Fibromyalgia Survey Criteria Are Associated with Increased Postoperative Opioid Consumption in Women Undergoing Hysterectomy

POSTOPERATIVE pain after hysterectomy is a major issue contributing to patient discomfort, decreased patient satisfaction, risk for postoperative morbidity and mortality, and increasing medical costs due to extended hospital stays.^1,2  It has been estimated that there are 433,621 hysterectomies performed each year in the United States,³  of which 12 to 17% are done for chronic pelvic pain.^4,5  With a large number of cases performed annually and the possibility of same-day discharge for some patients undergoing hysterectomy, there are economic and patient-safety implications in determining ways to avoid postoperative pain and any factors that lengthen hospital stays. It has been shown in a multicentered, retrospective study that factors predictive of severe acute postoperative pain include younger age, female sex, increased body mass index, preoperative surgical site pain severity, preoperative use of opioids, and general anesthesia.⁶  Cognitive factors and pain coping strategies, such as pain catastrophizing, and mood disorders such as depression and anxiety have also been shown to influence acute pain in patients undergoing hysterectomy.² 

Fibromyalgia is a commonly studied pain disorder, which is associated with aberrant central pain processing.^7–9  Fibromyalgia is associated with higher levels of central nervous system neurotransmitters that facilitate pain and lower levels of those that down-regulate pain.^8,10–12  In addition, fibromyalgia patients have higher endogenous opioids levels with lower opioid receptor availability, which could hypothetically make them less responsive to opioids.^13,14  In 2011, the fibromyalgia survey criteria were described for use in lieu of the 1990 criteria, which included a tender point examination.^15,16  Although not intended to be used for diagnosing fibromyalgia without a clinical evaluation, it has been suggested that these criteria can be used in an epidemiological manner.¹⁷  Our group previously demonstrated that the survey criteria for fibromyalgia independently predicted increased opioid consumption after total hip and knee arthroplasty, even when controlling for other predictors such as demographics, preoperative pain intensity, mood, anxiety, catastrophizing, anesthesia type, and surgery.¹⁸ 

Whereas higher fibromyalgia survey scores were predictive of increased opioid consumption in arthroplasty,¹⁸  it is not known whether a fibromyalgia-like phenotype would also predict increased opioid consumption in other surgical cohorts, such as patients undergoing hysterectomy. It is important to demonstrate the generalizability of this finding among a separate surgical cohort to increase acceptance of the preoperative fibromyalgia score in clinical decision making. The objective of this prospective, observational cohort study was to assess the impact of the fibromyalgia survey score on acute pain outcomes in subjects undergoing hysterectomy. We hypothesized that higher fibromyalgia survey scores would predict higher postoperative opioid consumption after hysterectomy. Length of stay was studied as a secondary outcome. These data could help to establish preoperative algorithms that personalize pain medicine to the individual patient phenotype rather than using the current care patterns that are dictated primarily by the surgical condition and not tailored based on patient characteristics.

Institutional review board (University of Michigan, Ann Arbor, Michigan) approval was obtained. Adult patients (≥18 yr old) scheduled for hysterectomy for a benign indication between October 2011 and December 2013 were prospectively recruited on the day of surgery in the preoperative waiting area. No power analysis was performed when the study was planned. All patients recruited during the period of the study were considered for inclusion subject to the exclusion criteria and availability of the measurements required for the analysis. Exclusion criteria included inability to provide written informed consent, non-English speakers, and prisoners. Written informed consent was obtained from all participants. Study reporting conforms to the STrengthening the Reporting of OBservational studies in Epidemiology Statement.¹⁹ 

Patients completed preoperative phenotyping using validated self-report measures. The primary measure of interest was the 2011 fibromyalgia survey criteria.¹⁷  The fibromyalgia survey is composed of the number of painful body areas assessed by using the Michigan Body Map (0 to 19) and comorbid symptoms such as fatigue, trouble thinking, and headaches assessed by using the Symptom Severity Index (0 to 12). Hence, the total fibromyalgia score ranges from 0 to 31. The measure has demonstrated good reliability, convergent validity, and discriminant validity.²⁰  Previous cut points have been described to categorize a person as “fibromyalgia positive”; however, the continuous score was used for the analyses in the current study. Additional phenotyping included pain severity (Brief Pain Inventory; 0 to 10 composite score using the mean of the average, least, worst, and pain right now; overall body pain and surgical site pain assessed separately)^21,22 ; neuropathic pain descriptors (PainDETECT),²³  anxiety and depression (Hospital Anxiety and Depression Scale),²⁴  and catastrophizing (Coping Strategies Questionnaire).^25,26  The medication list from the preoperative records was printed and reviewed with the patient by a research assistant. All medications administered as needed were reviewed in detail to ensure the ability to differentiate the varied patterns of opioid use (e.g., differentiating patient who uses one hydrocodone per week from the person who takes three daily). The average daily opioid consumption was converted to a 24-h oral morphine equivalent (OME) total.^27,28  Opioids administered by the anesthesia team before and during surgery were also converted to OMEs as a covariate. Demographics, body mass index, and American Society of Anesthesiologists (ASA) physical status score were collected from the electronic medical record (Centricity; General Electric Healthcare, USA). Because all patients received general anesthesia, the anesthetic technique was not assessed. The surgical approach was recorded as vaginal, laparoscopic, robotic, or open.

The total postoperative opioid consumption (primary outcome) was obtained from the institutional electronic order entry system (Carelink^*) and nursing records throughout time in the postanesthesia care unit and the inpatient course, as previously described by Brummett et al.¹⁸  Opioid use in the postanesthesia care unit was assessed separately and included in the total postoperative opioid consumption (combined with opioids administered after the postanesthesia care unit). All opioid data (preoperative, intraoperative, and postoperative) were converted to morphine equivalents (OME) for analyses. The length of stay was included as a covariate for postoperative opioid consumption to account for differences in opioids that were merely due to longer inpatient stays.

The cohort was also prospectively followed longitudinally for 6 months for chronic pain outcomes, which are not included in the current article. The opioid consumption data represent the primary outcome analysis for the acute pain portion of this prospective, observational cohort study.

Data were entered into the Assessment of Pain Outcomes Longitudinal Electronic Data Capture system.²⁹  Missing data for the validated instruments were handled as recommended by instrument authors,^23,30,31  and further described in a previous publication.¹⁸  Other missing data were handled as described by the authors of each individual instrument.^23,30,31  For the Brief Pain Inventory and PainDETECT tools, if patients were missing more than one item, the subject was excluded. For the Hospital Anxiety and Depression Scale, anxiety and depressive symptom scores, a single value for the missing item was inferred by imputation of the mean if six of the seven values were present. For all other tools, one missing question was allowed. Patients were not excluded for having one incomplete questionnaire.

Data were analyzed by using R 3.1.1 and SPSS (version 19; SPSS Inc., USA).

The cohort was divided into tertiles based on the estimates of 1/3 and 2/3 percentiles of the distribution of the fibromyalgia survey score for descriptive data and univariate outcomes analyses. Mean and SDs are presented for preoperative covariates. Between-groups comparisons were based on univariate models specific to the scale of the variable being considered (linear regression with continuous data, logistic with binary data). Histograms and Q–Q plots were examined to assess normality. Nonparametric analysis of the overall differences by group was performed by using the Kruskal–Wallis test. Bonferroni adjustment for multicomparisons was used in reporting the statistical significance of test results.

Total postoperative opioid consumption throughout the entire postoperative course was converted into OMEs for outcomes analyses. Multivariate linear regression models were used to analyze postoperative opioid consumption. All preoperative covariates, as well as intraoperative opioid administration and length of stay, were included as covariates in the multivariate modeling. Model-based hypotheses testing and a search for the best parsimonious model (variable selection) were performed using likelihood ratio tests and the Bayesian Information Criterion (BIC). BIC is preferred in the theory-based approach as it provides consistent estimates of the true model. Furthermore, a 10-fold cross-validation procedure was conducted to provide an assessment of model’s performance in future observations and predict shrinkage effects when the model is applied out-of-sample. The variable selection procedure was included in the cross-validation loop and was done anew for each training subset of the data. Models were conducted with missing data on relevant data excluded. Backward search for the best model was conducted starting from all variables included in the initial full model and supervised by medical experts within the limits of uncertainty allowed by the technical procedure. Interactions were tested by the likelihood ratio tests. Two-sided tests and 0.05 significance levels were used throughout.

A total of 208 patients of the 307 patients approached agreed to participate (68% participation rate). In analysis of participants versus nonparticipants, there were significant differences for age (participants 47.6 yr vs. nonparticipants 50.4 yr; P = 0.033) and race (84.6 vs. 66.7% Caucasian; P < 0.0001). One patient withdrew on the day of surgery, one patient was excluded from analysis due to additional surgeries during the same admission, and three patients were excluded because they did not undergo the planned hysterectomy and instead underwent a different procedure as outlined in the flowchart (fig. 1). Regarding missing data, eight patients of the 203 analyzed were missing data for the fibromyalgia survey score and were not included in the analysis. Data from 195 patients were analyzed.

A histogram of fibromyalgia scores is shown in figure 2. The mean fibromyalgia score for the entire cohort was 6.5 (±4.4) with a range of 0 to 24. A total of 17 patients (8.2%) met the previously described cut points of being termed “fibromyalgia positive.”¹⁷  The subjects were divided into tertiles based on fibromyalgia survey score (low, 0 to 4, n = 68; moderate, 5 to 7, n = 59; and high, 8 to 24, n = 68). As shown in table 1, there were no significant differences in age, race, body mass index, surgical approach, or ASA physical status score between fibromyalgia score tertiles. In addition, there were differences in surgical site body pain, overall body pain, depressive symptoms, anxiety, and catastrophizing between all groups (P < 0.0001 for all listed). Patients in the high fibromyalgia group were taking more daily opioids preoperatively and were on a higher average opioid dose (OME) when compared with the low and moderate groups (P < 0.01 for each comparison).

There was an increase in postoperative opioid consumption between low versus high fibromyalgia score groups (P = 0.013) as seen in table 2. Figure 3 demonstrates the scatterplot of total postoperative opioid consumption and the fibromyalgia survey score. Length of stay was not significantly different between groups (table 2).

The primary outcome, postoperative opioid consumption, was positively associated with the continuous fibromyalgia survey score in the multivariate linear regression model (P < 0.0001). Multivariate analysis also showed that higher ASA physical status score, catastrophizing, laparotomy surgical approach, and preoperative opioid consumption were all predictive of increased postoperative opioid consumption (table 3). Backward variable selection using likelihood ratio tests and the BIC-based model selection procedure resulted in one and the same best model with BIC = 2,231 presented in table 3. Over the set of cross-validation runs, the best variables presented in table 3 were selected 80 to 100% of the time, whereas selection frequency of other variables did not exceed 10%. Shrinkage from the raw R² of 0.53 based on the best model to the cross-validated R² of 0.35 was observed. The linear regressions with the log-transformed response were also considered. However, there was no improvement in fit as judged by the plots of residuals. The results of modeling without the transformation were confirmed, and the fibromyalgia survey score still was statistically significantly correlated with postoperative opioid use (data not shown).

This study demonstrated that increased fibromyalgia survey scores are predictive of increased postoperative opioid use after hysterectomy. These data support the findings by Brummett et al.,¹⁸  which showed this survey score as independently associated with increased postoperative opioid consumption in total hip and knee arthroplasty cohorts, and support the fact that the fibromyalgia survey score is more broadly useful in predicting opioid consumption after surgery. Each 1-point increase on the 31-point fibromyalgia survey score was associated with an adjusted increase of almost 7-mg oral morphine in the postoperative inpatient period (table 3), which is close to the estimated 9 mg OME increase in the total knee and hip arthroplasty cohort.

The use of the fibromyalgia survey criteria in a surgical cohort represents a relatively novel approach to predicting postoperative pain and opioid consumption. It is important to stress that this is not a cohort of fibromyalgia patients. Instead, the current study assessed the spectrum of signs and symptoms associated with fibromyalgia as a means to place patients on a continuum of sensitivity. Based on that which is known about fibromyalgia,⁹  our group believes that patients higher on this measure represent a population with pain that is more “centralized” in nature (e.g., altered central nervous system pain processing). Not surprisingly, the fibromyalgia survey scores in this cohort are much lower than a pure fibromyalgia cohort. In the 2011 survey criteria for fibromyalgia, patients were classified as “fibromyalgia positive” if their scores were widespread body pain 7 or greater (score ranges 0 to 19, body map assessment) and symptom severity index 5 or greater (score range 0 to 12) or widespread pain index of 3 to 6 and symptom severity index 9 or greater.¹⁷  More simply, a total score of 13 or greater has also been termed positive for fibromyalgia. The validity of the survey criteria in a fibromyalgia cohort when compared with the 1990 American College of Rheumatology criteria (which included the tender point examination) has been established.^32,33  Preoperative diagnoses of fibromyalgia were not assessed, as the diagnosis can be influenced by patients’ access to health care and the treating physician. Whether the strength of these relations would be stronger in a “pure” fibromyalgia cohort is challenging to estimate.

Whereas only 8.2% of this cohort met these criteria of being termed “fibromyalgia positive,” the associations between the increased fibromyalgia score and opioid consumption were demonstrated across the continuum of scores in the multivariate modeling (table 3). These data further underscore the importance of assessing patients who are sometimes termed “subclinical” or “subthreshold,” as the defined cut points fail to account for the middle of the distribution. Our group is interested in larger studies of a broader group of surgical cohorts to try to consider thresholds above which patients tend to require more opioids or have differential perioperative outcomes. In clinical care, it is more likely that higher scores on the measure would trigger deeper phenotyping and assessment to consider appropriateness of alternative acute pain treatments, whereas patients lower on the measure could get a more typical perioperative algorithm.³⁴ 

It has been shown that patients with fibromyalgia have lower μ-receptor availability,¹⁴  and higher endogenous opioid levels (cerebrospinal fluid levels),¹³  compared with those without fibromyalgia. A study by Bruehl et al.³⁵  found that individuals with preexisting high endogenous opioid function had a smaller benefit from morphine administration than those with lower opioid function, with individual differences in endogenous opioids accounting for 29% of the variance in responsiveness to opioids in the acute setting. The degree of endogenous opioid function was determined by finding the difference between pain values after administration of placebo and administration of naloxone, an opioid receptor antagonist.³⁶  This same group investigated chronic pain as well and found that those with lower natural opioid levels had better acute relief from morphine in the treatment of chronic back pain than those patients found to have higher endogenous opioid levels.³⁵ 

These physiologic findings may have great clinical significance in guiding how clinicians treat pain in this population of patients. It has long been believed that opioids are less effective in fibromyalgia patients due to these known central nervous system physiologic abnormalities.⁹  In this study, we attempted to capture these biological differences at a phenotypic level through the fibromyalgia survey score. The identification of this patient population exhibiting fibromyalgia-like symptoms, paired with data showing that this group of patients has a decreased response to opioid medications, may allow clinicians to use a simple phenotypic battery to practice personalized pain medicine. The fibromyalgia survey score may serve as a biomarker for an individual’s endogenous opioid tone and may therefore predict their response to exogenous opioids. Future studies including imaging and prospective follow-up are needed to better elucidate the clinical applicability of the findings.

Validation of the fibromyalgia survey score in a second surgical cohort increases the clinical applicability of this scoring system in practice. The importance of demonstrating generalizability of predictive modeling was described in a recent editorial by Drs. Eisenach and Houle.³⁷  This survey is an easily administered tool that can better prepare providers for the postoperative course of a patient at the individual level allowing for a personalized care plan. It is possible that patients with higher fibromyalgia survey scores would receive substantial benefit from regional analgesia and/or adjunct pharmacological therapies (e.g., serotonin–norepinephrine reuptake inhibitors, gabapentinoids). This one-page survey would be a simple and clinically relevant tool to administer before surgery to aid in development of the anesthetic and pain management plan. Although the measure does a good job of distinguishing patients with worse preoperative phenotypes and independently predicts increased opioid consumption, there are patients higher on the measure that do well (fig. 3). It is possible that additional domains of pain and affect need to be included to increase the predictive value of the measure. Our group is currently exploring these concepts.

It is also possible that patients who are more fibromyalgia-like would benefit from adjunctive medications that go beyond the acute, inpatient period to further address subacute pain after hospital discharge. Aiding surgeons in the identification of patients prone to more severe acute pain and creating individualized care plans is an opportunity for anesthesiologists to add value in the Perioperative Surgical Home model.^†

Additional predictors of increased postoperative opioid consumption include higher ASA class, catastrophizing, laparotomy surgical approach, and preoperative opioid use. The laparotomy surgical approach being a predictor of increased opioid consumption is clinically relevant to obstetricians and gynecologists in their surgical decision making for each patient. Because this surgical approach is predictive of increased opioid use, providers should attempt to use alternative surgical approaches unless otherwise contraindicated.

This study is limited in that it is a single-center study and therefore is limited by our institutional guidelines, patient population, and surgeon preferences and techniques. Further studies with multiple centers would add generalizability of the fibromyalgia survey criteria and its clinical application. In this study, we used postoperative OME consumption as a proxy variable for postoperative pain, and we were unable to collect individual pain scores for each patient due to inconsistent recording in the patient record. Opioid consumption is not likely a perfect proxy for the severity of pain. In the future, we plan to prospectively collect pain score data.

As was previously demonstrated in a total knee and hip arthroplasty cohort,¹⁸  this study demonstrated that increased fibromyalgia survey scores were predictive of postoperative opioid consumption in the posthysterectomy surgical population during their hospital stay. This finding reiterates that the fibromyalgia survey score is an important clinical tool that can be used to detect phenotypic differences in patients with centralized pain syndromes, such as fibromyalgia, and better treat pain in the postoperative period.

The authors thank Sachin Kheterpal, M.D., M.B.A., and Tyler Tremper, B.S.E. (both from the Department of Anesthesiology, University of Michigan, Ann Arbor, Michigan), for assistance with the electronic medical record query. The authors also thank Deborah Wagner, Pharm.D., F.A.S.H.P. (Department of Anesthesiology and the School of Pharmacy, University of Michigan), for assistance in opioid conversions.

Supported by the Department of Anesthesiology, University of Michigan, Ann Arbor, Michigan. Research funding was received from Neuros Medical Inc. (Willoughby Hills, Ohio) (to Dr. Brummett) and from Eli Lilly (Indianapolis, Indiana), Merck Pharmaceuticals (Whitehouse Station, New Jersey), and Forest Pharmaceuticals (New York, New York) (to Dr. Clauw).

Dr. Clauw is a consultant for Pfizer, Inc. (New York, New York), Johnson and Johnson (New Brunswick, New Jersey), Forest Pharmaceuticals (New York, New York), Merck (Whitehouse Station, New Jersey), Nuvo Research, Inc. (Mississauga, Ontario, Canada), Eli Lilly, Inc. (Indianapolis, Indiana), Grunenthal Pharma Ltd. (Dublin, Ireland), Jazz Pharmaceuticals, Inc. (Palo Alto, California), and Purdue Pharma (Stamford, Connecticut). The other authors declare no competing interests.