Kuppermann, Miriam PhD, MPH; Learman, Lee A. MD, PhD; Schembri, Michael MA; Gregorich, Steven E. PhD; Jackson, Rebecca MD; Jacoby, Alison MD; Lewis, James MD; Washington, A. Eugene MD, MSc
Although approximately 574,000 women in the United States undergo hysterectomy for noncancerous pelvic problems each year,1 questions remain regarding the factors that underlie their decisions to undergo this elective procedure. Similarly, there remains a dearth of data from prospective studies delineating which women are satisfied after hysterectomy.
The need to answer these critical questions, along with others, prompted the Agency for Health Care Policy and Research (now the Agency for Healthcare Research and Quality) in 1994 to convene a conference titled “Treatment Effectiveness of Hysterectomy and Other Therapies for Common Non-Cancerous Uterine Conditions,” during which clinical and methodologic experts reviewed the evidence on the use and effectiveness of hysterectomy and other treatments to make recommendations for research in this area.2 The conferees concluded that prospective studies were needed to investigate the natural history of these conditions and to compare the effects of hysterectomy with other treatment strategies, and that, in addition to “traditional endpoints,” these studies should use measures of psychological well-being, sexual function, health-related quality of life (HRQOL), and treatment satisfaction.2 They also noted the need for research into how patients' perceptions and expectations affect treatment choice and satisfaction.
Now, 15 years later, we have completed such an investigation: the Study of Pelvic Problems, Hysterectomy, and Intervention Alternatives (SOPHIA). This was an 8-year longitudinal study of English-, Spanish-, or Chinese-speaking premenopausal women with intact uteri who at the time of their enrollment were experiencing noncancerous pelvic conditions for which hysterectomy might be an option. The goals of SOPHIA were to 1) analyze the effect of abnormal uterine bleeding (with or without fibroids), chronic pelvic pain, and pressure due to fibroids on HRQOL and sexual function; 2) assess static and time-varying predictors of use of, and satisfaction with, hysterectomy and alternative management approaches to these conditions; and 3) describe the intermediate-term (4- to 8-year) clinical, HRQOL, and sexual function outcomes of these treatments.
Baseline characteristics of the SOPHIA cohort (none of whom had undergone hysterectomy at the time of enrollment) were reported in 2007.3 In this article, we present findings from our analysis of predictors of use of and satisfaction with hysterectomy. By focusing on a wide range of sociodemographic, HRQOL, and attitudinal characteristics among women who were followed for up to 8 years (and thus included several years of observation before hysterectomy for some of them), we aimed to identify new information that could be used to optimize clinical management decision making for noncancerous conditions often treated by hysterectomy.
MATERIALS AND METHODS
Patient recruitment methods, measures, and baseline characteristics of SOPHIA participants have been described.3 Briefly, English-, Spanish-, and Chinese (Cantonese or Mandarin)-speaking premenopausal women aged 31 to 54 years who had sought care in the previous year for noncancerous pelvic problems at clinics and practices affiliated with an academic medical center (the University of California, San Francisco), a county facility serving primarily indigent patients (San Francisco General Hospital), a closed-panel health maintenance organization (HMO) (Kaiser Permanente Northern California), or one of several community hospitals in San Francisco and had not undergone hysterectomy were recruited in two waves: 1998/1999 (cohort I) and 2003/2004 (cohort II), with the final interview conducted in January 2008. The study was approved by the University of California, San Francisco, Committee on Human Research and the Kaiser Permanente Northern California and San Francisco General Hospital Institutional Review Boards.
Participation in the study consisted of a baseline and up to eight annual face-to-face follow-up interviews (mean duration 30.3±13.3 minutes), and permission for investigators to review the participant's medical records. After signing informed consent, each participant completed an interviewer-administered questionnaire that included items related to 1) sociodemographic and clinical characteristics; 2) prior and current use of treatments for their pelvic problems (including Western medications, invasive and noninvasive surgeries, and complementary and alternative medicine); and 3) pelvic problem symptom resolution and impact, HRQOL, sexual function, and hysterectomy- and uterus-related attitudes. The questionnaire was readministered at each annual face-to-face interview. Use of hysterectomy was primarily ascertained during these interviews and supplemented by chart reviews for participants who could not be reached for their final interview. Women who completed the baseline interview and at least one follow-up interview, or for whom we were able to conduct chart review, were included in this analysis.
The SOPHIA measures have been described.3 They included a 12-item Pelvic Problem Impact Questionnaire adapted from the Urinary Incontinence Impact Questionnaire,4 which asked the extent to which pelvic problems had interfered with the respondent's mood, ability to exercise, walk, or move about, and ability to travel short and long distances; their sleep, normal work, recreational activities, enjoyment of life, clothes they chose to wear, and relationships with friends and family members; and how much these problems had caused them to be embarrassed or ashamed during the past 4 weeks (Cronbach α=.94), a three-item measure of pelvic problem impact on sex and a two-item sexual satisfaction measure (Cronbach α=.78). Our primary measures of HRQOL were the mental and physical component summaries of the 12-Item Short Form Health Survey (SF-12)5; these were supplemented by a single-item global measure adapted from the Maine6 and Maryland7 Women's Health Studies (“Which of the following answers describes how you feel about your health right now?”; response options ranged from 1=terrible to 6=delighted).
Hysterectomy- and uterus-related attitudes were assessed with a set of items addressing themes that emerged during focus group discussions.3 Factor analysis of the attitude statements yielded three interpretable factors: “benefits of not having a uterus,” consisting of three items (menstruation is one of the downsides of being a woman, the uterus is useless to women who have completed childbearing, if I had a hysterectomy I would feel happy not to have to worry about birth control; Cronbach α=.50); “value of uterus,” consisting of two statements about the perceived importance of having a uterus for sexual enjoyment and its role in feeling complete as a woman (Cronbach α=.74); and “hysterectomy concerns,” consisting of three statements about feeling older, violated, and sad about losing one's fertility if one had a hysterectomy (Cronbach α=.73). Satisfaction was assessed by asking women who had undergone hysterectomy, “Overall, how do you feel about having had a hysterectomy?”; response options ranged from 1=very dissatisfied to 5=very satisfied. For the analysis of predictors of hysterectomy satisfaction, we used a binary satisfaction outcome (“very” or “somewhat” satisfied compared with all other categories).
To estimate the effects of sociodemographic and clinical characteristics, treatments, HRQOL, sexual function, and attitudes on future hysterectomy use, we fit univariable and multivariable Cox proportional hazard models, with participant age defining the time scale. Values of the hysterectomy event outcome were either observed or censored (uterus retained at end of follow-up). Explanatory variables were grouped into four categories based on how they were modeled: 1) baseline indicators of sociodemographic and clinical characteristics; 2) a time-varying, absorbing state indicator of menopausal status; 3) four time-varying type-specific indicators of nonhysterectomy treatment use; and 4) time-varying indicators of symptom resolution, pelvic problem impact, HRQOL, sexual function, and hysterectomy- and uterus-related attitudes. The nonhysterectomy treatment types included uterine-preserving surgeries that are considered to be alternatives to hysterectomy (eg, myomectomy, endometrial ablation); uterine-preserving surgeries not considered alternatives to hysterectomy (eg, dilation and curettage, polyp removal) but also used to treat symptoms; gonadotropin-releasing hormone (GnRH) agonist; and prescription medications. Each type of treatment was indexed by two variables: a binary indicator of use before enrollment and a time-varying count of subsequent use across follow-up assessments.
We used Cox models to test equivalence of between- and within-person effects of time-varying indicators describing symptom resolution, pelvic problem impact, HRQOL, sexual function, and hysterectomy- and uterus-related attitudes. Each corresponding time-varying variable was initially replaced by two derived explanatory variables: a between-persons variable of baseline values and a within-person variable describing individual change since baseline at each follow-up. The multivariable model was modified in two stages. First, a backward elimination procedure removed explanatory variables with P values exceeding .20. Second, we tested the difference between corresponding between- and within-person parameter estimate pairs. Any nonsignificant difference (P>.05) suggested equivalent between- and within-person effects. In these cases, both corresponding explanatory variables were dropped from the model and replaced with the original time-varying variable; the model was then refit.8,9 As reported below, within the fourth category of explanatory variables, only the original time-varying variables had significant effects.
For the analysis of predictors of satisfaction, univariable and multivariable logistic models regressed the satisfaction outcome onto a set of explanatory variables measured at the interview immediately before hysterectomy and otherwise paralleled those included in the Cox models. The model also included covariates describing respondent age, the time span between the prehysterectomy and posthysterectomy assessments and surgery, and a covariate describing posthysterectomy symptom resolution; the latter covariate adjusted modeled effects for the “successfulness” of the surgical intervention. A final set of models compared prehysterectomy and posthysterectomy attitudes with unadjusted and adjusted paired t tests, which were fit within the linear mixed models framework, including random intercepts for respondents.
Each model was fit to 20 multiply imputed data sets created via a Markov chain Monte Carlo method10 using SAS PROC MI (SAS Institute Inc., Cary, NC). Separate imputation models were fit to support the Cox and logistic regression models, each of which included the modeled outcome and all potential predictors. Imputed values for binary, categorical, and ordinal variables were subsequently rounded to the nearest applicable category.10 All parameter and standard error estimates were calculated by combining results across the imputed data sets.11,12 After a significant omnibus test statistic for the effect of any explanatory variable with three or more categories, all possible corresponding pairwise comparisons were tested, and P values were adjusted using the Hochberg method.13
Of the 1,503 SOPHIA enrollees (although we originally reported that 1,493 women had enrolled in SOPHIA,3 in the process of linking baseline with longitudinal data we identified additional information on 10 participants who were previously excluded; our SOPHIA cohort thus consists of 1,503 women), 1,420 completed at least one follow-up interview or had available chart review data and were therefore included in this analysis. (Final hysterectomy status was obtained via chart review for 188 participants who did not complete all follow-up interviews, 14 of whom had a hysterectomy by the end of the study.) These participants yielded 5,175 person-years of follow-up, and together they constituted a sociodemographically diverse group of women obtaining care at a wide range of practice settings. A total of 207 (14.6%) had a hysterectomy during the study period, and these women were more likely to report symptomatic fibroids and that they did not want to become pregnant at baseline (Table 1).
Clinical symptoms at baseline and several time-varying clinical predictors of hysterectomy emerged as significant predictors in our univariable analyses (Table 2). In addition, six of the seven pelvic problem resolution and impact, sexual function, and HRQOL variables and all three attitudes were associated with hysterectomy use. Many of these variables persisted as predictors of hysterectomy use in the multivariable analyses. For example, participants were more likely to have a hysterectomy if they had fibroids and bleeding (adjusted hazard ratio [HR] 2.28, 95% confidence interval [CI] 1.33-3.91, P=.026) or pressure (adjusted HR 2.68, 95% CI 1.44-4.98, P=.018) compared with bleeding only; did not enter menopause during the study (adjusted HR for entering menopause 0.59, 95% CI 0.38-0.91, P=.017); were treated with GnRH agonist before enrollment (adjusted HR 1.98, 95% CI 1.30-2.99, P=.001); and reported lower levels of symptom resolution (adjusted HR 0.66, 95% CI 0.54-0.81, P<.001). Women who reported higher levels of pelvic problem impact on sex (adjusted HR 1.23, 95% CI 1.09-1.39, P=.001) or who had higher SF-12 mental component summary scores (HR 1.10, 95% CI 1.02-1.18, P=.010) also were more likely to undergo hysterectomy. In addition, higher scores on the “benefits of not having a uterus” scale (adjusted HR 1.24, 95% CI 1.12-1.38, P<.001) and lower scores on the “hysterectomy concerns” scale (adjusted HR 0.80, 95% CI 0.73-0.89, P<.001) were all found to be predictive of hysterectomy use in the multivariable analysis.
The majority of the women who underwent hysterectomy (63.9%) were very satisfied. However, nearly a quarter (21.4%) were only somewhat satisfied, 6.9% were neither satisfied nor dissatisfied, and 7.8% were somewhat or very dissatisfied. Four variables were found to be significantly associated with being very or somewhat satisfied in the multivariable analysis (Table 3). These included site for care (omnibus P=.036, compared with women receiving care at the HMO, women obtaining care at the county facility or a community practice tended to be less likely to be satisfied, odds ratio [OR] 0.22, 95% CI 0.06-0.85, and OR 0.10, 95% CI 0.01-0.75, respectively, P=.057, for both); degree of symptom resolution after the procedure (OR 4.53, 95% CI 1.87-10.97, P=.001); pelvic problem impact before their hysterectomy (OR 2.34, 95% CI 1.07-5.13, P=.035); and the “benefit of not having a uterus” scale score before hysterectomy (OR 2.08, 95% CI 1.22-3.56, P=.008).
We observed large and significant differences on all three attitude scores before and after hysterectomy. Specifically, after undergoing the procedure, these women had significantly higher scores on the “benefit of not having a uterus” scale (4.38 compared with 3.95, Cohen d=0.279, P=.003) and significantly lower scores on the “value of having a uterus” (2.97 compared with 3.77, Cohen d=−0.489, P<.0001) and “hysterectomy concerns” (2.73 compared with 3.08, Cohen d=−0.210, P=.034) scales.
Who undergoes hysterectomy for noncancerous uterine conditions in this country, and who among them is satisfied? Our study confirms previous reports that women who have symptomatic fibroids, who have been treated with GnRH agonists, and who report little or no resolution of their symptoms are more likely to undergo hysterectomy than other women with noncancerous pelvic problems.14 We also have identified a number of nonclinical predictors of hysterectomy use, including HRQOL, sexual function, and, perhaps most importantly, attitudinal factors that underlie women's decisions to undergo hysterectomy and feelings of satisfaction after the procedure.
Most of the HRQOL-related variables were found to be independent time-varying predictors of subsequent hysterectomy use. As we had previously demonstrated, women who reported less symptom resolution were more likely to have a hysterectomy.14 Not surprisingly, in this analysis we also found that women who reported more symptom resolution after hysterectomy were more likely to be satisfied with the procedure. However, after controlling for this variable (and all the other clinical, sociodemographic, HRQOL, and attitudinal variables), we found that greater pelvic problem impact on sex further increased the likelihood that a woman would choose to undergo a hysterectomy, and that greater impact of pelvic problems overall before hysterectomy predicted satisfaction with the procedure. These findings underscore the importance of determining the extent to which symptoms interfere with HRQOL and sexual functioning when counseling patients about hysterectomy and its outcomes. Interestingly, women with higher SF-12 mental component summary scores (indicating better mental well-being) also were more likely to undergo a hysterectomy than women with lower scores. These women may have been more inclined to make a definitive treatment choice than women with lower mental component summary scores. It also may be that providers view patients with better mental well-being as better candidates for major surgery.
Perhaps most noteworthy are our findings regarding the significant role of women's attitudes toward their uterus and hysterectomy in their decision making regarding and satisfaction with this surgery. Ultimately, women who could see the potential benefits of not having a uterus were significantly more likely to undergo and be satisfied with hysterectomy, whereas women who had concerns about this procedure were significantly less likely to make use of this procedure. These findings provide indirect evidence that the decision to undergo hysterectomy was informed by patients' attitudes and beliefs. We cannot comment, however, on the extent to which these attitudes were elicited by or shared with physicians. The substantial changes in these attitudes before and after hysterectomy suggest that some of the hesitations and concerns about undergoing hysterectomy may be reduced if patients are provided with data about how other women have fared, and how their attitudes change, after the surgery.
Clinicians often focus on “objective findings,” such as fibroid size and number, along with patient age and reproductive plans, when they recommend hysterectomy, rather than discussing the impact of symptoms on the patient's quality of life and her attitudes toward her uterus. Our findings further highlight the need to engage women in discussions regarding the potential upsides and downsides of hysterectomy, and to address both their expectations and concerns, to help them arrive at decisions that will lead them to better outcomes than might occur without meaningful discussion. Based on the findings of the current investigation, we infer that patients may benefit from focusing on the following key questions to guide counseling discussions: 1) To what extent has your (bleeding, pressure, pain) resolved on current therapy? 2) How much of an impact has your (bleeding, pressure, pain) had on your ability to have and enjoy sex? 3) To what extent do you see removal of your uterus as a positive and beneficial outcome? 4) What worries or concerns do you have about having surgery to remove your uterus? Patients who report the greatest effect of their pelvic problems and who believe that removal of their uterus is beneficial can expect high degrees of satisfaction from hysterectomy, particularly if their symptoms are likely to be resolved by the procedure. Although some of these drivers of satisfaction may seem self-evident, using a systematic rubric to guide counseling discussions may help patients make more informed and successful treatment choices. Expanding patients' focus from sexual functioning after hysterectomy to include the overall effect of their current symptoms may be particularly beneficial.
We also found that the setting in which care for noncancerous uterine conditions is provided may play a role in being satisfied with hysterectomy. During the SOPHIA study period, alternative treatments such as uterine artery embolization, and newer endometrial ablation options, were available to patients enrolled in the HMO, and the majority of patients receiving alternative surgical treatments at that site did not go on to have hysterectomy.15,16 Given the wider range of alternatives to hysterectomy at the HMO, it could be that the women who availed themselves of hysterectomy were having the most bothersome symptoms, leading to greater relief and satisfaction after the procedure. It also could be that having more options led to more perceived control over health decisions, less resentment of those decisions, and greater satisfaction.
Our study is limited by its lack of geographic diversity. Although we were successful in recruiting a sociodemographically diverse cohort obtaining care in a wide range of practice settings, all of our participants lived and received care in the San Francisco Bay area and therefore may not be representative of women or practice patterns in other parts of the country. This may be one of the reasons for the relatively low hysterectomy rate we observed. Also, of the 3,261 we contacted and were found to be eligible, approximately half (53%) did not enroll because of lack of time or interest. In addition, because we sought to follow the natural history of noncancerous pelvic problems, we included women whose symptoms may never have reached the point that hysterectomy was a reasonable option. This also may be reflected in the low hysterectomy rate. However, it is important to note that our patients were recruited into the cohort at differing stages in the progression of their conditions—they needed to have been seen only once in the previous 12 months. Consistent with our goal to describe natural history, there was no requirement that symptoms be refractory to first-line therapies. That the majority of these patients could be treated without hysterectomy is an important finding. Finally, during the time period that SOPHIA was conducted, a number of alternatives to hysterectomy became available, leading to temporal trends in the use of hysterectomy that might have affected our overall hysterectomy rate.
The Study of Pelvic Problems, Hysterectomy, and Intervention Alternatives was a comprehensive longitudinal study of women with noncancerous uterine conditions that explores a wide range of clinical, HRQOL, and attitudinal predictors of hysterectomy use and satisfaction among a diverse population of women receiving care in a wide range of practice types in a large metropolitan region. We have identified numerous important determinants of hysterectomy use and satisfaction that can be used to inform discussions between patients and their providers regarding the optimal use of hysterectomy and alternative treatments for noncancerous uterine conditions.
1. Merrill RM. Hysterectomy surveillance in the United States, 1997 through 2005. Med Sci Monit 2008;14:CR24–31.
2. Treatment of Common non-cancerous uterine conditions: issues for research. Conference summary. AHCPR Pub. No. 95-0067. Rockville (MD): Agency for Health Care Policy and Research; 1995.
3. Kuppermann M, Learman LA, Schembri M, Gregorich S, Jacoby A, Jackson RA, et al. Effect of noncancerous pelvic problems on health-related quality of life and sexual functioning. Obstet Gynecol 2007;110:633–42.
4. Shumaker SA, Wyman JF, Uebersax JS, McClish D, Fantl JA. Health-related quality of life measures for women with urinary incontinence: the Incontinence Impact Questionnaire and the Urogenital Distress Inventory. Continence Program in Women (CPW) Research Group. Qual Life Res 1994;3:291–306.
5. Gandek B, Ware JE, Aaronson NK, Apolone G, Bjorner JB, Brazier JE, et al. Cross-validation of item selection and scoring for the SF-12 Health Survey in nine countries: results from the IQOLA Project. International Quality of Life Assessment. J Clin Epidemiol 1998;51:1171–8.
6. Carlson KJ, Miller BA, Fowler FJ Jr. The Maine Women's Health Study, I: outcomes of hysterectomy. Obstet Gynecol 1994;83:556–65.
7. Kjerulff KH, Rhodes JC, Langenberg PW, Harvey LA. Patient satisfaction with results of hysterectomy. Am J Obstet Gynecol 2000;183:1440–7.
8. Neuhaus JM. Assessing change with longitudinal and clustered binary data. Annu Rev Public Health 2001;22:115–28.
9. Neuhaus JM, Kalbfleisch JD. Between- and within-cluster covariate effects in the analysis of clustered data. Biometrics 1998;54:638–45.
10. Schafer JL. Analysis of incomplete multivariate data. London: Chapman & Hall; 1997.
11. Meng XL, Rubin DB. Performing likelihood ratio tests with multiply-imputed data sets. Biometrika 1992;79:103–11.
12. Rubin DB. Multiple imputation for nonresponse in surveys. New York (NY): Wiley; 1987.
13. Hochberg Y. A sharper Bonferroni procedure for multiple tests of significance. Biometrika 1988;75:800–2.
14. Learman LA, Kuppermann M, Gates E, Gregorich SE, Lewis J, Washington AE. Predictors of hysterectomy in women with common pelvic problems: a uterine survival analysis. J Am Coll Surg 2007;204:633–41.
15. Gabriel-Cox K, Jacobson GF, Armstrong MA, Hung YY, Learman LA. Predictors of hysterectomy after uterine artery embolization for leiomyoma. Am J Obstet Gynecol 2007;196:588.e1–6.
16. Longinotti MK, Jacobson GF, Hung YY, Learman LA. Probability of hysterectomy after endometrial ablation. Obstet Gynecol 2008;112:1214–20.
17. Nakao K, Treas J. Updating occupational prestige and socioeconomic scores: how the new measures measure up. Sociol Methodol 1994;24:1–72.