Opioid Prescribing Patterns, Patient Use, and Postoperative Pain After Hysterectomy for Benign Indications

Opioid misuse and abuse has been described by the US Centers for Disease Control and Prevention as one of the top five public health threats in 2016. 1 While providing adequate postoperative pain control has been deemed a fundamental right, 2 surgeon overprescribing has been implicated in opioid diversion, potentially fueling the opioid epidemic. 3 Thus, surgeons must find the best balance between over- and under-prescription, highlighting the need for improved and personalized postoperative opioid prescribing guidelines. Currently, there are no evidence-based guidelines to inform surgeons on appropriate opioid prescribing for the management of postsurgical pain. Recent studies have found that most patients undergoing various surgical procedures 4 – 7 , including cesarean section 8 , 9 , are prescribed opioids in excess of what they use. Despite the era of personalized medicine, most studies did not investigate whether individual patient factors influence opioid use and postsurgical pain during the postoperative period. These data are critical for the development of evidence-based opioid prescribing guidelines tailored to individual patient characteristics. Hysterectomy is the most common surgical procedure performed in non-pregnant women 10 , 11 , making this a critical target for improvement in opioid prescribing. As part of a quality initiative project, we sought to prospectively quantify physician prescribing patterns and patient opioid use in the first two weeks after hysterectomy for benign indications. Our secondary aim was to determine whether individual patient and surgical factors predict patterns of postsurgical opioid use and pain after hysterectomy.

During the study period, 108 women underwent hysterectomy for benign indications. Of these, five women were not eligible to participate (one did not speak English, one had renal failure, and three had a history of addiction or buprenorphine use prior to hysterectomy). Of 103 eligible patients, 102 (99%) agreed to participate, including 44 (43.1%) laparoscopic, 42 (41.2%) vaginal, and 16 (15.7%) abdominal hysterectomies. Telephone surveys used to ascertain the primary outcome (total opioid consumption in the two weeks after hysterectomy) were successfully completed on 89 (87%) of participants. Diaries which collected secondary outcomes (daily pain and opioid use) were returned from 60 (59%) of participants ( Figure 1 ). Relative to women who returned their 2-week diary, diary non-responders were younger (44.7 ± 11. vs 51.8 ± 11.2 years, p=0.002), reported higher Fibromyalgia Survey Score (9.7 ± 5.5 vs 5.6 ± 3.6, p<0.001), and were more likely to be non-white (26.1% vs 10.7%, p=0.04), prior opioid user (21.7% vs 3.6%, p=0.005), smoker (22.2% vs. 3.6%, p=0.004), and have a history of depression (41.3% vs 21.4%, p=0.03). Clinical characteristics and intraoperative findings of the sample are summarized in Table 1 . Surgeon prescribing patterns and patient opioid use in the 2 weeks after hysterectomy by route of surgery are reported in Table 2 . Overall, the median amount of opioid prescribed upon discharge after hysterectomy at our institution was 200 oral morphine equivalents (OME), equivalent to 40 tablets of 5-mg hydrocodone. However, patients reported using approximately half of the opioids prescribed with a median excess of 110 OME per patient (interquartile range 40–150), equivalent to 22 unused tablets of 5-mg hydrocodone. In the univariate analyses, average opioid use and duration of opioid use in the first two weeks after hysterectomy did not differ significantly according to surgical approach in this sample. Overall, 97% women reported adequate or good postoperative pain control, and approximately 40% felt that they received too much opioid after surgery. Multivariable linear regression identified 7 factors that best predicted total postoperative total opioid consumption in the 2 weeks after hysterectomy ( Table 3 ). For example, interpretation of this model indicates that each 1-point increase in Fibromyalgia score is associated with 30.8 more OME’s used in the 2-weeks after hysterectomy, after controlling for all other variables in the model. Women with a history of endometriosis (but not necessarily those with endometriosis at the time of hysterectomy) used 190.8 more OME in the 2 weeks after hysterectomy, while those who were using medication for neuropathic pain used 178.9 OME less. There was no difference in predicted opioid use when comparing vaginal to laparoscopic hysterectomy. Predictors of total opioid consumption after excluding chronic opioid users were similar to those in the total sample, with preoperative Fibromyalgia Score remaining strongly predictive of total opioid consumption ( Table 4 ). The mean Fibromyalgia Survey Score prior to hysterectomy was 7.5 ± 5.0 with a range of 0 to 24. The cohort was divided into equal tertiles based on the distribution of the preoperative Fibromyalgia Score (Low: 0 to 4, Moderate: 5 to 10, High, 11–31), as previously described. 15 , 16 Average daily postoperative pain and opioid use in the first two weeks after hysterectomy, stratified by tertiles of preoperative Fibromyalgia Score, are shown in Figures 2 . Linear mixed regression models confirmed that patients in the highest tertile of Fibromyalgia Score reported significantly higher postoperative opioid consumption each day after hysterectomy (13.9 greater OME at baseline [95% CI 3.0 – 24.8], p=0.020) with the difference decreasing by 0.9 OME per day ([95% CI 0.3–1.5], p=0.002) relative to patients in the moderate or low Fibromyalgia groups. Furthermore, patients in the highest tertile of Fibromyalgia Score reported significantly higher postoperative average pain scores (2.4 points higher on 0–10 NRS at baseline, [95% CI 0.9 – 4.0], p=0.00400); relative to the moderate or low Fibromyalgia groups. While those in the moderate and low Fibromyalgia group reported improvement in pain over time after hysterectomy, the patients in the highest tertile reported worsening pain after surgery (increasing by 0.2 points per day [95% CI 0.1 – 0.3]).

We found that gynecologists in a large academic medical center prescribe approximately twice the amount of opioids than the average patient uses after hysterectomy. Individual patient factors that independently predicted two-week opioid consumption after discharge were similar to those we previously reported to predict opioid consumption prior to hospital discharge, 16 and included route of hysterectomy, Fibromyalgia Survey Score, preoperative opioid use, and history (but not presence) of endometriosis. Reasons for surgeons providing excess opioids are likely multifactorial, and probably relate to lack of evidence-based guidelines on appropriate postoperative opioid prescribing, desire to improve patient satisfaction, and reduce postoperative opioid refill requests. However, recent studies found no association between the amount of opioid prescribed and the likelihood of refill request after surgery 20 , and no association between the initial postoperative prescription, patient satisfaction 8 , or the Hospital Consumer Assessment of Healthcare Providers and Systems pain measure. 21 These findings should reassure surgeons that reducing the amount of opioid provided after surgery should still provide adequate pain management while maintaining patient satisfaction in most patients. Our findings of opioid over-prescription are similar to recent reports in other surgical cohorts 4 – 7 , including cesarean section. 8 , 9 However, few studies have examined whether individual patient factors predict outpatient opioid use. One recent study reported that individual in-hospital opioid consumption was correlated with post-discharge opioid use. 9 Similar to our findings, these data support the notion that individual patient factors influence postoperative opioid use and pain, independent of surgical factors. Thus, a “one-size-fits-all” approach to prescription guidelines may miss an important but feasible opportunity to personalize postoperative pain management. Our findings indicate that a validated, self-report measure of centralized pain, the 2011 Fibromyalgia Survey Score 12 , influences patient opioid use and pain recovery after hysterectomy. Centralized pain disorders refer to any chronic pain condition where central nervous system dysfunction plays a role in augmenting and maintaining pain. 14 This phenotype is characterized by multifocal pain and co-occurring somatic symptoms, including fatigue, sleep and mood disturbances. We have previously demonstrated that the Fibromyalgia Score predicts total opioid use during hospitalization after hysterectomy 16 and knee or hip arthroplasty 15 , as well as persistent pain 6 months after arthroplasty. 22 We now demonstrate this measure is also a strong predictor of total opioid use, time to opioid cessation, and pain recovery in the two weeks after hysterectomy. For example, a woman with a Fibromyalgia Score of 12 (who does not meet diagnostic criteria for fibromyalgia) is predicted to use 300 more OME (i.e. 60 more 5mg hydrocodone pills) in the two weeks after hysterectomy than someone with a score of 2, after adjustment for other covariates. Furthermore, women in the top Fibromyalgia Score tertile used more opioid for longer periods of time, with minimal improvement in pain in the two weeks after hysterectomy. These findings are consistent with published data suggesting opioid non-responsiveness in patients with centralized pain. 23 , 24 Strengths of the study include novel analysis of “centralized pain” as predictor of an individual patient’s postoperative opioid use and prospective data collection in consecutive women undergoing hysterectomy. Despite its strengths, there are several limitations. Data from our large, academic hospital may not be generalizable to other populations and hospital settings. The study group included relatively few abdominal hysterectomies, which limits interpretation of findings in this cohort. Opioid use after discharge was based on self-report, since we do not routinely ask patients to return to clinic for a pill count after surgery. Finally, only about 60% of patients returned their postoperative diaries which provided data for our secondary outcomes. Although this may limit interpretation of daily opioid use and pain scores after hysterectomy, we successfully collected total opioid use on nearly 90% of our participants through telephone surveys. Furthermore, the diary non-responders were more likely to have multiple characteristics that were associated with higher opioid use and pain scores (higher preoperative Fibromyalgia Score and chronic opioid use). Thus our findings likely underestimated the difference in pain and opioid recovery in women with high versus low Fibromyalgia Scores. The opioid epidemic may be partly fueled by over-prescription of opioids postoperatively. 3 While surgeons must adequately control postoperative pain, most surgeons prescribe opioids in excess of the amount women use, making this an important target for improvement for gynecologic surgeons. A personalized approach to prescribing opioids for postoperative pain may be feasible, and should focus on surgical approach and individual patient characteristics, such as measures of centralized pain. Multimodal postoperative analgesia that utilizes opioid and non-opioid strategies should be considered, and may be especially beneficial for patients with centralized pain. 25

The institutional review board at the University of Michigan determined that this quality initiative project was exempt and provided approval for publication because our study did not fit the definition of human subjects research requiring IRB approval. All consecutive, English-speaking patients scheduled for hysterectomy for a benign indication at a single, university hospital from August 31 st to December 18 th , 2015 were prospectively recruited on the day of surgery. Women with a cancer or obstetric indication, medical comorbidity that significantly altered postoperative opioid management, history of opioid addiction or substance abuse were excluded. Patients were informed that this quality initiative involved completing a brief set of questionnaires prior to surgery, a pain and medication daily diary for 14 days after surgery, and a brief telephone survey approximately 14 days after surgery. The primary outcomes were total opioid prescribed by the surgeon and total opioid consumption in the 2 weeks after hysterectomy. Secondary outcomes included daily opioid use and daily postoperative pain severity scores in the 2 weeks after hysterectomy. In the preoperative holding area, patients completed brief standardized validated self-report measures of pre-surgical pain, associated somatic symptoms, and preoperative opioid use. The primary predictor of interest was the 2011 Fibromyalgia Survey Score (Fibromyalgia Score) 12 , which is the sum of the total number of painful body areas (0 to 19 points) and the severity of fatigue, trouble thinking, and headaches (0 to 12 points). This measure has demonstrated good reliability, convergent and discriminant validity, and has been used as a continuous measure to assess the degree of “centralized pain” 12 , 13 , defined as any central nervous system dysfunction that plays a role in pain amplification. 14 Centralized pain disorders are characterized by widespread pain and co-occurring somatic symptoms, such as fatigue, sleep and memory difficulties. 14 The Fibromyalgia Score has been shown to be highly predictive of acute postsurgical opioid use during hospitalization among patients undergoing hysterectomy or knee or hip arthroplasty. 15 , 16 Patients also completed self-report measures of average pelvic and overall body pain in the month prior to surgery using an 11-point numeric rating scale (NRS) with 0 indicating “no pain” and 10 indicating “pain as bad as you can imagine.” Patients were classified as “preoperative opioid user” if they reported using opioids more than once per week for any reason in the month prior to hysterectomy. A standardized case report form was used to abstract patient and surgical characteristics from the electronic medical record by a gynecologist (SA, ST, or EM) blinded to the study results. We focused on characteristics prior to hospital discharge, as these would be the available data that could be used to develop postoperative opioid prescription guidelines. Demographic and medical history included age (years), race (white or non-white), current smoking status, current use of medications for depression, anxiety or neuropathic pain. Indications for hysterectomy were determined by review of operative report and outpatient clinic visits. Patients could have more than one indication for hysterectomy, and these diagnoses were not considered mutually exclusive except for dysmenorrhea and chronic pelvic pain. Dysmenorrhea was defined as pelvic pain that occurred primarily during menstruation, and chronic pelvic pain was defined as non-cyclic pelvic pain that was not limited to the time of menstruation. Route of surgery was categorized as abdominal, vaginal, or laparoscopic (which included traditional and robot-assisted laparoscopy). Additional surgical procedures performed at the time of hysterectomy were noted. When identified, endometriosis was assigned a stage according to the revised American Fertility Society endometriosis scoring system. 17 Adhesions were scored using the validated scoring system developed by the Adhesion Scoring Group. 18 The type of opioid, dose, and total number of pills prescribed in the first 14 days since discharge, were obtained from the electronic medical record. Postoperative opioid use and pain symptoms were collected by a telephone survey on postoperative day 14 and secondarily using daily diaries completed for the first 14 days after hysterectomy. Telephone interviews were conducted using a standardized script. If we were unable to reach the patient on the first attempt, a total of 3 attempts were made. The telephone survey queried the name(s), dose, and total number of opioid pills used since discharge, request for opioid refills since discharge and number of remaining opioid pills. If a patient requested a refill, all patients confirmed that they used all pills in the original prescription. Using daily “diaries”, participants also recorded their average pain (0–10 NRS), and the name, dose, and total number of opioid pills used at the end of each day for the first 14 days after hysterectomy. The last day of opioid use was determined from completed diaries. All opioid data used in this study was converted to oral morphine equivalents (OME) using the guidelines published by the Center for Disease Control. 19 Data analyses were performed using STATA 14.0 (StataCorp, College Station, TX) and R 3.3.1 (The R Foundation for Statistical Computing). Descriptive data were analyzed using x 2 or Fisher’s exact test where applicable for categorical variables, Student’s t-test or one-way analysis of variance (ANOVA) for normally distributed continuous variable, or via nonparametric Wilcoxon Rank-Sum test if the distribution was non-normal. Descriptive statistics on continuous variables showing no significant skewness (d’Agostino test) were reported as mean and standard deviation and statistics on skewed variables were reported as median and interquartile range. To explore the relationship between Fibromyalgia Score, daily opioid use and pain recovery, the cohort was divided into tertiles based on the distribution of the Fibromyalgia Survey Score, as previously described. 15 , 16 Mean plots with 95% Confidence Intervals (CI) were generated to demonstrate daily mean opioid consumption and daily mean “average pain” stratified by Fibromyalgia Score tertile in the 2 weeks after hysterectomy. Multivariate linear regression models were fit to estimate 2-week total opioid consumption. All preoperative covariates were examined for collinearity and considered as possible covariates in the multivariate modeling. We also examined the relationship between preoperative Fibromyalgia Score and daily postoperative pain and opioid use. Longitudinal multivariate analyses of the patterns of pain and opioid consumption during the 2 weeks after hysterectomy were done using linear mixed regression models with normal random intercept term shared by measurements on the same subject. Best models for all analyses were obtained using a backward selection procedure, repeated application of the likelihood ratio test, with results confirmed using Akaike Information Criteria (AIC, BIC). In the longitudinal models, time was represented as a categorical variable to allow enough flexibility to describe the complex dynamics of the pain and opioid responses. Differences between groups based on Fibromyalgia Score were expressed as the main effect of group and the group by time interaction. Plots of residuals were used to confirm no evidence of violations of model assumptions.