KJERULFF, KRISTEN H. MS, PhD; LANGENBERG, PATRICIA W. PhD; RHODES, JULIA C. MS; HARVEY, LYNN A.; GUZINSKI, GAY M. MD; STOLLEY, PAUL D. MD, MPH
Why some women experience positive outcomes and some do not is an aspect of hysterectomy that has not been well studied. Although prospective studies have generally reported positive results from hysterectomy, these studies have included some women who did not experience relief of the symptoms that had presumably led them to hysterectomy, and some women developed new problems and symptoms.1–4 It is not clear whether there are specific factors, such as surgical route, concomitant oophorectomy, or psychosocial factors, that can increase women's risk of poor outcomes after hysterectomy.
The Maryland Women's Health Study was a large prospective study designed to measure the outcomes and effectiveness of hysterectomy for benign conditions, which examined a wide variety of outcome variables including operative and postoperative complications, symptoms, psychologic function, sexual function, quality of life, patient satisfaction, and cost. Women were enrolled from a variety of socioeconomic and racial backgrounds and from urban, suburban, and rural settings, and were followed for 2 years. The objectives of this study were to measure the outcomes and effectiveness of hysterectomy from the women's perspective and to measure the effects of operative route (abdominal or vaginal), oophorectomy, and patient characteristics on outcomes and effectiveness. This article reports the outcomes and effectiveness results for three important aspects of health status: symptoms, psychologic function, and quality of life.
Materials and Methods
This study was approved by the Institutional Review Board at the University of Maryland and at each of the participating hospitals. All participants provided signed informed consent.
There were 1299 women enrolled in the Maryland Women's Health Study between February 1992 and December 1993. A woman was eligible if she was 18 years or older and was scheduled for hysterectomy for a benign indication, including cervical dysplasia and endometrial hyperplasia. Participants who were found to have cancer at hysterectomy were retained. There were 49 hospitals in the state in which hysterectomies were done at the time this study began. These hospitals were stratified on the basis of hysterectomy volume, ranging from stratum 1 (more than 600 hysterectomies per year) to stratum 6 (fewer than 100 per year), and 32 hospitals were randomly sampled within these strata. A total of 28 of 32 hospitals agreed to participate. The number of subjects obtained from each stratum was proportionate to the total number of hysterectomies occurring in that stratum across the entire state. All 663 attending gynecologists at these hospitals were asked to participate, and 406 (61%) agreed. During the enrollment period, 272 of the 406 physicians who agreed to participate performed one or more hysterectomies and provided women for the study. Hysterectomy patients of the participating physicians were identified from the surgical posting lists and were asked to participate. During enrollment, the names of 1823 eligible women were obtained. Interviewers were able to locate 88% of the eligible women, of whom 81% agreed to participate. The only information available concerning the nonparticipants was age, race, indication, date of posting, and date of scheduled hysterectomy. Comparison of participants with nonparticipants indicated that nonparticipants did not differ significantly in age, race, or indication but were significantly more likely to have been scheduled for hysterectomies within 1 week of their posting dates. The usual time between the date of posting and the date of surgery was 2–4 weeks.
Participants were interviewed shortly before surgery (n = 1299) and were reinterviewed 3, 6, 12, 18, and 24 months afterward. The 3-month interview focused on recovery from surgery and did not measure symptoms, quality of life, or psychologic function. Therefore, results from the 3-month interviews are not reported in this article. Medical records from each participant's hospital stay were abstracted by trained medical-records abstractors.
A 5% random sample of all interviews and a 10% sample of medical-record abstractions were validated on an ongoing basis. We did not assess interrater or intrarater reliability in this study.
Principal diagnoses were classified into the following mutually exclusive categories based on the International Classification of Diseases, 9th Revision-Clinical Modification diagnostic codes in the discharge summary: leiomyomas, uterine prolapse, endometriosis, cancer (eg, uterine, cervical, and ovarian), menstrual disorders (eg, menorrhagia and dysmenorrhea), infectious conditions (eg, salpingitis, oophoritis, parametritis, and cervicitis), adnexal conditions (eg, benign neoplasm of the ovary and ovarian cyst), and noninfectious conditions (eg, endometrial hyperplasia, cervical dysplasia, and stress incontinence).
The extent to which hysterectomy provided symptom relief was assessed in terms of eight symptoms: vaginal bleeding, pelvic pain, fatigue, back pain, abdominal bloating, sleep disturbance, urinary incontinence, and activity limitation. At each stage, participants were asked to indicate how much of a problem they had experienced during the previous month with “vaginal bleeding”; “pain, cramps, or discomfort in the pelvic or abdominal area”; and “feeling tired or a lack of energy.” Response options were “big problem,” “medium problem,” “small problem,” and “no problem.” Participants were also asked how often in the previous month they “had pain in your lower back,” “had bloating or swelling of the abdomen,” “had difficulty getting to sleep or staying asleep (for whatever reason),” and “dripped or leaked urine.” Response options to these questions were “all of the time,” “most of the time,” “a good bit of the time,” “some of the time,” “a little of the time,” and “none of the time.” Participants were asked, “How much does your condition limit your daily activities?” with response options of “a lot,” “some,” “a little,” and “not at all.” These symptom questions had been validated in the Maine Women's Health Study.1
The Profile of Mood States is a highly reliable and valid measure of psychologic function often used in health studies.5 Two of the scales from this index were used for this report: depression-dejection and tension-anxiety. Total scores were rescaled to range from 0 (absence of the mood disorder) to 10 (maximal mood disorder) on continuous scales. At each stage, women were also asked “In the past 3 months, have you seen a psychiatrist, psychologist, or other type of therapist for psychologic or emotional problems such as depression or anxiety?”
Several aspects of quality of life were measured using the physical function, social function, and health perception scales from the Medical Outcome Study Short-Form General Health Survey,6 a well-validated and reliable instrument used frequently in outcomes studies. These scores were rescaled to range from 0 (maximal impairment) to 10 (no impairment) on continuous scales. Comorbidities were assessed with the Charlson Comorbidities Index,7 a well-validated weighted index that takes into account the number and seriousness of comorbid diseases.
Operative and postoperative complications were abstracted from the medical records. All in-hospital complications were graded on a scale of 1 (mild), 2 (moderate), or 3 (serious).8 Women were classified into four mutually exclusive categories: no complications, one or more mild complications but no moderate or serious complications, one or more moderate but no serious complications, and one or more serious complications.
For each of the eight symptoms, the midpoint of the response options was used as a cutoff point such that responses were dichotomized as occurring at problematic-severe levels versus mild-none levels. This dichotomization helped clarify the changes in symptom severity from before to after hysterectomy and simplified the presentation of results. Response options were categorized as indicating that the associated symptom occurred at a problematic-severe level “all of the time,” “most of the time,” or “a good bit of the time” for back pain, abdominal bloating, sleep disturbance, and urinary incontinence; was a “big problem” or “medium problem” for vaginal bleeding, pelvic pain, and fatigue; and was a problem “a lot” and “some” of the time for activity limitation. We also counted the total number of symptoms each woman reported having at problematic-severe levels, ranging from 0 to 8.
Scores for the psychologic function and quality-of-life scales were dichotomized also for some analyses. The cutoffs chosen were based on the potential range of scores and previous reliability and validity studies of the instruments.5,6 For the depression and anxiety scales, a score of 2.6 or greater indicated moderate to severe levels of the mood disorders of depression and anxiety and was used as the cutoff to delineate women with none to mild levels of these mood disorders versus those with moderate to severe levels. For the physical function scale, a score of 8.2 or less (the presence of one or more limitations for more than 3 months) indicated poor physical function. For the social function scale, a score of 5.0 or less (limited all, most, or a good bit of the time) indicated limitation in social function. For the health perceptions scale, a total score of 7.0 or less (corresponding to the lowest 20% of scores in a general population sample) indicated that the respondent perceived herself as being in poor health. These cutoffs are equivalent to those suggested by Stewart et al.6
All analyses, other than those in Table 1, were adjusted for within-hospital correlation and nesting of hospitals within strata defined by hospital volume, using generalized estimation equations,9 as implemented in the GENMOD procedure in the SAS statistical software, version 6.12 (SAS Institute, Cary, NC).10
A repeated-measures analysis was used to assess the statistical significance of changes in symptom levels, psychologic function, and quality of life from before hysterectomy to after hysterectomy. Trend analyses were used to examine the effects of trends from 6 to 24 months after hysterectomy.
To measure lack of symptom relief, we calculated the percentages of women who had the same number or more symptoms at the problematic-severe level after hysterectomy than before. Women with no symptoms at problematic-severe levels at both stages of data collection were omitted from calculation of this variable because we could not classify them as having fewer symptoms at problematic-severe levels after hysterectomy than before; nor did it make sense to classify them as lacking symptom relief, that is, having the same number or more symptoms at problematic-severe levels after hysterectomy than before.
Multiple logistic regression was used to identify factors associated with lack of symptom relief at 12 and 24 months, controlling for baseline symptom severity.
We found few significant or consistent differences in outcomes associated with principal diagnoses and therefore did not stratify the results by diagnosis.
Some of the 1299 participants who began this study moved out of state, dropped out of the study, or were otherwise lost to follow-up. By 3 months there were 1244 participants, by 6 months 1225, by 12 months 1188, by 18 months 1174, and by 24 months 1162. The attrition rate over 2 years was 10.6%. Comparison of those who remained in the study with those who did not indicated that those who left were more likely to have low annual income (59.8% at $35,000 or less for leavers versus 42.2% for stayers; P = .001) and were more likely to be black (51.1% for leavers versus 30.2% for stayers; P = .001). At baseline, 44.0% of the participants had annual incomes of $35,000 or less and 32.4% were black (Table 1), which decreased to 42.2% and 30.2%, respectively, 2 years after hysterectomy.
To assess the representativeness of the sample of hysterectomy patients, we used Maryland state discharge summary data. Study subjects were compared with all other women who had hysterectomies in Maryland in 1992 and 1993, excluding those with principal diagnoses of cancer or obstetric conditions. Study participants did not differ significantly from hysterectomy patients with respect to race, marital status, principal diagnosis, complications, comorbidities, or type of procedure (abdominal or vaginal). However, participants were slightly younger (mean age 43.3 versus 44.6 years; P < .001), had shorter hospital stays (mean of 3.4 versus 3.8 days; P < .001), and had lower total hospital charges (mean of $3226 versus $3721; P = .003) than the general population of hysterectomy patients. Participants and nonparticipants also differed with respect to insurance coverage (P = .001); participants were more likely to be insured by a health maintenance organization (33.6% versus 29.6%) and were less likely to be covered by Medicare (4.6% versus 7.1%).
Nearly half of the participants (48.1%) had principal diagnoses of uterine leiomyomas. Other common conditions were menstrual disorders (16.5%), uterine prolapse (13.2%), and endometriosis (8.6%). Less common were noninfectious conditions (5.2%), cancer (3.1%), adnexal conditions (3.0%), and infectious conditions (2.3%). The range of scores on the measure of comorbidities was 0–34. One fourth of the women (24.5%) had scores of 0, 25.6% had scores of 1, 19.3% had scores of 2, and 30.5% had scores of 3 or more.
More than one fifth of the women (21.4%) had no in-hospital complications; 66.8% had one or more mild complications with no moderate or serious complications, 11.1% had one or more moderate complications with no serious complications, and nine women (0.7%) had one or more serious complications. Forty-eight women (4.0%) reported being readmitted to the hospital for a reason related to their hysterectomies during the first year after surgery, and an additional 15 were readmitted during the second year after hysterectomy, for a total of 63 women (5.4%). The most common reasons were incision problems (dehiscence or infection), surgery for adhesions, intestinal blockage, and urinary tract problems.
There was a substantial decrease from before to after hysterectomy in the percentage of women with each symptom at problematic-severe levels (Table 2) and in the percentage with impaired psychologic function and limitations in quality of life (Table 3). For all eight symptoms, the pre- to posthysterectomy comparisons were highly significant across all data-collection stages, and rates of problematic-severe levels for these symptoms remained at decreased levels across the 2-year follow-up. Similar results were seen for the psychologic function and quality-of-life measures. One third of the women (30.2%) who had reported urinary incontinence all or most of the time before hysterectomy had planned procedures concomitant to hysterectomy to correct the problem. Trend analyses from the 6- to 24-months' data-collection stages showed significant improvement in pelvic pain (P < .001), activity limitation (P < .001), abdominal bloating (P = .02), anxiety (P = 007), and physical function (P = .008); borderline significance was found for social function (P = .063).
Although many hysterectomy patients showed substantial improvement in the problems under investigation, some did not. In addition, some women developed problems after hysterectomy that they did not have before in terms of symptoms, psychologic function, and quality of life. The percentage of women whose symptoms were relieved at 12 and 24 months after hysterectomy, and the percentage of women who acquired each symptom are shown in Table 4. Nearly three quarters of the women who scored as depressed before hysterectomy were no longer depressed at 12 months after hysterectomy, and 73% at 24 months after. In addition, 59.8% of those who scored as anxious before hysterectomy were no longer anxious at 12 months after, and this increased to 67.6% at 24 months after hysterectomy.
Most women also showed improvement in the quality-of-life measures, particularly social function. Of those who reported limited social function before hysterectomy, 87.5% no longer reported limited social function 1 year after hysterectomy, and 88.6% 2 years after. Women in the higher income category were less likely to have poor outcomes (the same number or more symptoms after hysterectomy than before) than women with lower household incomes at both 12 and 24 months after hysterectomy, based on multiple logistic regression (Table 5). Having a higher income reduced the odds of a poor outcome by 63% at 24 months after hysterectomy. Race, age, and education were not significantly associated with poor outcomes.
Baseline depression and therapy for emotional problems were significantly associated with poor outcomes. For each increase of 1 point on the 1 to 10 depression scale, there was a 32% increase in the odds of poor outcomes 1 year after hysterectomy. Women who were in therapy for psychologic or emotional problems before hysterectomy had 3.6 times higher odds of poor outcomes at 12 months and 3.5 times higher at 24 months after hysterectomy than women not in therapy.
The only hysterectomy characteristic that was significantly associated with poor outcomes was bilateral oophorectomy, which was statistically significant at 24 months, but not at 12 months after hysterectomy. Variation among hospitals was not significant for any of the analyses in this study.
We found a substantial decrease in depression and anxiety levels after hysterectomy. Three fourths of the women who were depressed before hysterectomy were no longer depressed afterward. These results confirm those in other studies.2,11–15 However, there were some women who were not depressed before hysterectomy but became depressed afterward. For some of these women, it is likely that changes in life circumstances were associated with posthysterectomy depression, but it is also possible that for some, hysterectomy instigated depression.
The results of this study confirm those of another recent prospective study of outcomes of hysterectomy conducted in Maine. In that study, most women undergoing hysterectomy reported substantial improvement in symptoms, quality of life, and psychologic function, with no evidence of regression of improvement over time.1 We found continued improvement over the 2-year follow-up for several factors including pelvic pain, anxiety, and physical function.
The Maine study enrolled a comparison group of women who received nonsurgical medical management for benign gynecologic conditions.15 Although the women who received medical management were substantially less symptomatic at the time of study enrollment than those who had hysterectomies, nearly one fourth underwent hysterectomies during the 1-year follow-up. Of women who continued with nonsurgical treatment, there were no significant improvements among those with leiomyomas over the 1-year follow-up, but those with other conditions, such as chronic pelvic pain, did show some improvement. However, the women enrolled in the comparison group were significantly different from the women who had hysterectomies in terms of education, parity, duration of condition, multiple diagnoses, psychologic function, quality of life, and symptom severity. With such differences between the treatment and comparison groups, it is difficult to delineate the effects of the treatment versus other factors.
The Maine and Maryland studies are the two largest prospective interview studies conducted in the United States to date of outcomes of hysterectomies among women with benign indications. These studies provide strong evidence that hysterectomy substantially improves symptoms for women with multiple and severe symptoms associated with gynecologic disorders such as leiomyomas, abnormal uterine bleeding, and endometriosis. However, in both studies there were some women whose specific symptoms were not relieved and some women who developed new symptoms or other problems after surgery.
Few studies have systematically investigated why some women do not benefit from hysterectomy in terms of symptom improvement or relief. We found that nearly 8% of the women who had hysterectomies in this study had the same number or more symptoms at problematic-severe levels 2 years after hysterectomy as they had before. Assuming that these results are relatively generalizable, this would indicate that approximately 40,000 women who have hysterectomies each year will fail to achieve desired results.16
The large size and diversity of the study sample allowed us to identify women who experienced poor outcomes and compare them with those who benefited from hysterectomy. Several factors were significantly associated with this outcome. We found that women with household incomes of $35,000 or less were less likely to benefit from hysterectomy than women with higher incomes, even 2 years after surgery. This effect was independent of race and education. It is not clear why this would be the case. There are many stressors inherent to poverty in the United States, including living in more dangerous neighborhoods, moving frequently, holding low-paying manual labor jobs, and simply worrying about money. These types of stressors would be present before and after hysterectomy for poorer women and could be associated with symptom severity independent of true condition severity.
Depression and being in therapy for psychologic or emotional problems at the time of hysterectomy substantially increased the risk that hysterectomy would not be effective for symptom relief. Depression increases the risk of mortality after myocardial infarction17 and stroke among hypertensive patients18 and is inversely associated with recovery from coronary artery bypass surgery.19
It is important to note which factors were not related to poor outcomes. Neither in-hospital complications nor readmission due to postdischarge complications was related to lack of symptom relief at 1 year and 2 years after hysterectomy. Abdominal hysterectomy, which is more invasive than vaginal hysterectomy, was not related to lack of symptom relief. This is not to say that these factors do not matter. It is likely that complications and operative route have a substantial impact on recovery from hysterectomy, cost, satisfaction, and other outcomes that were not covered in this study.
It is not clear why bilateral oophorectomy was associated with lack of symptom relief at 24 but not at 12 months. We looked at posthysterectomy hot flushes, which were more common among the women who had bilateral oophorectomies (37.4% at 12 months and 36.1% at 24 months) than those who had unilateral or no oophorectomies (24.3% at 12 months and 23.6% at 24 months), and found that even when we controlled for concurrent hot flushes, the association between bilateral oophorectomy and lack of symptom relief remained nonsignificant at 12 months and significant at 24 months. Postsurgical use of hormone replacement therapy was not associated with lack of symptom relief at 12 and 24 months after hysterectomy.
A limitation of this study was the lack of a control group of women with gynecologic disorders who did not have hysterectomies. We did not include controls because we considered it likely that women who did not have hysterectomies in a prospective study would differ in important ways from those who had hysterectomies, as found by Carlson et al.15 A randomized clinical trial of hysterectomy compared with watchful waiting would have been a stronger design for establishing the specific effects of hysterectomy versus the effects of time, but would not be an ethical study in our opinion. One could argue that without a comparable control group of women who were not treated, it is possible that the improvements in symptoms, quality of life, and psychologic function after hysterectomy would have occurred even if the women had received no treatment. However, these improvements were so substantial and persistent that it seems unlikely that they would have occurred spontaneously over time, and were more likely associated with hysterectomy.
Hysterectomy was not an effective treatment for all women in this study. The women whose symptoms were not relieved after hysterectomy were more likely to be in therapy for emotional problems at the time of surgery and were more likely to have a low household income. These results are important because they underscore the relevance of factors in patients' private lives that determine treatment effectiveness but of which gynecologists might not be aware.