Correlates of Hysterectomy among African-American Women

American Journal of Epidemiology Copyright O 99 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol. 150, Printed In USA. Correlates of Hysterectomy among African-American Women Julie R. Palmer, 1 R. Sowmya Rao, 1 Lucile L. Adams-Campbell, 2 and Lynn Rosenberg 1 Hysterectomy is the second most common surgery performed on US women. Baseline data from a large study of African-American women were used to examine correlates of premenopausal hysterectomy. Analyses were conducted on participants aged 30-49 years; 5,3 had had a hysterectomy and 29,77 were still menstruating. Multiple logistic regression was used to compute prevalence odds ratios for the association of hysterectomy with various factors. Hysterectomy was associated with region of residence: Odds ratios for living in the,, and relative to the rtheast were 2.63 (95% confidence interval (Cl): 2.3, 2.91), 2.02 (95% Cl: 1.1, 2.25), and 1.9 (95% Cl: 1.6, 2.), respectively. Hysterectomy was inversely associated with years of education and age at first birth: Odds ratios were 1.96 (95% Cl: 1.74, 2.21) for ^ years of education relative to > years and 4.33 (95% Cl: 3.60, 5.22) for first birth before age relative to age 30 or older. Differences in the prevalence of major indications for hysterectomy did not explain the associations. This study indicates that the correlates of hysterectomy among African-American women are similar to those for White US women. The associations with geographic region and educational attainment suggest that there may be modifiable factors which could lead to reduced hysterectomy rates. Am J Epidemiol 99; 150:09-15. blacks; hysterectomy; women Hysterectomy is one of the most frequently performed surgeries among women in the United States, second only to cesarean section (1). Hysterectomy rates are considerably higher in the United States than in most other ern countries (1-3), and rates also appear to vary by geographic region within the United States, being highest in the (4, 5). Several studies have indicated that women with low educational attainment or low income are more likely to have a hysterectomy (6, 7). Data on variation by race are less consistent (4-10), but most sources indicate a higher rate in African-American women than in White women, especially before age 45 (4, 7, 9, 10). Little information is available on the correlates of hysterectomy among African-American women, since most data sources have comprised primarily White samples. We assessed correlates of hysterectomy in data from 34,950 participants in the Black Women's Health Study. Received for publication January 11,99, and accepted for publication April, 99. Abbreviations: Cl, confidence interval; NHANES I, First National Hearth and Nutrition Examination Survey. 1 Slone Epidemiology Unit, Boston University School of Medicine, Brookline, MA. 2 Division of Epidemiology and Biostatistics, Howard University Cancer Center, Washington, DC. Reprint requests to Dr. Julie R. Palmer, Slone Epidemiology Unit, Boston University School of Medicine, 71 Beacon Street, Brookline, MA 026. MATERIALS AND METHODS In 95, 64,554 African-American women aged 21-69 years were enrolled in the Black Women's Health Study through mailed questionnaires sent to subscribers of Essence magazine (a women's magazine marketed to Black women), members of several professional organizations, and friends and relatives of respondents (11). The baseline questionnaire obtained information on many factors, including demographic characteristics, medical history, medication use, height and weight, and smoking and alcohol consumption. Among the conditions included under medical history were uterine fibroids and endometriosis. Participants were asked whether their menstrual periods had stopped permanently and, if yes, the age at which their periods had stopped and the reason for stopping (surgery, natural menopause, medication, or radiation). If their periods had stopped due to surgery, they were asked whether the ovaries were removed as well as the uterus and the reason for the surgery. In addition, there was a question on current use of contraception, with hysterectomy listed as one of the choices. The study protocol was approved by the Institutional Review Boards of Boston University. This analysis was restricted to study participants who were aged 30-49 years in 39,111); almost no participants under age 30 reported undergoing hysterectomy. Excluded from the analysis were 1,561 par- 09

10 Palmer et al. ticipants who had missing data for any of the menopause questions, 1,24 who reported a natural menopause or menopause due to medication or chemotherapy, 293 who had never menstruated, 535 who reported cancer of the cervix or uterus, and 4 who had conflicting data on the menopause questions. This left 34,950 women: 5,3 who reported permanent cessation of menses due to removal of the uterus (hysterectomized women) and 29,77 who were premenopausal and had never had a hysterectomy (nonhysterectomized women). The median age was 44 for the hysterectomized women and 37 for the nonhysterectomized women. Multiple logistic regression analysis was used to estimate prevalence odds ratios for the association of hysterectomy with particular factors (). Age (as a continuous variable), geographic region (rtheast,,, and ), years of education (<,,, and >17), history of uterine leiomyoma, history of endometriosis, age at menarche (<11, 11,, and years), age at first birth (<, -24, 25-29, and >30 years), parity (0, 1, 2, 3, and >4), and history of tubal ligation were controlled simultaneously. Ninety-five percent confidence intervals were calculated by the test-based method (). RESULTS The prevalence of hysterectomy increased from 1.9 percent among women aged 30-34 years to 3.9 percent among women aged 45^49 years (table 1). Of those who had had hysterectomies, the proportions who reported having both ovaries removed, one ovary removed, and no ovaries removed were similar across age strata; approximately 30 percent had had both ovaries removed and 15 percent had had only one ovary removed. As table 2 shows, women who lived in the were about 2.6 times as Likely to have a hysterectomy as women in the rtheast, and women in the and were about two times as Likely. Women with the least education were the most likely to have TABLE 1. Prevalence of hysterectomy among 34,950 women aged 30-49 years, Black Women's Health Study, 95* Age (years) 30-34 35-39 40M4 45-^*9 Total no. 10,00 9.6.906 6,150 %wfth hysterectomy 1.9 7.4.7 3.9 Women with hysterectomy %wtth bilateral oophorectomy 29 2 27 31 * Women who had a natural menopause were excluded. %wfth one ovary removed 15 1 had a hysterectomy: The prevalence odds ratios for <,, and years of education relative to >17 years were 1.96, 1.63, and 1.1, respectively, and all were statistically significant. As expected, the occurrence of hysterectomy was strongly associated both with history of uterine leiomyomas (prevalence odds ratio = 5.34; 95 percent confidence interval (CI): 4.96, 5.75) and history of endometriosis (prevalence odds ratio = 4.32; 95 percent CI: 3.90, 4.79). Early age at first birth was also strongly associated with risk of hysterectomy: Compared with women whose first birth took place at age 30 or later, the prevalence odds ratios were 4.33 (95 percent CI: 3.60, 5.22), 3.56 (95 percent CI: 2.95,4.2), and 2.30 (95 percent CI: 1.9, 2.1) for a first birth occurring at ages <, -24, and 25-29 years, respectively. Other reproductive variables were only weakly associated with hysterectomy: For menarche after age years relative to menarche before age 11, the prevalence odds ratio was (95 percent CI: 0.70,0.6); for no births relative to four or more births, it was 1.30 (95 percent CI: 1.10, 1.54); and for a history of tubal ligation, it was 0.92 (95 percent CI: 0.5, ). As table 3 shows, the associations of hysterectomy with geographic region, years of education, history of endometriosis, and age at first birth were strongest for hysterectomy before age 40, whereas the association with history of uterine leiomyoma was stronger at age 40 years and older. Differences in the prevalence of uterine leiomyoma and endometriosis (the two major indications for hysterectomy in premenopausal women) did not explain the strong associations with geographic region and educational attainment observed for hysterectomy (table 4). The prevalence of uterine leiomyoma was uniform across geographic region. Leiomyoma was somewhat less prevalent among the least educated women (2 percent as opposed to 33 percent, 35 percent, and 36 percent for the other categories), rather than more prevalent as would be expected if this explained the greater prevalence of hysterectomy. The prevalence of endometriosis was similar across both geographic regions and levels of education. It is possible that the stronger associations observed for hysterectomy before age 40 with geography and education reflect a cohort effect due to changes in medical practice over time. Overall, hysterectomies that were done before age 40 would have occurred further in the distant past than hysterectomies performed at ages 40-49. To examine whether this was in fact the explanation, we assessed the relation of geographic region and years of education to prevalence of hysterectomy according to both current age and age at Am J Epidemiol Vol. 150,., 99

Correlates of Hysterectomy in African-American Women 11 TABLE 2. Factors related to the prevalence of hysterectomy, Black Women's Health Study, 95. of women wtth hysterectomy. of women without hysterectomy Prevalence odds ratio' 95% confidence interval* rtheast 61 9 1,360 1,953,336 5,553 6,945,950 i.oot 1.9 2.02 2.63 1.6, 2. 1.1, 2.25 2.3, 2.91 217 1,06 2,056 4 1,044 4,47 10,29 7,177 6,79 1.96 1.63 1.1 1.74,2.21 1.4, 1.0 1.06, History of uterine leiomyoma 1,444 3,7 21,935 7,52 5.34 4.96, 5.75 History of endometriosis 4,06 1,077 2,14 1,603 4.32 3.90, 4.79 Age (years) at menarche <11 11 2 1,409 1,355 1,19 915 6,2,490 7,325 5,959 0.0 0.73, 0.9 0.70, 0.5 0.70, 0.6 Age (years) at first birth < -24 25-29 230 1,21 1,47 571 9 6,41 6,450 4,4 2,50 4.33 3.56 2.30 3.60, 5.22 2.95, 4.2 1.9,2.1 Parity 0 1 2 3 24 1,0 1,266 1,609 04 422 9,6 7, 7,577 3,639 1,71 1.30 1.34 1.21 1.10, 1.54 1., 1.55 1., 1.55 1.03, 1.41 Tubal ligation 2,535 2, 1,36 9,06 0.92 0.5, * Prevalence odds ratios and 95% confidence intervals were calculated in multiple logistic regression analyses that included terms for the following: current age (continuous variable), geographic region, years of education, history of uterine leiomyoma, history of endometriosis, age at menarche, age at first birth, number of births, and history of tubal ligation. t Referent. hysterectomy. As table 5 shows, the differences according to age did not appear to be simply a result of how long ago the hysterectomy occurred. In an analysis of hysterectomy at age 30-34, it was found that among both women with a current age of 40-49 years (whose hysterectomies would have taken place 6 years previously) and women aged 30-39 years (whose hysterectomies would have occurred mostly in the past 5 years and no more than 9 years previously), the associations with geographic region and years of education were of the same magnitude. Similarly, the odds ratios for hysterectomy at age 35-39 were of the same magnitude regardless of whether the current age was 40-49 years or 30-39 years. We repeated the main analyses among the 1,067 hysterectomized women and 21,109 nonhysterectomized women who had reported neither a history of leiomyoma nor a history of endometriosis (data not shown). The associations with geographic region, educational attainment, and other factors observed in the overall data were present in this subgroup. For example, the prevalence odds ratio for living in the relative to the rtheast was 2.31 (95 percent CI: 1.92, 2.79), and the odds ratio for the lowest category of education relative to the highest was 2.99 (95 percent CI: 2.41, 3.70). Am J Epidemiol Vol. 150,., 99

Palmer et al. TABLE 3. Factors related to the prevalence of hysterectomy, according to age, Black Women's Health Study, 95 Aged <40 years t Aged 40-49 years. of. of women > vomen wlttii without hysterectomy hysterectomy Prevalence odds ratio* 95% confidence Interval*. of women with hysterectomy. of women without hysterectomy Prevalence odds ratio* 95% confidence Interval* rtheast 495 652 971 1,37,336 5,553 6,945,950 i.oot 2. 2.46 3.15 1., 2.47 2.17,2.79 2.79, 3.55 33 299 373 517 2,95 2,1 2,71 3,030 i.oot 1.47 1.42 1.4 1.22, 1.76 1.,1.69 1.56,2.17 S 17 770 1,45 61 605 4,47 10,29 7,177 6,79 2. 1.2 1.31 1.93,2.53 1.62,2.04 1.15, 1.49 26 556 247 417 1,776 3,634 2,17 3,031 1.40 0.99 1.15, 1.69 1., 1.57 0.2, 1. History of uterine leiomyoma 1,5 2,370 21,935 7,52 4.54 4.1, 4.94 25 1,270 7,0 3,11.31 7.,9.61 History of endometrtosis 2,709 796 2,14 1,603 4.63 4.15,5.17 1,267 261 10,211 6 2.9 2.49, 3.5 Age (years) at menarche <11 11 974 91 11 622 6,2,490 7,325 5,959 0.76 0.6, 0.5 0.69, 0.7 0.6, 0.7 405 431 352 26 2,343 3,002 2, 2,245 0.92 0.75 0.74 0.7, 1.09 0.63, 0.90 0.62, 0.90 Age (years) at first birth < -24 25-29 30 1,32 93 327 92 6,41 6,450 4,4 2,50 5.77 4.35 2.43 4.57, 7.29 3.45, 5.50 1.9,3. 447 460 227 73 2,97 2,769 1,706 1,05 2.15 2.22 2.00 1.61,2.7 1.6,2.95 1.4,2.69 Parity 0 1 2 3 4 7 42 1,102 553 270 9,6 7, 7,577 3,639 1,71 1.40 1.29 1.09, 1.61 1.11, 1.5 1.1, 1.65 1.0, 1.54 279 393 470 229 9 2,0 2,604 3,32 1,739 1,007 1.44 1.44 1.26 1.03 1.0,1.90 1., 1.5, 1.60 0.0, Tubal ligation 1,33 1,44 1,36 9,06 0.1 0.74, 0.9 639 71 5,153 4,9 1.1 1.03,1.36 * Prevalence odds ratios and 95% confidence intervals were calculated in multiple logistic regression analyses that included terms for the following: current age (continuous variable), geographic region, years of education, history of uterine leiomyoma, history of endometriosis, age at menarche, age at first birth, number of births, and history of tubal ligation. t Referent. The reasons for having had a hysterectomy did not differ appreciably according to age, geographic region, or education (table 6). Overall, 75 percent of participants reported uterine fibroids as the reason for the hysterectomy. DISCUSSION The high incidence of hysterectomy in the United States, coupled with the great differences in rates among countries and even among different regions of the United States, has prompted efforts to assess factors that may influence hysterectomy. Most of the information on hysterectomy in the United States comes from the National Hospital Discharge Survey (5,9), an annual probability sample of discharges from nonfederal, short-stay hospitals. Information on up to seven diagnoses and up to four procedures is collected for each hospital stay. These data allow estimation of rates of hysterectomy by age, race, and geographic area. They also provide information on the indications for hysterectomy. Data on the period from 70 to 93 have been reported (4, 5, 9): The highest rate was seen for the age group 35 44 years, and rates were Am J Epidemiol Vol. 150,., 99

Correlates of Hysterectomy in African-American Women TABLE 4. Prevalence* of fibroids and endometriosls, according to geographic region and years of education, Black Women's Hearth Study, 95 History of uterine fibroids History of endometrtosis TABLE 6. Reasons for hysterectomy (%) among 5,3 cases, Black Women's Health Study, 95 Fibroids Endometrtosls Bleeding Other Overall 75 21 rtheast 5S 217. 3,037 2,092 2,75 3,653 1,637 4,1 2,663 2,994 * Adjusted to overall age distribution (5-year categories). % 34 31 33 34 2 33 35 36. 557 42 61 959 379 999 59 661 % 6 7 9 7 Age (years) 30-39 40-49 rtheast 6 77 71 75 79 0 77 74 76 76 25 17 21 17 21 22 22 21 23 11 highest in the and lowest in the rtheast. The hysterectomy rate was higher in Black women than in White women, although the difference was not statistically significant in the most recent time period (-93); rates differed the most for hysterectomy at ages 30-44 years, being about 30 percent higher in Black women. In addition, the reasons for hysterectomy differed by race: Uterine leiomyoma was the most common indication in each group, but it accounted for 61 percent of hysterectomies in Black women as compared with 29 percent in White women (5). Other sources of data on hysterectomy, the Behavioral Risk Factor Surveillance Survey (6) and the First National Health and Nutrition Examination Survey (NHANES I) Epidemiologic Follow-up Study (), indicate th is most common among women with the least education. However, too few African-American women were included in these other surveys for a meaningful analysis to be carried out by race. Information is also available from a telephone survey of a random sample of 2,7 residents of Pittsburgh, Pennsylvania aged 40-52 years, 326 of whom were African-American (10). Overall, hysterec- TABLE 5. Relation of geographic region and years of education to the prevalence of hysterectomy, according to current age and age, Black Women's Health Study, 95 Prevalence odds ratio* Current age 30-39 years Current age 40-49 years Aged 30-34 years (n = 59) Aged 35-39 years (n=307) Aged 30-34 years (0=1,274) Aged 35-39 years (n= 1,326) Aged 240 years (n= 1,52) rtheast LOOT 2.94 2.99 3.50 1.3 2.33 2.75 2.45 2.79 3.41 1.63 2.07 2.6 1.46 1.42 1.3 17 3.34 2.46 1.41 T 3.33 1.4 1.10 3.3 2.64 1.52 2. 1.91 1.41 1.46 1.36 0.99 * Prevalence odds ratios were calculated in multiple logistic regression analyses that included terms for the following: current age (continuous variable), geographic region, years of education, history of uterine leiomyoma, history of endometriosis, history of tubal ligation, age at menarche, age at first birth, arid number of births. All prevalence odds ratios presented in this table except 1.10 and 0.99 were statistically significant (two-sided, p < 0.05). t Referent. Am J Epidemiol Vol. 150,., 99

Palmer et al. tomy was positively associated with Black race, high body mass index, early age at menarche, early age at first birth, and lower educational attainment. Among the Black participants, the only factor significantly associated with hysterectomy was nulliparity. Our study indicates that the most important correlates of hysterectomy among African-American women are the same as those for White US women and for overall US population samples. We found that premenopausal hysterectomy was strongly associated with geographic region of residence, with die highest rates being seen in the, and with level of education, with the highest rates being seen among the least educated women. It was also strongly associated with an early age at first birth, a history of uterine fibroids, and a history of endometriosis. Hysterectomy was only weakly associated with number of births and age at menarche, and it was not associated with history of tubal ligation. It was not possible to evaluate the relation of body mass index to hysterectomy, because the hysterectomy antedated reported weight. Because the study was restricted to African- American women, our data do not provide information on whether African-American women are more likely than White women to undergo hysterectomy. An interesting finding in our study is that the associations most likely to reflect the role of physician practice and uncertainty about the appropriate indications for hysterectomy the associations with region of residence and level of education were strongest for the earliest hysterectomies, those taking place before age 40. This finding persisted regardless of how recently the hysterectomies had occurred. The association with geographic region suggests regional differences in practice. The association with patient's level of education may have to do with how well alternatives to surgery are presented and understood, and with differences in access to medical care. The present finding of an increased prevalence odds ratio associated with early age at first birth is consistent with data from the mostly White participants in the NHANES I Epidemiologic Follow-up Study () and the random sample from Pittsburgh (10), although the associations were weaker in those studies. Why women who had their first birth before age should have more than a twofold risk of hysterectomy even in their forties, an age when childbearing has probably ceased is unclear. The association was particularly strong among women who had a hysterectomy before age 40. One possibility is that some of the early hysterectomies may have been performed as a means of sterilization. Previous studies of US women have suggested that women who have had a tubal ligation may be more likely to undergo hysterectomy, perhaps because of a greater acceptance of surgical procedures (, 15). We did not find an association between tubal ligation and premenopausal hysterectomy. In accordance with previous data on African- American women (9), uterine leiomyoma was by far the most common indication for hysterectomy, accounting for 75 percent of all hysterectomies. By contrast, in the National Hospital Discharge Survey data, uterine leiomyoma was the indication for only 29 percent of hysterectomies performed in White women (5). There is evidence from other studies that the incidence of leiomyoma is higher in African-American women and that the fibroids tend to be larger and more numerous at presentation (, 17). Certain limitations of the present study should be recognized. The Black Women's Health Study cohort is not a random sample of all African-American women. Thus, the overall prevalence of hysterectomy and agespecific prevalences may not accurately represent the true prevalence among all African-American women. However, comparisons within the cohort e.g., prevalence among women with < years of education compared with women with college degrees are likely to be valid. Such comparisons would be biased only if women chose to participate in the study on the basis of both characteristics i.e., if college-educated hysterectomized women enrolled in the study at rates quite different from those of hysterectomized women with < years of education. There is no reason to suspect that this was the case. In addition, the age-specific prevalences of hysterectomy in our data (table 1) were similar to those reported from representative samples (6). The analyses for which results are presented here were based on self-reports of hysterectomy, without documentation from medical records. Therefore, it is possible that some participants were misclassified with regard to hysterectomy status. We tried to minimize such misclassification by excluding women whose responses to the several questions about menopausal status and surgeries contained conflicting information or missing data. In addition, previous US studies have indicated a high degree of accuracy in the self-reporting of hysterectomy (1, ). Hysterectomy carries a great cost in terms of risk of complications, time needed for recovery, and health care costs to the individual and community (). The present findings suggest that there may be nonmedical factors that could be modified to reduce the rate of hysterectomy among African-American women. ACKNOWLEDGMENTS This work was supported by grant RO1 CA54 from the National Cancer Institute. Am J Epidemiol Vol. 150,., 99

Correlates of Hysterectomy in African-American Women 15 REFERENCES 1. National Center for Health Statistics. National Hospital Discharge Survey: annual summary, 90. (Vital and health statistics, series, no. 1). Washington, DC: US GPO, 92. (DHHS publication no. (PHS) 92-1773). 2. Bachmann GA. Hysterectomy: a critical review. J Reprod Med 90;35:39-62. 3. Coulter A, McPherson K, Vessey M. Do British women undergo too many or too few hysterectomies? Soc Sci Med ;27: 97-94. 4. Dicker RC, Scally MJ, Greenspan JR, et al. Hysterectomy among women of reproductive age: trends in the United States, 70-7. JAMA 2;24:323-7. 5. Lepine LA, Hillis SD, Marchbanks PA, et al. Hysterectomy surveillance United States, 0-93. MMWR Morb Mortal Wkly Rep CDC Surveill Summ 97;46:1-15. 6. Kjerulff K, Langenberg P, Guzinkis G. The socioeconomic correlates of hysterectomies in the United States. Am J Public Health 93;3:106-. 7. Kjerulff KH, Guzinski GM, Langenberg PW, et al. Hysterectomy and race. Obstet Gynecol 93;2:757-64.. Brett KM, Marsh JV, Madans JH. Epidemiology of hysterectomy in the United States: demographic and reproductive factors in a nationally representative sample. J Womens Health 97;6:309-. 9. Wilcox LS, Koonin LM, Pokras R, et al. Hysterectomy in the United States, -90. Obstet Gynecol 94;3:549-55. 10. Meilahn EN, Matthews KA, Egeland G, et al. Characteristics of women with hysterectomy. Maturitas 9; 11:3-29. 11. Rosenberg L, Adams-Campbell L, Palmer JR. The Black Women's Health Study: a follow-up study for causes and prevention of illness. J Am Med Womens Assoc 95;50:56-.. Schlesselman JJ. Case-control studies: design, conduct, analysis. New York, NY: Oxford University Press, 2:217-67.. Miettinen O. Estimability and estimation in case-referent studies. Am J Epidemiol 76;103:226-35.. Hillis SD, Marchbanks PA, Tylor LR, et al. Higher hysterectomy risk for sterilized than nonsterilized women: findings from the U.S. Collaborative Review of Sterilization. Obstet Gynecol 9;91:241-6. 15. Stergachis A, Shy KK, Grothaus C, et al. Tubal sterilization and the long-term risk of hysterectomy. JAMA 9O;264:293-.. Kjerulff KH, Langenberg P, Seidman JD, et al. Uterine leiomyomas: racial differences in severity, symptoms and age at diagnosis. J Reprod Med 96;41:43-90. 17. Marshall LM, Spiegelman D, Barbieri RL, et al. Variation in the incidence of uterine leiomyoma among premenopausal women by age and race. Obstet Gynecol 97;90:967-73. 1. Colditz GA, Stampfer MJ, Willett WC, et al. Reproducibility and validity of self-reported menopausal status in a prospective cohort study. Am J Epidemiol 7;6:3-25.. Paganini-Hill A, Ross RK. Reliability of recall of drug usage and other health-related information. Am J Epidemiol 2; 1:1-22.. Pinn VW. Additional material. Unnecessary hysterectomies, the second most common major surgery in the United States May 93. In: Subcommittee on Aging, US Senate Committee on Labor and Human Resources Senate hearing. Congressional Record 93(May 5): 105-. Am J Epidemiol Vol. 150,., 99