Kuppermann, Miriam PhD, MPH; Learman, Lee A. MD, PhD; Schembri, Michael BS; Gregorich, Steven E. PhD; Jackson, Rebecca A. MD; Jacoby, Alison MD; Lewis, James MD; Washington, A. Eugene MD, MSC
Hysterectomy remains the most common major nonobstetric surgical procedure for U.S. women.1 Over the past 15 years, global endometrial ablation techniques, the levonorgestrel intrauterine system, and uterine artery embolization have been introduced, and advances in laparoscopic surgery have allowed for removal of large myomas and uteri that previously would have required abdominal procedures. As a result, patients with noncancerous uterine conditions now have many more options than were previously available.
With the advent of these alternatives to hysterectomy, deciding on the best treatment for an individual patient has become more complex. Clearly, physicians and patients make decisions across a broad range of treatment options, including the option of no treatment, and these decisions frequently change over time when symptoms persist, yet data are lacking to help patients understand the likely long-term effect of treatment alternatives on the outcomes they value most. Although numerous randomized clinical trials have compared different routes of hysterectomy, uterus-preserving surgical options, and alternative treatments, these studies do not compare symptom trajectories in the years before and after treatment. Longitudinal cohort studies are particularly well suited to studying multiple treatments simultaneously and examining changes in patient outcomes and treatment choices over time. To date, however, such studies have been relatively short term,2,3 limited to a single symptom such as menorrhagia or leiomyomas,4,5 or a comparison of two treatments.6–8
The Study of Pelvic Problems, Hysterectomy, and Intervention Alternatives, conducted from 1998 to 2008, was a longitudinal study of English-speaking, Spanish-speaking, or Chinese-speaking premenopausal women with intact uteri who at the time of their enrollment were experiencing noncancerous pelvic problems (abnormal uterine bleeding with or without leiomyomas, chronic pelvic pain, or pressure resulting from leiomyomas) for which hysterectomy might be an option. The goals of the study were to 1) explore the effect of these symptoms on health-related quality of life (HRQOL) and sexual function; 2) identify static and time-varying predictors of use of, and satisfaction with, hysterectomy; and 3) describe the natural history of these conditions, with a focus on patient-reported outcomes, leading up to the use of hysterectomy and surgical alternatives along with the intermediate-term (4- to 8-year) outcomes of these treatments. The first two topics were addressed in prior publications.9,10 In this article, we report on changes in condition-specific and overall HRQOL experienced by study participants and the extent to which women who underwent hysterectomy and surgical alternatives experienced immediate and sustained improvements in these outcomes.
PATIENTS AND METHODS
Recruitment methods, measures, and baseline characteristics of the Study of Pelvic Problems, Hysterectomy, and Intervention Alternatives participants have been described.9,10 Briefly, we recruited premenopausal women aged 31–54 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), an integrated health delivery system (Kaiser Permanente Northern California), or one of several community hospitals in San Francisco and had not undergone hysterectomy. Women enrolled in the study during one of two recruitment waves: 1998–1999 (Cohort I) and 2003–2004 (Cohort II). Women who had been recruited into Cohort I and were still being followed at the time of the second recruitment wave were invited to reenroll in the study. The final interview was conducted in January 2008, allowing for up to 8 years of follow-up for women in Cohort I who reenrolled in Cohort II and 4 years for those who enrolled only in Cohort I or Cohort II. 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.
After signing informed consent, each participant completed an interviewer-administered questionnaire, which included items related to 1) sociodemographic and clinical characteristics; 2) prior and current use of treatments for their pelvic problems; and 3) eight HRQOL outcomes assessed using three condition-specific measures and five generic measures. To measure occupational prestige, one of our indicators of socioeconomic status, we used an open-ended question regarding the participant's own occupation and that of her spouse or partner (if she indicated that she was married or living with a partner). We then mapped the responses onto U.S. census codes used in the National Opinion Research Center General Social Survey, assigning the associated prestige score from that survey and selecting the highest of the scores for the respondent or spouse.11 The three condition-specific HRQOL measures included 1) a single-item measure of Pelvic Problem Resolution (“To what extent would you say your pelvic problems have been resolved?”); 2) Pelvic Problem Impact Overall (using a 12-item Pelvic Problem Impact Questionnaire); and 3) Pelvic Problem Impact on Sex (using three items: “please tell us how much [bleeding, pelvic pain or discomfort, pelvic problems overall] interfered with your sexual activity during the past 4 weeks”; the score for this scale was the maximum response value to an individual item). The five generic HRQOL outcomes included 1) the Physical and 2) Mental Component Summaries of the Short Form-1212; 3) a two-item sexual satisfaction measure; 4) a single-item global assessment of the participant's feelings about her health; and 5) a computerized preference elicitation exercise that used the time tradeoff metric13 to rate the participant's current health.
At the beginning of the baseline interview, participants were placed into one of five clinical subgroups based on their symptoms and diagnoses (abnormal uterine bleeding without leiomyomas; bleeding with chronic pelvic pain, with or without leiomyomas; pelvic pain only; leiomyomas with bleeding; or leiomyomas with pelvic pressure). They were then asked about treatments they had undergone before study enrollment and, during annual face-to-face follow-up interviews, they were asked about the treatments they had undergone in the previous year. The eight HRQOL measures also were administered at the baseline and all follow-up interviews. Chart reviews for participants who could not be reached for their final interview were conducted to identify the treatments they had undergone since the time of their last interview.
We first divided the sample into three mutually exclusive “observed groups” based on the most invasive surgical intervention (if any) they had undergone during the study period: hysterectomy, uterus-preserving surgery, or no surgery. Within the uterus-preserving surgery group, the last observed surgery was designated as the index intervention. Analyses proceeded in three stages: 1) comparison of baseline characteristics across the three groups; 2) comparison of baseline and last observed HRQOL values within and between the three groups; and 3) examination of presurgical and postsurgical HRQOL trajectories within the uterus-preserving surgery and hysterectomy groups. The primary outcomes for Stages 2 and 3 were the eight HRQOL outcomes described previously; covariates included recruitment wave (cohort), age, race or ethnicity, education, occupational prestige, site for care, clinical subgroup, desire for future pregnancy, and the number of prior uterus-preserving surgeries including those that occurred before enrollment.
After comparing the baseline characteristics of the hysterectomy, uterus-preserving surgery, and no surgery groups (Stage 1), we fit a repeated-measures linear model for each HRQOL outcome, testing effects of the three groups, binary time point (baseline compared with last observed follow-up), and the group-by-time interaction (Stage 2). These models included the covariates listed previously plus a covariate describing the length of time between the baseline and last follow-up interview.
In Stage 3 modeling, the index surgical intervention for each participant was designated as occurring at time zero, and the timing of all interviews relative to the index surgery was calculated. In other words, negative relative time values were used to represent interviews that occurred before surgery and positive relative time values were used to represent those that followed surgery. Across all participants, these relative times ranged from roughly −8 years to +8 years with no individual participant providing more than eight annual observations. Because the timing of the interviews and index surgeries was not systematically linked, relative times could occur at any point along this time scale. Before Stage 3 analyses, we restricted the range from −4 years to +7 years, because the data beyond this range were very sparse.
Within each surgery group (hysterectomy and uterus-preserving surgery), a linear mixed model of each outcome estimated separate presurgery and postsurgery trajectories as well as an effect describing the “immediate” HRQOL change after surgery (the average predicted outcome difference immediately before and immediately after surgery). Initially, these models included four random effects for intercepts, presurgery and postsurgery trajectories, and immediate change; all random effects were allowed to covary. Preliminary modeling suggested that the variance components associated with presurgery and postsurgery trajectories were near zero and nonsignificant for all outcomes within the hysterectomy and uterus-preserving surgery groups. Therefore, the random trajectory effects were dropped from all models, retaining random effects for intercepts, immediate change, and their covariation. In Stages 2 and 3, reported P values were based on z scores.
For each HRQOL outcome within the uterus-preserving surgery group, we report the presurgical and postsurgical slopes, the average predicted immediate change, and the predicted means at 4 years before surgery, immediately before surgery, immediately after surgery, and 7 years after surgery. Within the hysterectomy group, some trajectories were nonlinear and we sought a modeling framework that would approximate the shapes of those trends and ease interpretation by clinicians and their patients. Nonlinear trajectories were represented using segmented regression with linear trajectory segments connected at “turning points.” Each model considered candidate turning points positioned every 3 months from −2 years through +2 years and every 12 months from −4 years to −2 years and +2 years to +7 years. Candidate turning points that improved the model R2 by 0.02 or greater were retained.
A total of 1,503 women enrolled in the Study of Pelvic Problems, Hysterectomy, and Intervention Alternatives.9,10 For this analysis, we included the 1,491 participants who had complete information on all covariates used in the multivariable adjustment. On average, each woman was followed for 3.58 years for a total of 5,338 person-years of followup. These women constituted a sociodemographically diverse population that had undergone a wide range of prior treatments for their conditions and came into the study with varying clinical symptoms, symptom severity, and HRQOL. Most participants (88.3%) had used medications to control their symptoms, and 22.2% had undergone at least one prior uterus-preserving surgery. In addition, the majority (82.6%) felt their symptoms were not at all (47.3%) or only somewhat (35.3%) resolved, and only half were satisfied with their health (47.9%) and their ability to have and enjoy sex (48.7%). Nearly one-third (29.7%) desired a future pregnancy, and over half (59.1%) indicated they would consider having a hysterectomy.
By the end of the study, 13.7% of the women had undergone hysterectomy, 9.0% underwent uterus-preserving surgery as their most invasive treatment, and the remainder (77.3%) had not undergone surgery to treat their symptoms. Stage 1 analyses suggested that the three groups did not differ much with respect to their sociodemographic characteristics (Table 1) with one notable exception: a greater percentage of the women in the uterus-preserving surgery group were college graduates (60.4%) compared with 50.2% in the hysterectomy group and 47.3% in the no surgery group (overall P=.03).
Table 1-a Baseline C...Image Tools
Not surprisingly, differences in the baseline clinical characteristics of these participants did vary by group. Women who reported having leiomyomas with bleeding at baseline, eg, were disproportionally represented in the hysterectomy group: 49.8% of these women had this symptom at baseline compared with only 44.0% of the uterus-preserving surgery group and 35.2% of the no surgery group (Table 1, overall P<.001). Women who at baseline reported desiring a future pregnancy were least represented in the hysterectomy group (20.6% compared with 42.4% of the uterus-preserving surgery group and 29.9% of the no surgery group, P<.001). Whereas 83.6% of the women in the hysterectomy group had indicated at baseline that they would be willing to consider this procedure, only approximately half of the women in the uterus-preserving surgery group (54.7%) and the no surgery group (56.3%) shared this view (P<.001).
Table 1-b Baseline C...Image Tools
Stage 2 analyses compared mean baseline and last observed follow-up values within the hysterectomy, uterus-preserving surgery, and no surgery groups. All groups reported statistically significant improvement on the three condition-specific HRQOL outcomes (Fig. 1, top panel). However, women in the hysterectomy and uterus-preserving surgery groups, who on average were much more affected by their conditions at baseline, experienced substantially more improvement over the course of the study than women in the no surgery group. For example, on the four-point Pelvic Problem Resolution scale, the mean score for women in the hysterectomy group went from 1.9 (corresponding to “somewhat resolved”) at baseline to 3.5 (halfway between “mostly” and “completely” resolved) at the last follow-up, yielding an improvement of 1.6 points (P<.001). Women in the uterus-preserving surgery group had a similarly low score on this measure at baseline (1.8), which improved by 0.9 points to 2.7 (P<.001). Women in the no surgery group, on the other hand, had a score of 2.1 on this scale at baseline, which improved by only 0.2 points by the final interview (P for group-by-time interactions <.001 for all three groups).
Changes in condition...Image Tools
As far as the generic HRQOL outcomes (Fig. 1, panel 2), all groups reported significant improvements on the Current Health Utility and Feelings about Health measures (all P<.01). As before, the hysterectomy and uterus-preserving surgery groups had lower baseline scores and greater improvements than the no surgery group. Interestingly, only the uterus-preserving surgery group demonstrated significant improvement on the Satisfaction with Sex scale (0.5 points, P=.003).
Condition-specific HRQOL trajectories from Stage 3 analyses are shown in Figure 2, and estimated mean scores at various time points are presented in Table 2. We observed postsurgical improvements in all three of these outcomes in the hysterectomy group (indicated by the solid lines in the figure). For example, 4 years before their hysterectomy, these women had an estimated average score of 1.87 on the Pelvic Problem Resolution scale (Table 2). Just before hysterectomy, the predicted mean score had significantly decreased to 1.56; the slope (b) of this line was −0.08 points per year (Fig. 1; 95% confidence interval [CI] −0.13 to −0.02, P=.006). Immediately after hysterectomy, the predicted mean score for this outcome increased to 3.54. In other words, among women who underwent hysterectomy, the average predicted improvement in Pelvic Problem Resolution (the immediate change score) was 1.98 (95% CI 1.82–2.14, P<.001), corresponding to a change from between “not at all” and “somewhat” resolved to between “mostly” and “completely” resolved. Moreover, this improvement was maintained over time, as evidenced by a posthysterectomy trajectory that did not significantly differ from zero (b=−0.02, 95% CI −0.01 to 0.04, P=.18). Whereas the estimated mean immediate changes for Pelvic Problem Impact Overall and Pelvic Problem Impact on Sex were not significant, by 6 months, substantial improvements were reported (b=−2.62, 95% CI −3.23 to −2.02, P<.001 for Pelvic Problem Impact Overall and b=−3.54, 95% CI −4.53 to −2.55, P<.001 for Pelvic Problem Impact on Sex).
Compared with women who opted for hysterectomy, women in the uterus-preserving surgery group demonstrated significant, but smaller, immediate improvement on the Pelvic Problem Resolution scale (the immediate change score was 0.68, 95% CI 0.42–0.93, P<.001; Fig. 2, top panel, dashed line). Unlike the women in the hysterectomy group, these participants also showed immediate improvement on the other two condition-specific measures (immediate change scores were −0.63, 95% CI −0.88 to −0.38, P<.001 and −0.64, 95% CI −1.04 to −0.24, P=.002 for Pelvic Problem Impact Overall and Pelvic Problem Impact on Sex respectively; Fig. 2, panels 2 and 3). Scores on all three condition-specific measures continued to improve over time (b=0.06, 95% CI 0.02–0.10, P=.001; b=−0.10, 95% CI −0.13 to −0.06, P<.001; and b=−0.15, 95% CI −0.20 to −0.10, P<.001, respectively).
Generic HRQOL trajectories are displayed in Figure 3. Presurgical compared with postsurgical changes on these outcomes were less substantial than for the condition-specific outcomes. Among women in the hysterectomy group, the model-estimated mean immediate change was significant for all generic HRQOL outcomes (Table 2). Short Form 12 Physical Component Summary scores suggested a significant decline (−7.68, 95% CI −11.28 to −4.09, P<.001), but by 6 months postsurgery, mean Physical Component Summary scores were significantly higher than presurgery levels. Immediate improvements were observed for the other generic HRQOL measures: Short Form 12 Mental Component Summary (immediate postprocedure change in score=2.20, 95% CI 0.44–3.97, P=.015), Current Health Utility (change=0.10, 95% CI 0.05–0.15, P<.001), Feelings about Health (change=0.86, 95% CI 0.66–1.06, P<.001), and Satisfaction with Sex (change=0.56, 95% CI 0.32–0.79, P<.001). Postsurgical slopes tended to be relatively flat although significantly positive for the Mental Component Summary and significantly negative for the Physical Component Summary and Feelings about Health scale.
Trajectories of gene...Image Tools
For the most part, these outcome trajectories displayed similar patterns among women who underwent uterus-preserving surgery: relatively small—often nonsignificant—immediate postsurgical changes in mean score coupled with relatively flat presurgical and postsurgical HRQOL slopes. In particular, among women in the uterus-preserving surgery group, presurgery and postsurgery trajectories slightly but significantly improved on the Mental Component Summary and Satisfaction with Sex scale, slightly but significantly worsened for Current Health Utility, and stayed generally flat for the Physical Component Summary and Feelings about Health scale.
In this comprehensive prospective study of 1,491 women experiencing abnormal uterine bleeding (with or without leiomyomas), chronic pelvic pain, or pressure resulting from leiomyomas, we observed across-the-board improvements over time on all condition-specific measures and some generic measures of HRQOL with more dramatic changes experienced by women who underwent surgery during the observation period (up to 8 years). Nearly one-fourth of the cohort had surgical treatment with more women undergoing hysterectomy (13.7%) than uterus-preserving surgery (9.0%) as their most invasive treatment. Women who underwent surgery had been experiencing more substantial effects of their symptoms on HRQOL at baseline than women who did not end up having surgery, and these women, on average, also experienced more dramatic improvement in their HRQOL. Those who underwent hysterectomy generally experienced more improvement in the first few years after surgery than women whose most invasive treatment undergone was uterus-preserving surgery, although improvements in Satisfaction with Sex were more substantial in the uterus-preserving surgery group, and most differences tended to converge over time. Although women who did not undergo surgery also showed significant improvements from baseline to the last follow-up interview on most of the measures for which significant improvements were observed among women who did have surgery, these improvements were smaller in magnitude, and, at the time of the final follow-up interview, their mean estimated scores were almost always lower.
How can our findings be used to help patients engage in informed shared decision-making with their gynecologists? First, by focusing on eight different measures of HRQOL, we believe our study provides substantial and nuanced evidence on the patient-centered outcomes women care most about. These findings could be explained to patients to help them better understand the effect that their treatment is likely to have on their HRQOL. Moreover, in addition to comparing changes in HRQOL from baseline to final interview, we modeled trajectories over time before and after surgical events, providing a more complete view of the HRQOL effect experienced by women who had a hysterectomy or uterus-preserving surgery. The resulting plots show temporal patterns in the 4 years before hysterectomy and uterus-preserving surgery, the postsurgical changes, and subsequent patterns over 7 years. These graphic depictions could be used to develop materials to help counsel women as they consider a variety of treatment approaches.
Additionally, our findings can help patients understand the potential effect of treatment alternatives on their overall HRQOL, of which the condition being treated is just a part. Generic HRQOL measures change in response to an individual's overall health, which over time is more related to the effect of chronic diseases and other factors than to treatment of pelvic problems earlier in life. So it is not surprising that the effects of treatment were more readily observed on the condition-specific measures. Finally, because our study population was large, diverse, and recruited from a variety of practice types, the findings can be applied broadly to the wide range of patients seen by practicing gynecologists.
Several limitations of our study deserve comment. First, by conducting an observational study, inferences about the effect of hysterectomy and uterus-preserving surgery on HRQOL are constrained by selection bias among those who underwent these procedures. In our earlier work, we found that women's attitudes toward their uterus and hysterectomy were predictive of hysterectomy use and satisfaction.10 Treatment decisions for the rest of the cohort were likely based on variety of specific factors, including the underlying cause of the symptoms, the gynecologist's expert opinion on which treatments to offer, and the patient's desires and preferences. In addition, numerous other factors that we did not measure may have affected the HRQOL of the participants over time. If we had conducted a randomized study, much more definitive conclusions about the effect of these procedures on HRQOL could be made. From our study, we can only comment on HRQOL trajectories before and after surgery among women who underwent hysterectomy or uterus-preserving surgery.
Second, we relied solely on patient-reported data for all of the independent variables; only hysterectomy use was verified by chart review. As such, we could not confirm whether the symptoms the participants reported experiencing were what the health care provider was treating them for, and we could not control for other clinical or biologic factors that may have played roles in their decisions to undergo hysterectomy or uterus-preserving surgery as well as in their HRQOL over the course of the study. Finally, although we were successful in recruiting a sociodemographically diverse population of women seeking care at differing types of medical practices, all of the participants were from the San Francisco Bay area, potentially limiting the generalizability of the study.
Despite these limitations, we believe our results shed important new light on the HRQOL effect of hysterectomy and alternative approaches to the management of noncancerous pelvic problems, which can be used by gynecologists and their patients in making informed treatment decisions that reflect the outcomes women value most. Because we followed participants who underwent uterus-preserving surgery or hysterectomy for several years before and after their procedures, and were able to compare their improvements over time with those of women who did not undergo surgical treatment, our study allows for greater context in interpreting intermediate-term gains from hysterectomy and uterus-preserving surgery than others have been able to provide. In particular, although we observed small but statistically significant improvement over time among women who did not undergo surgery, the improvements with uterus-preserving surgery and hysterectomy were substantially greater. We believe these findings are of particular value to patients and should be incorporated in counseling regarding treatment approaches for noncancerous pelvic problems.
© 2013 by The American College of Obstetricians and Gynecologists.