Incidence, time trends, laterality, indications, and pathological findings of unilateral oophorectomy before menopause

Between 1950 and 2007, 1,838 women younger than age 50 years underwent UO while living in Olmsted County, MN. About 60% of the UOs occurred at the time of hysterectomy. Age-specific incidence rates are provided in Table 1 for all women who underwent UO and separately for the subgroup of women who underwent UO with hysterectomy over six time periods (five decades and eight additional years). Pooling all oophorectomies over the study period of 58 years, UO was more common than bilateral oophorectomy through age 39 years, at which time the incidence of bilateral oophorectomy surpassed that of UO ( Figure 1 , panel A). The left ovary was removed more frequently than the right ovary for all ages, but particularly for women between the ages of 35 to 49 years (overall 55.3% left vs. 44.7% right; P < 0.0001; Figure 1 , panel A). Temporal trends of age-standardized incidence rates showed that a decrease in UO rates coincided with an increase in bilateral oophorectomy rates ( Figure 1 , panel B). The incidence rates for UO increased from 1950 to 1974, but decreased after 1974. By contrast, bilateral oophorectomy was performed less frequently than UO from 1950 through 1969, was performed as frequently as UO from 1970 through 1984, and then surpassed UO starting in 1985 ( Figure 1 , panel B). 1 Temporal trends also demonstrated a difference in laterality of UO. Age-standardized incidence rates of left UO were higher than of right UO from 1950 through 1984 (79.9 left vs. 54.9 right per 100,000 person-years), but were similar in the most recent 23 years (34.5 left vs. 34.6 right per 100,000 person-years; 1985–2007). We divided the cohort into two age groups to approximate the premenopausal (0–39 years) and perimenopausal (40–49 years) periods. Left UO was performed more commonly than right UO in women of age 0–39 years until 1979 (45.4 left vs. 34.4 right per 100,000 person-years; Figure 2 , panel A) and for women aged 40–49 years until 1989 (170.2 left vs. 102.3 right per 100,000 person-years; Figure 2 , panel D). We investigated time trends for women who underwent UO with and without concurrent hysterectomy. There was no clear trend among women younger than 40 years ( Figure 3 , panel A). By contrast, among women 40–49 years of age, UO was more commonly performed concurrently with hysterectomy through 1989 ( Figure 3 , panel B). Figure 2 also shows the trends over time in laterality for the women who had a UO with a medical indication, divided into two age groups (panels B and E). In both age groups, the left ovary was removed more commonly than the right ovary through 1979. The most common indication for UO was a benign condition of the ovary for all women and for women with concurrent hysterectomy ( Table 2 ). Among women with concurrent hysterectomy, the right ovary was removed for the indication of a benign condition more frequently than the left ovary (69.6% right vs. 63.3% left, P = 0.03, Table 2 ). All medical indications were similar in the right and the left ovary, except “other” medical indications which were more common in the right ovary (3.0% right vs.1.5% left; P = 0.02). Women who underwent UO without a medical indication were the second largest group by indication in the overall incident series (20.2%) and among women with concurrent hysterectomy (31.8%; Table 2 ). The left ovary was removed more frequently without a medical indication than the right ovary in both the overall group (22.8% left vs.17.0% right, P=0.002; Table 2 ) and among women with concurrent hysterectomy (34.8% left vs. 27.8% right; P = 0.01). In women of age 0–39 years without a medical indication, the laterality differences were small throughout the study period ( Figure 2 , panel C). By contrast, in women aged 40–49 years, the left ovary was removed more commonly than the right through 1989 ( Figure 2 , panel F). Table 3 shows the pathology found in the removed ovary over the entire study period by laterality. In the overall group, cysts were the most common pathologic finding and had similar frequency on both sides (29.2% left vs. 29.4% right; P = 0.92; Table 3 ). Apparently normal ovaries were the second most common pathologic finding overall and occurred similarly in the right and left ovary (19.4% left vs.19.5% right; P = 0.97). By contrast, endometriosis occurred more frequently in the left ovary than in the right ovary (19.0% left vs. 13.4% right; P = 0.001). Although not statistically significant, primary ovarian cancer occurred more frequently in the right than in the left ovary (1.5% right vs. 0.7% left, P = 0.10). Other pathology, such as ectopic pregnancy, was more common in the right than left ovary (7.1% right vs. 4.4% left, P = 0.01). Table 3 also shows the pathology found in the ovary removed from women with or without a medical indication. The pathology differed by laterality among women with a medical indication for UO. Endometriosis was more common in the left ovary (23.9% left vs. 15.8% right, P = 0.0001). “Other” pathologies were more common in the right ovary (7.8% right vs. 4.7% left, P = 0.02). By contrast, most ovaries removed without a medical indication were apparently normal on pathology (52.2% left vs. 55.7% right, P = 0.51), and the pathology did not differ significantly by side ( Table 3 ). Indication and resultant pathology were consistent. Of the 1,352 women with an indication of benign condition ( Table 2 ), 1,197 (88.5%) had pathology consistent with a benign process (e.g., endometriosis or benign tumor). Among the remaining 155 women (11.5%), 137 had normal ovaries, 13 had ovarian cancer, and 5 had unknown pathology. Of the 12 women with an indication of primary or metastatic cancer, 8 (66.7%) were confirmed by pathology.

Secular trends in the incidence of UO are influenced by trends in the incidence of bilateral oophorectomy, 1 which in turn are determined by evolving evidence and beliefs about the risk-to-benefit balance between conserving the ovaries to reduce menopausal symptoms and sequelae 4 – 6 and removing the ovaries to decrease the risk of ovarian or breast cancer and the risk of reoperation. 8 , 9 Many studies have influenced this balance over time, leading to major trends in gynecological practices. 2 , 19 – 20 Our study showed a link between the trends in bilateral oophorectomy and the trends in UO rates over the most recent 58-year period in a well-defined US population. UO without a medical indication has been influenced by surgical preferences and traditions. Between 1950 and 1985, preservation of ovarian function during concurrent hysterectomy was the tradition, so there was an attempt to save at least one ovary. 21 – 22 In our study, the incidence of left oophorectomy was higher than the incidence of right oophorectomy until approximately 1985. In reviewing medical records, we learned that it was customary to remove one ovary to reduce the risk of future reoperation. The left ovary was preferentially removed if both ovaries appeared normal, because of its proximity with the sigmoid colon. 23 The decline in this practice over time coincides with the increased use of ultrasound for diagnostic purposes. 24 In the women who underwent UO for a medical indication, the pathology differed between the left and right side. We confirmed the findings from previous studies that endometriosis is more common in the left than in the right ovary. 25 – 26 Two theories behind the formation of endometriosis may explain this laterality. First, the retrograde menstruation theory postulates that endometrium is expelled from the fallopian tubes into the peritoneum. Elimination of this endometrial tissue may be less efficient on the left side because of the anatomic proximity with the colon 27 or the decreased flow of peritoneal fluid on the left. 28 Therefore, tissue deposited on the left ovary would have greater opportunity to adhere and implant. The second theory is based on the functional difference between ovaries. The right ovary ovulates more frequently than the left and therefore, has higher localized progesterone production which suppresses endometriosis. 29 – 31 Although the initial time segment of our study was based on previously collected data (38 years), our focus on laterality, indication, and pathologic findings in UO was novel and prompted a complete re-analysis of the data. We combined the older data and the new data from the most recent 20 years to provide a long-term perspective on surgical practices (58 years). The strengths of our study include documentation of the laterality of oophorectomy and of whether a contralateral oophorectomy had previously been done in the medical records, avoiding recall bias and inaccuracy of reporting. Additionally, availability of incidence rates for bilateral oophorectomy in the same population from a previous study allowed the comparison of rates for both surgeries over time. 1 , 13 A first limitation of this study is the local nature of UO practices. Surgical practices typical of the Mayo Clinic and other care providers in Olmsted County may not reflect national trends. A literature review failed to reveal national or regional guidelines for UO in the first 4 decades of our study. Thus, local surgical preferences and traditions may have influenced the decision to remove the right or the left ovary without a medical indication. Second, we may have undercounted women who resided in Olmsted County but underwent a unilateral oophorectomy at a medical facility not included in the REP, or women who were incorrectly coded in their medical records or in the electronic indexes of the REP system. Based on our experience in manually abstracting medical records, 32 – 33 these two events are rare and have not changed over time. Finally, our population includes primarily white women of northern and central European descent, thus, our findings may not be generalizable to other populations or ethnic groups. However, the population is similar to a large segment of the US population. 16

UO is thought to be safer than bilateral oophorectomy in premenopausal women because of the assumption that the remaining ovary continues to provide adequate amounts of circulating hormones. However, several studies of UO have demonstrated diminished function of the contralateral ovary following surgery either with or without concurrent hysterectomy. 10 – 12 The effect of the removal of only one ovary on the timing of menopause and on subsequent mortality is understudied. Further investigation of outcomes associated with UO will provide better practice guidelines, especially for those women who undergo oophorectomy without a medical indication. One study showed no difference in the risk of myocardial infarction outcomes after UO, 34 and a study from this Olmsted County cohort showed a mild reduction of subsequent cardiovascular mortality. 14 However, UO in women younger than 42 years was associated with increased risk of parkinsonism and of cognitive impairment or dementia in this Olmsted County population. 32 – 33 Future research should confirm these initial associations of UO with cardiovascular disease and neurologic disease, and should focus on laterality. We may be overlooking the importance of anatomical and functional differences in the right and left ovaries. Viewing the ovaries as interchangeable and symmetrical may lead to surgical practices that have detrimental long-term effects.

This study was based on a previously described cohort of women who underwent UO from 1950 to 1987, 1 , 6 , 13 – 14 and on a new cohort of UO performed from 1988 to 2007 that has not been previously reported. For the 1950–1987 cohort, all analyses and results were novel because of the focus on laterality, indication, and pathologic findings in UO. Thus, in total, we studied the geographically defined population of Olmsted County, Minnesota, USA from January 1, 1950 through December 31, 2007 (58 years). The county population was predominately urban, with more than 70% of county residents living in the city of Rochester. The remainder of the county was suburban or rural, consisting of several smaller communities. Extensive details about the Olmsted County population were reported elsewhere. 15 – 18 Women who underwent UO before 50 years of age were identified using the Rochester Epidemiology Project, a unique population-based medical records-linkage system. 15 – 18 Residents of Olmsted County are served by a limited number of healthcare facilities, and all these facilities participate in the medical records-linkage system. The nearest metropolitan area is 90 miles away and the population is stable. A tertiary care center is among the healthcare facilities (Mayo Clinic), so that referrals out of the county are limited. 15 – 18 The methods used to identify women who underwent UO have been described previously. 1 , 13 Briefly, the electronic indexes of the system were searched for the surgical procedure codes for UO. We included all women undergoing a first UO during the study period, defined as complete removal of one ovary. Women who underwent a first unilateral oophorectomy before the study period, and a second UO during the study period were excluded from incidence calculations (considered equivalent to bilateral oophorectomy). Patient’s date of birth, date of surgery, and the co-occurrence of hysterectomy at the time of oophorectomy were recorded. A nurse or a physician abstractor reviewed complete inpatient and outpatient records, including the surgeon’s narrative report of the operation, to confirm the type of surgery and to define the primary ovarian indication for the surgery. Indications included cancer or cancer-related diagnoses, benign condition, torsion, and inflammatory disease or oophoritis. A woman was considered to have no medical indication if the oophorectomy was performed in the course of another surgery but without any ovarian condition, or was defined as elective by the surgeon. We report the results separately for women who had a UO with and without a medical indication because laterality and pathology were hypothesized to differ by indication. Because UO was commonly performed at the time of hysterectomy, the incidence rates and indications are also shown separately for the subgroup of women who underwent UO with concurrent hysterectomy (or had hysterectomy before the UO). The pathology report was based on the evaluation of frozen sections of the ovary at the time of the surgery and on the subsequent confirmation on permanent sections. 13 The medical records abstractor categorized pathology as primary ovarian cancer or metastatic cancer, endometriosis, cyst, benign tumor, oophoritis or inflammatory process, atrophy, or other lesion (including ectopic pregnancy, adhesions, abscess, and infarction). Ovaries that had no pathologic diagnoses were labeled as “apparently normal.” Incidence rates were calculated by dividing the number of women who underwent a first UO by the number of women at risk in the Olmsted County population and then multiplying by 100,000. Population denominators were derived from the US Decennial Censuses for the years 1950 to 2010, and intercensal year population denominators were linearly interpolated. Where appropriate, overall incidence rates were age-adjusted to the total US female population from the 2000 Census. The distributions of indications and pathologies for the left and the right ovary were compared using chi-square tests. All analyses were conducted using SAS v. 9.2 (SAS, Cary, NC), and tests of significance were undertaken at the two-tailed alpha level of 0.05. The institutional review boards at Mayo Clinic and Olmsted Medical Centers approved the study.