, brain, and other metastatic sites. The conception-based denominators can also influence rate determinations. In this study, the best available published estimates of live birth and pregnancy for the
Smith et al
Gestational Choriocarcinoma Trends
985
entire US population were used because there is no centralized data for registry-specific fertility denominators.15 Therefore, no statement can be made regarding the effects of reproductive trends over time and between registries. The alternative approach, to include live birth data from the five state registries with published live birth rates, ignores the fact that fertility rates are likely to vary across registries. In a recent analysis of all gestational trophoblastic disease in which one stage registry was used, age-adjusted incidence rate determinations were found to be a more reliable estimate than rates based on live births or pregnancies.33 Nonetheless, when the best published estimates for live births and pregnancies in the United States are used, the ratios per live births and pregnancy reported in this and the previous Surveillance, Epidemiology and End Results study are similar.5 The absolute number of conceptions is the ideal denominator to use when comparing rates, but the number of conceptions is rarely known, and the number of clinically recognized pregnancies greatly underestimates the total number of conceptions. In this study, we calculated incidence ratios by using women-years, live births, and clinically recognized pregnancies according to published denominator data for the middle 11-year period (1981–1991) and corresponding cases from the Surveillance, Epidemiology and End Results study. Our results are in agreement with those of Brinton et al,5 who reported that live birth denominators overestimated the ratio by 40%. In this study, significant differences in ratios by race per live births but not by pregnancies were found. This suggests that higher conception rates alone do not account for the observed differences. In conclusion, data derived from the Surveillance, Epidemiology and End Results study indicate that gestational choriocarcinoma incidence rates in the United States seem to be declining and that survival has improved. There are significant differences in risk by race, stage, age at diagnosis, and geographic region. Although both blacks and others have higher incidence rates compared with whites, blacks alone are at greater risk for death. Factors such as access to contraception or timely health care, as well as genetic, dietary and other environmental influences, warrant further investigation.
REFERENCES 1. Berkowitz RS, Goldstein DP. Chorionic tumors [review]. N Engl J Med 1996;335:1740–8. 2. Lage JM, Wolf NG. Gestational trophoblastic disease. New approaches to diagnosis. Clin Lab Med 1995;15: 631–64.
986
Smith et al
Gestational Choriocarcinoma Trends
3. Aziz MF, Kampono N, Moegni EM, Sjamsuddin S, Barnas B, Samil RS. Epidemiology of gestational trophoblastic neoplasm at the Dr. Cipto Mangunkusumo Hospital, Jakarta, Indonesia. Adv Exp Med Biol 1984;176:165–75. 4. Palmer JR. Advances in the epidemiology of gestational trophoblastic disease. J Reprod Med 1994;39:155–62. 5. Brinton LA, Bracken MB, Connelly RR. Choriocarcinoma incidence in the United States. Am J Epidemiol 1986;123:1094–100. 6. Hertig AT, Mansell H. Tumors of the female sex organs, part I: Hydatidiform mole and choriocarcinoma. In: Atlas of tumor pathology, section IX, fascicle 33. Washington: Armed Forces Institute of Pathology, 1956. 7. Shanmugaratnam K, Muir CS, Tow SH, Cheng WC, Christine B, Pedersen E. Rates per 100,000 births and incidence of choriocarcinoma and malignant mole in Singapore Chinese and Malays: Comparison with Connecticut, Norway, and Sweden. Int J Cancer 1971;8:165–75. 8. Fritz A, Percy C, Jack A, Shanmugaratnam K, Sobin L, Porkin DM, Whelan S, Van Holten V, Muir C, eds. International classification of diseases for oncology (ICD0), 3rd ed. Geneva: World Health Organization, 2000. 9. Cunningham J, Ries L, Hankey B, Seiffert J, Lyles B, Shambaugh E, et al, eds. The SEER program code manual, revised edition. National Institutes of Health (PHS) 92–1999. Bethesda, Maryland: National Institutes of Health, 1999. 10. Smith PG. Comparison between registries: Age-standardized rates. In: Parkin DM, Muir CS, Whelan SL, Gao Y-T, Ferlay J, Powel J, eds. Cancer incidence in five continents, vol. VI (IARC scientific publication no. 120). Lyon, France: IARC, 1992. 11. Kleinbaum DG, Kupper LL, Muller KE. Applied regression analysis and other multivariate methods. North Sciutate, Massachusetts: Duxbury Press, 1988. 12. McCullagh P, Nelder JA. Generalized linear models, 2nd ed. London: Chapman & Hall, 1989. 13. Collett D. Modelling survival data in medical research. London: Chapman & Hall, 1994. 14. Kalbfleisch JD, Prentice RL. The statistical analysis of failure time data. New York: John Wiley & Sons, 1980. 15. Saraiya M, Berg CJ, Shulman H, Green CA, Atrash HK. Estimates of the annual number of clinically recognized pregnancies in the United States, 1981-1991. Am J Epidemiol 1999;149:1025–9. 16. Hayashi K, Bracken MB, Freeman DH Jr, Hellenbrand K. Hydatidiform mole in the United States (1970-1977): A statistical and theoretical analysis. Am J Epidemiol 1982; 15:67–77. 17. Yen S, MacMahon B. Epidemiologic features of trophoblastic disease. Am J Obstet Gynecol 1968;101:126–32. 18. Westerhout FC Jr, Morel ES, Slate WG. Observations on 138 molar pregnancies. Am J Obstet Gynecol 1969;103: 56–9.
OBSTETRICS & GYNECOLOGY
19. Massad LS, Abu-Rustum NR, Less SS, Renta V. Poor compliance with postmolar surveillance and treatment protocols by indigent women. Obstet Gynecol 2000;96: 940–4. 20. Dignam JJ. Differences in breast cancer prognosis among African-American and Caucasian women. CA Cancer J Clin 2000;50:50–64. 21. Hicks ML, Phillips JL, Parham G, Andrews N, Jones WB, Shingleton HM, et al. The National Cancer Data Base report on endometrial carcinoma in African-American women. Cancer 1998;83:2629–37. 22. Digman JJ, Ye Y, Colangelo L, Smith R, Manounas EP, Wieand HS, et al. Prognosis after rectal cancer in blacks and whites participating in adjuvant therapy randomized trials. J Clin Oncol 2003;21:413–20. 23. Howell EA, Chen YT, Concato J. Differences in cervical cancer mortality among black and white women. Obstet Gynecol 1999;94:509–15. 24. Grisby PW, Hall-Daniels L, Baker S, Perez CA. Comparison of clinical outcome in black and white women treated with radiotherapy for cervical carcinoma. Gynecol Oncol 2000;79:357–61. 25. Cross CK, Harris J, Recht A. Race, socioeconomic status, and breast carcinoma in the U.S.: What have we learned from clinical studies. Cancer 2002;95:1988–99. 26. O’Malley CD, Le GM, Laser SL, Shema SJ, West DW. Socioeconomic status and breast carcinoma survival in four racial/ethnic groups. Cancer 2003;97:1303–11. 27. Katz SJ, Hofer TP. Socioeconomic disparities in preventive care persist despite universal coverage: Breast and cervical cancer screening in Ontario and the United States. JAMA 1994;272:530–4. 28. Bagshawe KD, Dent J, Webb J. Hydatidiform mole in England and Wales 1973-1983. Lancet 1986;11:673–7. 29. Flam F, Lindsted-Lundstro¨m V, Rutqvist LE. Incidence of gestational trophoblastic disease in Stockholm county, 1975-1988. Eur J Epidemiol 1992;8:173–7. 30. Marquez-Monter H, de la Vega GA, Ridaura C, Robles M. Gestational choriocarcinoma in the general hospital of Mexico: Analysis of forty cases. Cancer 1968;22:91–8.
VOL. 102, NO. 5, PART 1, NOVEMBER 2003
31. Aguero O, Kizer S, Pinedo G. Hydatidiform mole in Concepcion Palacios Maternity Hospital. Am J Obstet Gynecol 1973;116:1117–20. 32. Martin PM. Hydatidiform mole in Nigeria, Iran, and Alaska. III. High frequency of hydatidiform mole in native Alaskans. Int J Gynaecol Obstet 1978;15:395–6. 33. Smith HO, Hilgers RD, Bedrick EJ, Qualls CR, Wiggins CL, Rayburn RF, et al. Ethnic differences in risk for gestational trophoblastic disease in New Mexico: A 25-year population-based study. Am J Obstet Gynecol 2003;188:357–66. 34. Kim SJ, Bae SN, Kim JH, Kim CJ, Han KT, Chung JK, et al. Epidemiology and time trends of gestational trophoblastic disease in Korea. Int J Gynecol Obstet 1998;60: S33–8. 35. Hando T, Ohno M, Kurose T. Recent aspects of gestational trophoblastic disease in Japan. Int J Gynecol Obstet 1998;60:S71–6. 36. Sakumoto K, Kigashi M, Kanazawa K. Hydatidiform mole in Okinawa Islands and mainland Japan. Int J Gynaecol Obstet 1999;64:309–10. 37. Sivanesartnam V. The management of gestational trophoblastic disease in developing countries such as Malaysia. Int J Gynaecol Obstet 1998;60:S105–9. 38. McCorriston CC. Racial incidence of hydatidiform mole. Am J Obstet Gynecol 1968;101:377–82. 39. Matsuura J, Chiu D, Jacobs PA, Szulman AE. Complete hydatidiform mole in Hawaii: An epidemiological study. Genetic Epidemiol 1984;1:271–84. 40. Herts BR, Yee JM, Porges RF. Primary cervical choriocarcinoma: Case report and review of the literature. J Ultrasound Med 1993;12:59–62. Address reprint requests to: Harriet O. Smith, MD, University of New Mexico Health Sciences Center, Department of Obstetrics and Gynecology, Division of Gynecologic Oncology, 2211 Lomas Boulevard, NE, Albuquerque, NM 87131-5286; Email: [email protected]. Received December 26, 2002. Received in revised form June 4, 2002. Accepted June 9, 2002.
Smith et al
Gestational Choriocarcinoma Trends
987
978
Gestational choriocarcinoma is a relatively rare and highly malignant variant of gestational trophoblastic disease.1– 4 Although choriocarcinoma might be preceded by any gestational event, hydatidiform mole is the most common precursor. Throughout the world, incidence rates for choriocarcinoma differ widely.5 For example, in Europe and North America, choriocarcinomas are reported to affect one in every 30,000 – 40,000 pregnancies, and one in 40 molar pregnancies,6 whereas in Southeast Asia, rates as high as one in every 500 –3000 pregnancies have been reported.7 Comparisons of gestational trophoblastic disease incidence rates among different regions of the world are limited by the methods used to compute rates. The denominators most often used to compute rates (live births, pregnancies, deliveries) are hospital-based, whereas the numerator, incident cases, tends to reflect referral patterns within a given continent or region. Although of greater clinical relevance, it is even more difficult to compare choriocarcinoma incidence rates, because pathologic confirmation is not always feasible, and the disease is rare.5 Population-based data that can be applied consistently between regions is necessary to consistently determine rates and to accurately compare differences. The best available published estimate of choriocarcinoma incidence rates in the United States was derived from data obtained from the Surveillance, Epidemiology and End Results program (1973–1982).5 Differences in age-adjusted incidence rates by age at diagnosis and race (black, white, other) were reported, but there were insufficient data to analyze trends in incidence or survival rates. The objective of this study was to analyze trends in incidence and survival rates for gestational choriocarcinoma with the most recent version of the Surveillance, Epidemiology and End Results program’s cancer incidence publicuse database (SEERSTAT 4.2; Surveillance, Epidemiology and End Results 9 Registries public-use, November 2001 submission), which includes all cases recorded between 1973 and 1999. Although the original nine Surveillance, Epidemiology and End Results registries
VOL. 102, NO. 5, PART 1, NOVEMBER 2003 © 2003 by The American College of Obstetricians and Gynecologists. Published by Elsevier.
0029-7844/03/$30.00 doi:10.1016/S0029-7844(03)00669-0
record carcinoma cases for only approximately 10% of the US population, data derived from the Surveillance, Epidemiology and End Results registry is regarded as the best estimate of cancer incidence and mortality trends in the United States. MATERIAL AND METHODS The Surveillance, Epidemiology and End Results program is a set of geographically defined, population-based central registries for the United States, operated by local nonprofit organizations under contract to the National Cancer Institute. The original nine Surveillance, Epidemiology and End Results registries have been in operation since 1973. Each year, certified abstractors review the medical records and pathology reports of all hospitals and laboratories and identify all incident invasive carcinoma cases. Only cases affecting residents within the individual Surveillance, Epidemiology and End Results registry are included. Ethnicity is assigned according to the best available information, usually by the racial preference stated on the hospital charts, on the basis of identifiers in the record, or by surname. Because there is no mechanism for centralized pathology review, the potential for variability in experience and expertise of pathologists and clinicians involved is substantial—a known limitation of this data set. Registry data are submitted electronically without personal identifiers to the National Cancer Institute on a biannual basis. Surveillance, Epidemiology and End Results public-use data and software are provided at no charge to individuals interested in conducting scientific research (http://seer. cancer.gov/seerstat). Approval to perform this study was obtained from the internal review board of the University of New Mexico Health Sciences Center. We identified all gestational choriocarcinoma cases in the Surveillance, Epidemiology and End Results 9 registries public-use database by selecting for International Classification of Diseases for Oncology (ICD) morphology codes specific for hydatidiform mole (9100/0), invasive mole (9100/1), partial mole (9103/0), choriocarcinoma (9100/3), and placental site trophoblastic disease (9104).8 Because the Surveillance, Epidemiology and End Results national registry includes only malignant cases (terminal digit /3), there were no partial moles. Cases identified with the typography code for placenta and female genital tract were included, except for 24 ovarian primaries (569), which were more likely to be of germ cell origin,8 and 13 placental site trophoblastic tumors. Choriocarcinomas arising in all other sites were not included. Over the 27-year period, there have been additions and modifications to Surveillance, Epidemiology and
VOL. 102, NO. 5, PART 1, NOVEMBER 2003
End Results coding, but variables exist for stage and race that have not been modified since the program’s inception, and these were selected for this analysis. Stage of disease was therefore defined as localized (confined to the organ of origin), regional (spread to adjacent organs or regional lymph node involvement), distant (discontinuous organ spread, distant lymph nodes), and unstaged (stage of disease unknown).9 Race was defined as white, black, other nonwhite populations, and unknown. The diagnostic method was determined with the Surveillance, Epidemiology and End Results diagnostic confirmation variable. The proportion of cases by year of diagnosis, method of diagnosis, and stage were analyzed by logistic regression. Overall, and for each sequential 5-year period, age-specific and age-adjusted incidence rates per 100,000 woman-years standardized to both the United States 2000 and 1970 census were calculated.10 Of note, regardless of the standard, the rates obtained were the same to within two decimal places, and age-adjusted standardization had no impact on P values for trends. Therefore, only rates with the United States 2000 standard are reported. Estimated annual percent change in incidence rates by race, stage, Surveillance, Epidemiology and End Results registry, and age at diagnosis was calculated by fitting a least-squares regression line to the natural logarithm of the rates, with calendar years as a regressor variable.11 Differences between trends between two time periods were compared by 2 test.11 Because multivariate analyses are not feasible with the use of the statistical program in the Surveillance, Epidemiology and End Results database, the “list option” in the Surveillance, Epidemiology and End Results database was used to abstract case-specific data, including age at diagnosis, race, year of diagnosis, geographic region, and histology, as well as survival in months and causes of death. Poisson regression (SAS/STAT User’s Guide version 8, SAS Institute, Cary NC, 1999) was then used to construct multivariate models of the effect of calendar time, age, race, stage, and registry and their potential interactions on incidence rates.12 Overall and 5-year relative survivals and 95% confidence intervals (CI) within two standard errors were calculated with the Surveillance, Epidemiology and End Results program software. Data regarding disease recurrence is not available. Cause of death in the Surveillance, Epidemiology and End Results data is based on death certificate data. The Surveillance, Epidemiology and End Results program can be used to analyze overall survival (alive versus dead), including or excluding cases without microscopic confirmation and those with no survival time (ie, the date of diagnosis was either the date of the last recorded information within Surveillance,
Smith et al
Gestational Choriocarcinoma Trends
979
Table 1. Gestational Choriocarcinoma Incidence Rates by Race (1973–1999) Race Blacks
No. of cases
% of all races
Woman-years
AAIR*
Standard error
Relative risk† (95% CI)
86
19.1
33,696,739
0.217
0.024
2.14 (1.60, 2.86)
Others
62
13.8
25,387,535
0.225
0.029
2.30 (1.67, 3.18)
Whites
301
66.9
253,334,776
0.110
0.006
1.00
1
0.2 312,419,050
0.133
0.006
Unknown All races
450
100
AAIR ⫽ age-adjusted incidence rates; CI ⫽ confidence interval. * Per 100,000 woman-years, adjusted to the US 2000 census using the SEERStat 4.23 statistical program. † Adjusted for age and for calendar time.
Epidemiology and End Results or the date of death), but multivariate analyses by age, race, and stage are not feasible. Therefore, the data were transferred into SAS, and the Life test procedure was used to calculate censored survival curves by race, stage, registry, and over calendar time, counting deaths from choriocarcinoma as events, and censoring for deaths from unknown causes or causes other than choriocarcinoma. Censored survival differences were compared by log-rank tests.13 Multivariate survival analyses of the effects of race, stage, age, registry, and calendar time were performed with Cox’s proportional hazards model; only those variables found to be statistically significant main effects were included in analyses of potential interaction effects.14 The Surveillance, Epidemiology and End Results registry includes five state and four city or county registries, and annual live birth data are available only for the five state registries. Therefore, the effects of live births and pregnancy denominators were not included in the Poisson model. An overall estimate of live birth and pregnancy denominators was made with the use of published estimates of the annual number by year and race in the United States between 1981 and 1991.15 Because the Surveillance, Epidemiology and End Results program represents population-based data that encompasses 9.6% of the US population, an assumption was made that women within the Surveillance, Epidemiology and End Results program accounted for 9.6% of all live births and pregnancies. The numerator for this estimate was therefore the absolute count of choriocarcinoma cases and woman-years within the Surveillance, Epidemiology and End Results program between 1981 and 1991. RESULTS After excluding ovarian and placental site tumors, a total of 450 cases registered between 1973 and 1999 consistent with gestational choriocarcinoma were identified, 401
980
Smith et al
Gestational Choriocarcinoma Trends
(89.1%) of which were pathologically confirmed. Pathologically confirmed cases included 399 cases with positive histology, and one case each with microscopic confirmation, method not specified, and positive exfoliative cytology. By morphology, 446 were choriocarcinomas (ICD code 9100). The four cases identified with the ICD code for choriocarcinomas with mixed germ cell elements (9101) were included as gestational choriocarcinoma cases because of their topography sites (placenta, three cases, uterus, one case). By sites of origin, 383 (85.1%) were identified with the topography code for placenta, and 67 (14.9%) with codes for female genital tract organs except ovary: cervix (530 –539) (two cases), corpus (540 –549) (40 cases), uterus, not otherwise specified (559) (12 cases), broad ligament (one case), fallopian tube (570) (nine cases), adnexa (574) (one case), and female genital tract not otherwise specified (579) (two cases). The average annual age-adjusted incidence rates per 100,000 woman-years were 0.133 per 100,000 womanyears for all races, and 0.110 for whites, 0.217 for blacks, and 0.225 for others (Table 1). Overall, the estimated annual percentage change in age-adjusted incidence rates over 29 years was ⫺49.7%, an estimated annual percentage decrease of 2.8% per year. Trends in the 5-year average age-adjusted incidence rates per 100,000 woman-years by race (1973–1997) are depicted in Figure 1A. Among whites, incidence rates declined by 62.3% over 27 years; among blacks and others, by 27.2% and 54.3%, respectively. Although incidence data are available for other racial groups, the denominator of woman-years is available only for the original Surveillance, Epidemiology and End Results code for race. When analyzed according to Surveillance, Epidemiology and End Results stage, the age-adjusted incidence rates per 100,000 woman-years were 0.042, 0.0013, 0.059, and 0.0181, respectively, for localized, regional, distant, and unstaged disease. Localized disease rates decreased by 54.5%, with an estimated annual percent-
OBSTETRICS & GYNECOLOGY
Figure 1. Trends in 5-year average gestational choriocarcinoma incidence rates per 100,000 by race (A) and stage (B) according to Surveillance, Epidemiology and End Results (1973–1997) program data. Smith. Gestational Choriocarcinoma Trends. Obstet Gynecol 2003.
age change of 2.82% (P ⫽ .002), regional disease rates increased by 300.7% (P ⬎ .05), distant disease rates decreased by 54.7%, with an estimated annual percentage change of 3.25% (P ⫽ .006), and unstaged disease rates decreased by 100% (P ⬍ .05). No estimated annual percentage change was calculated for regional and unstaged disease, because there were no significant trends. Trends in 5-year age-adjusted incidence rates by stage are depicted in Figure 1B. Differences in the distribution of choriocarcinoma incidence rates by stage and race were compared by Fisher exact tests. The difference was not significant (P ⫽ .09), and in subset analyses, a higher proportion of unstaged cases among whites (P ⫽ .02) were the only significant findings. The 27-year annualized incidence rates and relative risks (95% CI) by geographic registry and by age at diagnosis are depicted in Table 2. The lowest rates were found in Iowa and the highest in Utah. Relative risk
VOL. 102, NO. 5, PART 1, NOVEMBER 2003
(95% CI) was calculated with respect to Iowa (the relative risk for Iowa is 1.00). Significantly higher rates were found in Detroit, Seattle–Puget Sound, San Francisco– Oakland, Hawaii, and Utah. Rates in Iowa were not significantly different from those found in Connecticut, Atlanta, or New Mexico. Similarly, age-specific incidence rates in 5-year increments were calculated and compared. Because women 20 –24 years old have the highest fertility rate, the relative risk (95% CI) in other age groups were calculated with respect to women 20 –24 years old. While rates were highest in women 25–29 years old, they were not significantly different from those of women 20 –24, or 30 –34 years old (all P ⱖ .40). There were significantly lower rates in women 10 –14, and in all age groups over 40 (all P ⬎ .001), with marginally significant differences in women 15–19 (P ⫽ .03) and 35–39 (P ⫽ .04) compared with 20 –24 year olds.
Smith et al
Gestational Choriocarcinoma Trends
981
Table 2. Incidence Rates by SEER Registry and by Age at Diagnosis (1973–1999)
Registry
Count
AAIR*
Relative risk, (95% CI)†
Utah Hawaii San Francisco–Oakland Seattle–Puget Sound Detroit New Mexico Atlanta Connecticut Iowa
46 26 92 76 94 21 28 36 31
0.191 0.176 0.176 0.166 0.152 0.098 0.083 0.080 0.079
3.46 (1.94, 6.18) 1.04 (0.48, 2.27) 2.38 (1.39, 4.08) 2.11 (1.20, 3.71) 1.86 (1.07, 3.22) 1.32 (0.64, 2.72) 1.07 (0.55, 2.11) 1.43 (0.79, 2.58) 1.00
450
0.133
Totals
Age at diagnosis (y)
Count
ASIR
0–9 10–14 15–19 20–24 25–29 30–34 35–39 40–44 45–49 50–54 55–59 60⫹ Totals
0 5 62 91 105 79 56 24 15 5 3 5 450
0.000 0.022 0.263 0.371 0.407 0.309 0.240 0.115 0.084 0.031 0.021 0.009 0.144
Relative Risk, (95% CI)‡ 0 0.06 (0.02, 0.18) 0.71 (0.47, 1.07) 1.00 1.23 (0.88, 1.75) 0.99 (0.68, 1.43) 0.78 (0.52, 1.17) 0.41 (0.24, 0.70) 0.31 (0.17, 0.59) 0.09 (0.03, 0.28) 0.10 (0.03, 0.30) 0.01 (0.00, 0.07)
SEER ⫽ Surveillance, Epidemiology, and End Results; ASIR ⫽ age-specific incidence rate; other abbreviations as in Table 1. * Per 100,000 woman-years, adjusted to the US 2000 census. † Relative risks calculated, adjusting for a relative risk of 1.0 for Iowa. ‡ Relative risks calculated adjusting force for a relative risk of 1.0 for women 20 –24 years of age.
In the Poisson model evaluating incidence rates by age at diagnosis, race, stage, registry, and calendar time, all of these factors were found to be significant main effects (all P ⬍ .001). However, there were no statistically significant interaction effects identified between calendar time and age at diagnosis, race, stage, or registry (all P ⬎ .15). The effect of calendar time was found to be linear, and in the Poisson model adjusting for age at diagnosis and calendar time, the relative risk (95% CI) for choriocarcinoma in blacks and others compared with whites was 2.14 (1.60, 2.86) and 2.30 (1.67, 3.18), respectively (Table 1). Although there was a significant difference in trends in incidence rates over calendar time between registries (P ⫽ .04), no significant interaction effects by age at diagnosis (P ⫽ .76) or by race (P ⫽ .91) were identified. With the Poisson model, the estimated annual decrease in incidence rates (percent per year plus or minus standard error) was 3.3% (⫾ 0.9%) among whites (P ⫽ .004), 4.6% (⫾1.7%) among blacks (P ⫽ .006), and 3.4% (⫾ 1.9%) among other nonwhites (P ⫽ .08), respectively. The proportion of cases identified by positive laboratory or marker (nine cases, P ⫽ .004) but not those identified by radiographic studies (27 cases, P ⫽ .71), relative to the 399 histologically confirmed cases, significantly increased over the 27-year period. However, the increase in cases without (49 cases) compared with those with (401 cases) pathologic confirmation (histologic, cytologic, or microscopic, method not specified) was not significant (P ⫽ .46). Analyzed by stage, 139 (30.9%) of the pathologically confirmed cases had localized, 207 (46.0%) had regional or distant, and 55 (12.2%) had unstaged disease. Six (12.2%) cases without pathologic
982
Smith et al
Gestational Choriocarcinoma Trends
confirmation had localized disease, 38 (77.6%) had regional or distant metastases, and five (10.2%) were unstaged. The proportion of cases with distant (P ⫽ .49) or localized (P ⫽ .82) disease relative to the total cases did not change over calendar time. Gestational choriocarcinoma must be preceded by a gestational event, and ratios have almost always been reported with live birth and pregnancy denominators. With the use of the best published estimates for live birth and pregnancy denominators, the ratio of gestational choriocarcinoma cases per live births and clinically recognized pregnancies among all races was 1:26,000 live births and 1:41,000 pregnancies; 1:29,000 and 1:45,000 for whites, and 1:27,000 and 1:50,000 for blacks plus others (Table 3). The relative risk with live birth denominators was significantly higher for blacks and others than for whites (relative risk 1.10, 95% CI 1.02, 1.22, P ⫽ .01). However, with pregnancy denominators, the relative risks were not significantly different (relative risk for blacks and others, 1.06, 95% CI 0.98, 1.17, P ⫽ .18). We used the Surveillance, Epidemiology and End Results program to calculate 5-year observed and relative survivals by selecting for actively followed, malignant behavior and excluding second and later primaries, death certificate and autopsy cases, and adjusting for cumulative rates over one and increases in cumulative rates from a prior interval. The 27-year annualized 5-year relative survival (plus or minus two standard errors) was 89.5% (⫾ 3.15%), including those without microscopic confirmation (60 cases), and excluding those identified by second or later primary codes (25 cases), or based on death certificate or autopsy data (five cases). Stratified by race, the 5-year average relative
OBSTETRICS & GYNECOLOGY
Table 3. Estimated Choriocarcinoma Incidence Rates per Live Births and Pregnancy Within SEER (1981–1991)*
Race Whites Blacks and others Totals
Estimated Estimated mean mean Case to Case to live births in pregnancies in live birth pregnancy Count Woman-years SEER (SE) SEER (SE) ratio (SE) ratio (SE) 128
103,465,000
49
24,107,000
177
127,572,000
294,000 (3,700) 74,000 (2,000) 368,000 (5,600)
451,000 (5,400) 138,000 (3,800) 589,000 (9,000)
29,000 (3,400) 27,000 (6,800) 26,000 (3,000)
45,000 (5,300) 50,000 (12,600) 41,000 (5,000)
Relative Relative risk (95% risk (95% CI) per CI) per live births pregnancies 1.00
1.00
1.10 1.06 (1.02, 1.22) (0.98, 1.17)
SE ⫽ standard error; other abbreviations as in Tables 1 and 2. * Estimates based on annual number of live births and pregnancies in the US (1981–1991) (14). These estimates assume that 9.6% of total US live births and registries occurred within regions covered by the nine SEER registries.
survival was 92.1% (⫾ 3.36%) for whites, 81.0% (⫾ 9.23%) for blacks, and 87.6% (⫾ 9.28%) for other nonwhites (P ⫽ .02). The differences in odds ratios for survival among blacks versus whites was significant (2.69; 95% CI 1.23, 5.74, P ⫽ .006), but not for blacks versus others (1.73; 95% CI 0.61, 5.42, P ⫽ .37) or for others versus whites (1.54; 95% CI 0.53, 3.98, P ⫽ .46). According to case-specific data abstracted by stage, race, registry, age, survival in months, and cause of death based on death certificate information, 378 patients (84.0%) were listed as alive, and there were 72 deaths— 44 (9.8%) from choriocarcinoma, 20 (4.4%) from known causes other than choriocarcinoma, and eight (1.8%) for which the cause was unknown. Twenty (45.5%) deaths from choriocarcinoma occurred before 12 months of follow-up, and in five cases the survival time was less than 1 month. Survival by race (Figure 2), censored for deaths from intercurrent illnesses other than choriocarcinoma (unknown, and causes of death other than choriocarcinoma), showed significant differences (whites 92.4%, blacks 84.9%, others 87.1%, P ⫽ .045). Censored survival decreased with advancing stage of disease (localized 94.5%, regional 92.9%, distant 87.1%, P ⫽ .02) and with increasing age at diagnosis
(10 –20 years 97.0%; 20 –29 years, 87.2%; 30 –39 years, 93.3%; 40 – 49 years, 87.2%; 50 –59 years, 87.5%, and 60 or more years, 60.0%; P ⫽ .017). Censored survival was highest in Iowa (96.7%) and lowest in Hawaii (84.6%), but differences in survival between registries were not significant (P ⫽ .61). The median censored survival was 115 months (⫾ 96 months [standard error], range 0 –205 months). We analyzed the effect of calendar time on survival, subdividing the cases into three 9-year periods and censoring the analysis with the median follow-up for the last time period, 42 months, so that there were comparable follow-up intervals for each group. The difference (1973–1981, 174 cases, 23 deaths, survival rate 86.8%; 1982–1991, 145 cases, 12 deaths, survival rate 91.7%; 1992–1999, 130 cases, four deaths, survival rate 96.9%, log-rank P ⫽ .01) was statistically significant. In the Cox’s proportional hazards model (hazards ratio, 95% CI), calendar time, age at diagnosis, and race were significant main effects on survival. There was no interaction effect with calendar time and stage (P ⫽ .38). The effect of calendar time by age of diagnosis was significant (P ⫽ .01), but only for the 1973–1981 period (hazards ratio 0.51, 95% CI 0.33, 0.80, P ⫽ .004), and was not
Figure 2. Censored survival by race according to Surveillance, Epidemiology and End Results (1973–1999) program data. Smith. Gestational Choriocarcinoma Trends. Obstet Gynecol 2003.
VOL. 102, NO. 5, PART 1, NOVEMBER 2003
Smith et al
Gestational Choriocarcinoma Trends
983
significant by race (P ⫽ .90) or stage (P ⫽ .38). The effect of age at diagnosis was significant among whites (hazard ratio 1.06, 95% CI 1.03, 1.10), which indicates that mortality increased by 6.2% for each additional 10 years of age. However, age at diagnosis did not impact survival for blacks (hazard ratio 1.0, 95% CI 0.94, 1.06) or others (hazard ratio 0.97, 95% CI 0.90, 1.05). The effect of age at diagnosis did not change when we adjusted for either registry or stage of disease. DISCUSSION According to population-based data recorded over 27 years (Surveillance, Epidemiology and End Results, 1973–1999) and the best available published estimates of live births and pregnancies in the United States, the age-adjusted incidence rate for gestational choriocarcinoma per 100,000 woman-years for all races was 0.133, one in 25,674 live births, and one in 41,094 pregnancies. Over all, there was a 49.7% decrease, with an annual percentage decrease of 2.8% per year. In the Poisson model, calendar time, age at diagnosis, stage, registry, and race were all significant main effects, but no significant interaction effects between these variables and calendar time were identified. Although trends by registry were significant, because these data could not be controlled for live birth or pregnancy denominators, no conclusions can be drawn regarding the effects of fertility trends by race, age, or registry. Age-adjusted incidence rates were higher among blacks and other nonwhites compared with whites, but only blacks were found to have significantly poorer survival rates. Brinton et al,5 using Surveillance, Epidemiology and End Results data, have previously reported that the rates of choriocarcinoma, after adjusting for race and estimated birth rates, were 2.1- and 1.8-fold higher among blacks and other races compared with whites. Among women of Asian descent, these findings were as expected, because these women are known to have a higher incidence of hydatidiform mole,4 and most choriocarcinoma cases are preceded by a molar pregnancy.6 However, although the risk of hydatidiform mole in blacks is less well characterized, most studies conducted within the United States have not found significantly higher rates in blacks.16 –18 In Rhode Island, molar pregnancy rates from 1956 through 1965 were higher among blacks than for whites, but the difference was not statistically significant, possibly because of the small number of cases in blacks.17 Similarly, a case– control study that compared molar pregnancy rates among black, white, and Latin American women in Los Angeles also found no racial differences, but again, the sample size was small.18 Of interest, Hayashi et al16 estimated that the ratio of
984
Smith et al
Gestational Choriocarcinoma Trends
gestational trophoblastic disease for US blacks (58.2 per 100,000 pregnancies), was approximately half that of whites and others (118 and 117.9 per 100,000 pregnancies, respectively). This is especially intriguing, because it implies that the higher incidence of choriocarcinoma among blacks might be a reflection of greater risk for choriocarcinoma after a nonmolar pregnancy. This could potentially account for lower survival among blacks, because the prognosis for choriocarcinoma after a normal gestation is poorer.1 Reduced access to surveillance might also account for higher incidence and mortality rates among blacks.19 Although persistent gestational trophoblastic disease has an excellent prognosis, noncompliance with follow-up can lead to a delay in diagnosis, and inconsistent compliance with treatment can lead to the development of chemotherapy-resistant disease.1,19 Among blacks, survival rates are poorer for cervical, breast, endometrial, and colorectal carcinoma, and in some cases these disparities persist after correcting for stage of disease.20 –25 It is now recognized that differences in survival by race are a reflection of socioeconomic status and class, because blacks are disproportionately represented among the poor and disadvantaged.25,26 Indigent women, especially those belonging to ethnic and racial minority groups, are less likely to fully participate in breast and cervical carcinoma screening programs,27 and postmolar surveillance, including routine human chorionic gonadotropin determinations and timely delivery of chemotherapy are also poorer.19 With respect to other racial groups, data is also limited regarding trends in gestational trophoblastic disease incidence rates. In the United Kingdom, where centralized registration of all gestational trophoblastic disease has been standard practice since 1973, the molar pregnancy ratio is 1.54 per 1000 live births.28 In Sweden, which also maintains a population-based registry, there were no significant trends in incidence rates between 1975 and 1988.29 Population-based data indicate that there are higher gestational trophoblastic disease rates among nonwhite Hispanic,30,31 American Indian or Eskimo,32,33 and Asian34 –37 women. Alaskan natives32 and Southwestern American Indians33 have rates three to four times greater than United States whites. Japanese and Philippino women in Hawaii also have higher rates than do Hawaiian whites.38,39 The significance of Hispanic descent is uncertain. In New Mexico and Los Angeles,18,33 Hispanic women were not found to be at increased risk, whereas hospital-based studies from Mexico and Central America have consistently reported higher rates.30,31 In Asian populations, both populationbased and hospital-based studies indicate that rates are declining.4,34 –37 In Korea, with live births as the denom-
OBSTETRICS & GYNECOLOGY
inator, a 17.5-fold decline in all gestational trophoblastic disease incidence rates was recently reported (from 40.2 [1971–1975] to 2.3 [1991–1995]).34 The decline in incidence rates found in the Surveillance, Epidemiology and End Results data might be a reflection of more wide-spread use of standardized protocols for surveillance, clinical evaluation, and therapy for low-risk gestational trophoblastic disease, or might be a reflection of lower conception rates, especially in older women. Differences in rates by registry could not be accounted for by the effects of age at diagnosis, race, or stage. These unexpected findings might indicate that other factors not evaluated in this study, including registry-specific conception rates, environmental influences (climate, diet, other exposures), and possibly differences in access to quality medical care, account for differences by registry. Ascertainment bias might also contribute to the perceived decline in incidence rates. The proportion of cases without microscopic confirmation (49 of 450 cases, 10.9%), was the same as that of the previous Surveillance, Epidemiology and End Results study (20 of 203 cases, 9.7%).5 Although there was an increase in the number of cases diagnosed based on laboratory markers, the proportion of histologically confirmed cases did not change over time. Analysis of these data within the context of other studies is limited, because most population-based studies include all gestational trophoblastic disease variants, and choriocarcinoma cases account for only a small percentage. The vast majority of gestational trophoblastic disease cases are managed based on the number and location of metastases and the human chorionic gonadotropin titer instead of a tissue diagnosis. Therefore, an analysis limited primarily to histologically confirmed and metastatic cases potentially underestimates the total number of cases. Because the disease is rare, any variation in criteria for case selection could influence rate determinations. However, because the proportion of cases identified with regional/metastatic disease and with microscopic confirmation did not change over calendar time, it is probable that this analysis accurately reflects Surveillance, Epidemiology and End Results registry trends. Other limitations to this study deserve acknowledgment. A comparison of trends for all gestational trophoblastic disease cases is not possible with Surveillance, Epidemiology and End Results data because partial or invasive mole variants are not included in the public-use database. Most registries do not maintain data for molar pregnancies, and these cases are categorically excluded from the Surveillance, Epidemiology and End Results public-use database by coding definitions. Nevertheless, we have attempted to estimate the ratio of choriocarci-
VOL. 102, NO. 5, PART 1, NOVEMBER 2003
noma cases relative to gestational trophoblastic disease within the Surveillance, Epidemiology and End Results registry, extrapolating from an analysis of all gestational trophoblastic disease cases in one state registry that were systematically collected over 25 years. Given 939 incident gestational trophoblastic disease cases including 22 choriocarcinomas, the ratio of choriocarcinoma to gestational trophoblastic disease (excluding placental site trophoblastic tumors) is 1:42.7.33 In the same period, there were 422 choriocarcinoma cases in the Surveillance, Epidemiology and End Results study. If this ratio is reasonably representative of national registry incidence rates, hypothetically there were 18,019 cases of gestational trophoblastic disease. This estimation is imprecise, and overlooks the influence of geographic region, race, and number of cases arising from a normal antecedent pregnancy. However, this is the best estimate available, which emphasizes the potential for ascertainment bias. Of note, the three unpublished choriocarcinoma cases comprising the New Mexico numerator were found with the SEERStat program when topography codes for uterus (two case) and fallopian tube (one case) were used. The choice of cases selected for inclusion could influence incidence and survival rates. Unlike the previous Surveillance, Epidemiology and End Results study, all ovarian cases were excluded, and it is possible that some of these were gestational choriocarcinomas. The two cases identified as cervical in origin might have been primary cervical choriocarcinomas.40 However, because this condition is exceedingly rare, it is more likely that they were gestational choriocarcinomas that were coded as cervical in origin because the cervix was the site of biopsy. Similarly, the one case identified as arising in the adnexa may have been a germ cell neoplasm. All three of these cases were histologically confirmed and were unstaged (two cases) or had distant disease (one case). There were five, eight, and 15 cases, respectively, in women over 60, 50 –59, and 45– 49 years of age—women unlikely to have recently been pregnant. It is feasible that cases found in some or all postmenopausal women might be a reflection of delayed-onset choriocarcinoma, misinterpretation of pathologic specimens, or misclassification of other neoplasms as choriocarcinoma. The staging system available in the Surveillance, Epidemiology and End Results database does not take into account the number and location of metastases, which limits its usefulness, because women with pulmonary metastases alone have an excellent prognosis, but survival is poorer in those with liver, brain, and other metastatic sites. The conception-based denominators can also influence rate determinations. In this study, the best available published estimates of live birth and pregnancy for the
Smith et al
Gestational Choriocarcinoma Trends
985
entire US population were used because there is no centralized data for registry-specific fertility denominators.15 Therefore, no statement can be made regarding the effects of reproductive trends over time and between registries. The alternative approach, to include live birth data from the five state registries with published live birth rates, ignores the fact that fertility rates are likely to vary across registries. In a recent analysis of all gestational trophoblastic disease in which one stage registry was used, age-adjusted incidence rate determinations were found to be a more reliable estimate than rates based on live births or pregnancies.33 Nonetheless, when the best published estimates for live births and pregnancies in the United States are used, the ratios per live births and pregnancy reported in this and the previous Surveillance, Epidemiology and End Results study are similar.5 The absolute number of conceptions is the ideal denominator to use when comparing rates, but the number of conceptions is rarely known, and the number of clinically recognized pregnancies greatly underestimates the total number of conceptions. In this study, we calculated incidence ratios by using women-years, live births, and clinically recognized pregnancies according to published denominator data for the middle 11-year period (1981–1991) and corresponding cases from the Surveillance, Epidemiology and End Results study. Our results are in agreement with those of Brinton et al,5 who reported that live birth denominators overestimated the ratio by 40%. In this study, significant differences in ratios by race per live births but not by pregnancies were found. This suggests that higher conception rates alone do not account for the observed differences. In conclusion, data derived from the Surveillance, Epidemiology and End Results study indicate that gestational choriocarcinoma incidence rates in the United States seem to be declining and that survival has improved. There are significant differences in risk by race, stage, age at diagnosis, and geographic region. Although both blacks and others have higher incidence rates compared with whites, blacks alone are at greater risk for death. Factors such as access to contraception or timely health care, as well as genetic, dietary and other environmental influences, warrant further investigation.
REFERENCES 1. Berkowitz RS, Goldstein DP. Chorionic tumors [review]. N Engl J Med 1996;335:1740–8. 2. Lage JM, Wolf NG. Gestational trophoblastic disease. New approaches to diagnosis. Clin Lab Med 1995;15: 631–64.
986
Smith et al
Gestational Choriocarcinoma Trends
3. Aziz MF, Kampono N, Moegni EM, Sjamsuddin S, Barnas B, Samil RS. Epidemiology of gestational trophoblastic neoplasm at the Dr. Cipto Mangunkusumo Hospital, Jakarta, Indonesia. Adv Exp Med Biol 1984;176:165–75. 4. Palmer JR. Advances in the epidemiology of gestational trophoblastic disease. J Reprod Med 1994;39:155–62. 5. Brinton LA, Bracken MB, Connelly RR. Choriocarcinoma incidence in the United States. Am J Epidemiol 1986;123:1094–100. 6. Hertig AT, Mansell H. Tumors of the female sex organs, part I: Hydatidiform mole and choriocarcinoma. In: Atlas of tumor pathology, section IX, fascicle 33. Washington: Armed Forces Institute of Pathology, 1956. 7. Shanmugaratnam K, Muir CS, Tow SH, Cheng WC, Christine B, Pedersen E. Rates per 100,000 births and incidence of choriocarcinoma and malignant mole in Singapore Chinese and Malays: Comparison with Connecticut, Norway, and Sweden. Int J Cancer 1971;8:165–75. 8. Fritz A, Percy C, Jack A, Shanmugaratnam K, Sobin L, Porkin DM, Whelan S, Van Holten V, Muir C, eds. International classification of diseases for oncology (ICD0), 3rd ed. Geneva: World Health Organization, 2000. 9. Cunningham J, Ries L, Hankey B, Seiffert J, Lyles B, Shambaugh E, et al, eds. The SEER program code manual, revised edition. National Institutes of Health (PHS) 92–1999. Bethesda, Maryland: National Institutes of Health, 1999. 10. Smith PG. Comparison between registries: Age-standardized rates. In: Parkin DM, Muir CS, Whelan SL, Gao Y-T, Ferlay J, Powel J, eds. Cancer incidence in five continents, vol. VI (IARC scientific publication no. 120). Lyon, France: IARC, 1992. 11. Kleinbaum DG, Kupper LL, Muller KE. Applied regression analysis and other multivariate methods. North Sciutate, Massachusetts: Duxbury Press, 1988. 12. McCullagh P, Nelder JA. Generalized linear models, 2nd ed. London: Chapman & Hall, 1989. 13. Collett D. Modelling survival data in medical research. London: Chapman & Hall, 1994. 14. Kalbfleisch JD, Prentice RL. The statistical analysis of failure time data. New York: John Wiley & Sons, 1980. 15. Saraiya M, Berg CJ, Shulman H, Green CA, Atrash HK. Estimates of the annual number of clinically recognized pregnancies in the United States, 1981-1991. Am J Epidemiol 1999;149:1025–9. 16. Hayashi K, Bracken MB, Freeman DH Jr, Hellenbrand K. Hydatidiform mole in the United States (1970-1977): A statistical and theoretical analysis. Am J Epidemiol 1982; 15:67–77. 17. Yen S, MacMahon B. Epidemiologic features of trophoblastic disease. Am J Obstet Gynecol 1968;101:126–32. 18. Westerhout FC Jr, Morel ES, Slate WG. Observations on 138 molar pregnancies. Am J Obstet Gynecol 1969;103: 56–9.
OBSTETRICS & GYNECOLOGY
19. Massad LS, Abu-Rustum NR, Less SS, Renta V. Poor compliance with postmolar surveillance and treatment protocols by indigent women. Obstet Gynecol 2000;96: 940–4. 20. Dignam JJ. Differences in breast cancer prognosis among African-American and Caucasian women. CA Cancer J Clin 2000;50:50–64. 21. Hicks ML, Phillips JL, Parham G, Andrews N, Jones WB, Shingleton HM, et al. The National Cancer Data Base report on endometrial carcinoma in African-American women. Cancer 1998;83:2629–37. 22. Digman JJ, Ye Y, Colangelo L, Smith R, Manounas EP, Wieand HS, et al. Prognosis after rectal cancer in blacks and whites participating in adjuvant therapy randomized trials. J Clin Oncol 2003;21:413–20. 23. Howell EA, Chen YT, Concato J. Differences in cervical cancer mortality among black and white women. Obstet Gynecol 1999;94:509–15. 24. Grisby PW, Hall-Daniels L, Baker S, Perez CA. Comparison of clinical outcome in black and white women treated with radiotherapy for cervical carcinoma. Gynecol Oncol 2000;79:357–61. 25. Cross CK, Harris J, Recht A. Race, socioeconomic status, and breast carcinoma in the U.S.: What have we learned from clinical studies. Cancer 2002;95:1988–99. 26. O’Malley CD, Le GM, Laser SL, Shema SJ, West DW. Socioeconomic status and breast carcinoma survival in four racial/ethnic groups. Cancer 2003;97:1303–11. 27. Katz SJ, Hofer TP. Socioeconomic disparities in preventive care persist despite universal coverage: Breast and cervical cancer screening in Ontario and the United States. JAMA 1994;272:530–4. 28. Bagshawe KD, Dent J, Webb J. Hydatidiform mole in England and Wales 1973-1983. Lancet 1986;11:673–7. 29. Flam F, Lindsted-Lundstro¨m V, Rutqvist LE. Incidence of gestational trophoblastic disease in Stockholm county, 1975-1988. Eur J Epidemiol 1992;8:173–7. 30. Marquez-Monter H, de la Vega GA, Ridaura C, Robles M. Gestational choriocarcinoma in the general hospital of Mexico: Analysis of forty cases. Cancer 1968;22:91–8.
VOL. 102, NO. 5, PART 1, NOVEMBER 2003
31. Aguero O, Kizer S, Pinedo G. Hydatidiform mole in Concepcion Palacios Maternity Hospital. Am J Obstet Gynecol 1973;116:1117–20. 32. Martin PM. Hydatidiform mole in Nigeria, Iran, and Alaska. III. High frequency of hydatidiform mole in native Alaskans. Int J Gynaecol Obstet 1978;15:395–6. 33. Smith HO, Hilgers RD, Bedrick EJ, Qualls CR, Wiggins CL, Rayburn RF, et al. Ethnic differences in risk for gestational trophoblastic disease in New Mexico: A 25-year population-based study. Am J Obstet Gynecol 2003;188:357–66. 34. Kim SJ, Bae SN, Kim JH, Kim CJ, Han KT, Chung JK, et al. Epidemiology and time trends of gestational trophoblastic disease in Korea. Int J Gynecol Obstet 1998;60: S33–8. 35. Hando T, Ohno M, Kurose T. Recent aspects of gestational trophoblastic disease in Japan. Int J Gynecol Obstet 1998;60:S71–6. 36. Sakumoto K, Kigashi M, Kanazawa K. Hydatidiform mole in Okinawa Islands and mainland Japan. Int J Gynaecol Obstet 1999;64:309–10. 37. Sivanesartnam V. The management of gestational trophoblastic disease in developing countries such as Malaysia. Int J Gynaecol Obstet 1998;60:S105–9. 38. McCorriston CC. Racial incidence of hydatidiform mole. Am J Obstet Gynecol 1968;101:377–82. 39. Matsuura J, Chiu D, Jacobs PA, Szulman AE. Complete hydatidiform mole in Hawaii: An epidemiological study. Genetic Epidemiol 1984;1:271–84. 40. Herts BR, Yee JM, Porges RF. Primary cervical choriocarcinoma: Case report and review of the literature. J Ultrasound Med 1993;12:59–62. Address reprint requests to: Harriet O. Smith, MD, University of New Mexico Health Sciences Center, Department of Obstetrics and Gynecology, Division of Gynecologic Oncology, 2211 Lomas Boulevard, NE, Albuquerque, NM 87131-5286; Email: [email protected]. Received December 26, 2002. Received in revised form June 4, 2002. Accepted June 9, 2002.
Smith et al
Gestational Choriocarcinoma Trends
987