Chronic Opiod Use after Hysterectomy

Specific Aims

The rapid increase in the incidence of opioid-related overdoses and deaths has become a big public health threat in the United States. This opioid epidemic affects more women than men due to higher opioid prescribing and dose, longer period use, and more opioid dependence in women.¹ From 1999 to 2010, the rate of opioid overdose caused death increased 5-fold among women.² Besides the illicit purchase for some women, the initial exposure to opioids for many others may likewise come from the regular medical treatment as prescription opioids are widely used for pain management after surgery.³ As such, the critical research gap on opioids use is that the evidence is insufficient to draw conclusions about optimal strategies for initiation and titration of opioid therapy.⁴ Several observational studies have investigated the patterns of opioid use for noncancer pain in patients pre and post-surgery, and identified the risk factors of chronic opioid use post-surgery.^5-11 Most of these studies examined the prediction of baseline demographic and clinical characteristics of patients for prolonged opioid postoperative use and suggested that surgery is a risk factor for chronic opioid postoperative use.^5-9 Only two studies examined the initial opioid exposure within 6 or 7 days of the surgery date and results are controversial.^10,11 One study reported that initial exposure of opioid poses 44% increase risk of chronic postoperative opioid use and another one suggested that this risk would be low and statistically insignificant.^10,11 Similar investigation has not been done specifically for hysterectomy, the most frequently performed non-obstetric surgeries in the United States for women of reproductive age.^3,12 Pain has been demonstrated as a common symptom before and after hysterectomy.^13,14 A small cross-sectional study has examined the predictors for opioid prescription in women of productive age and identified that opioid use was significantly associated with hysterectomy status and pain-related dysfunction.¹⁵ Another cross-sectional survey study reported that 32% patients had chronic pain after hysterectomy.¹⁴ Neither chronic opioid postoperative use nor initial opioid exposure for acute surgery pain was examined in these two studies.^14,15 Filling this gap in knowledge is critical since identifying the risk factors of chronic opioid postoperative use could lead to optimized initial opioid prescribing for acute pain management and reduced chronic opioid postoperative use and improved women health.

Our long-term goal is to help reduce chronic use of opioids and optimize the pain management in women after obstetric and gynecologic surgery through identification and dissemination the safer initial opioid prescribing for acute post-operative pain. Our objective here, which is the next step in our long-term goal, is to compare the patters of opioid use pre and post hysterectomy and determine the important risk factors that associate with chronic use of opioids in women post hysterectomy.

The national OptumInsight Clinformatics data offers an essential resource to investigate these aims. The availability of clinical diagnoses and pharmacy medical dispensing offers a significant advantage for investigating drug utilization with corresponding clinical conditions in large population.

Our team is well suited to conduct this research given extensive expertise in contemporary pharmacoepidemiology, many years of experience on opioid abuse research, prior drug utilization studies using large claims data, and clinical expertise from obstetric and gynecologic physicians.

Our specific aims are to investigate patterns and predictors of chronic opioid use in women pre and post hysterectomy with the following analyses:

• Aim 1: To characterize the patterns of chronic opioid use in women after hysterectomy.
• Aim 2: To determine the significant risk factors for chronic use of opioids in women after hysterectomy.

The first specific aim for this study is to identify patients who chronically take opioids during six months post hysterectomy, and investigate the time and geographical patterns of chronic opioid use in women after hysterectomy. The pattern of chronic opioid use pre- and post-surgery will also be compared in women with varied age, comorbidities, co-medications, as well as types and doses of initial opioid prescribing. The second specific aim is to determine the risk factors that significantly relate to chronic opioid use, and determine if characteristics of initial filled opioid prescriptions significantly associate with the chronic use of opioids after adjusting for other potential risk factors. Many pain related studies have demonstrated that long term opioid prescribing was significantly predicted by patients’ clinical characteristics and psychosocial factors, including pain conditions, psychiatric disorders, frequency of medical visits, smoking, and pain-related dysfunction.^16-18 Therefore, the adjusted covariates in this study would include both demographic and clinical characteristics.

B. Significance and Innovation

Currently the United States is experiencing an unprecedented opioid epidemic. According to the Centers for Disease Control and Prevention (CDC), opioid-related deaths increased 200% overall from 2000 to 2014.¹⁹ During 2014, about 1.9 million people had an addiction of opioids, with overall 47,055 overdose deaths, 18,243 overdose deaths in women, and 18,893 overdose deaths related to prescription pain relievers.^20,21 Based on the data reported by the CDC, women are more likely to have chronic pain, be prescribed opioids with higher dose and longer period, and hence progress to dependence.²² Although opioid abuse is a public health crisis, opioid analgesics are still the mainstay for treatment for acute pain after major surgery. In 2014, total 245 million prescriptions for opioids were dispensed from U.S. retail pharmacies.^23,24

During 1998-2010, approximately 7.4 million hysterectomies were performed, making hysterectomy one of the most frequently performed surgeries for women in the United States.²⁵ With over 60% of hysterectomies performed abdominally and up to 85% of patients experiencing moderate-to-severe pain after hysterectomy, postoperative pain management becomes very important.^26,27 Evidence suggests that intense and long-lasting postsurgical pain can increase postsurgical morbidity, delay recovery, and lead to chronic pain.²⁸ Opioids such as morphine, meperidine, and oxycodone are widely used postoperatively to reduce and manage pain in patients after hysterectomy.²⁹ Women with hysterectomy and high levels of pain-related dysfunction were almost twice as likely to have opioid prescription. More than 85% of women with hysterectomy and a high level of pain-related dysfunction were found to use opioid.¹⁵ Hormonal disturbance, hyperalgesia, and iatrogenic effects are potential adverse effects from opioid use in women after hysterectomy.^14,15 The interaction of opioid-induced adverse effects and post-hysterectomy hormonal dysregulation may intensify pain and promote continued use of opioids.¹⁵

There are no known studies that evaluate the connection between acute and chronic postsurgical opiate prescription in women experienced hysterectomy. In consideration of the current opioid epidemic it is important to understand how post-surgical pain is managed, and if this setting presents increased risk of opioid addiction among certain groups, or related to particular prescribing practices. The goal of this proposed study is to examine whether opioids prescribed in women following hysterectomy is associated with chronic use of opioids, and to evaluate which factors may predict patterns of opioid use that indicate overuse or addiction. Accordingly, this study first aims to describe the incidence pattern of chronic opioid use in women post hysterectomy. The second aim is to explore the significant risk factors and determine if the characteristics of initial opioid prescribing significantly associate with chronic opioid use after adjusting for other potential risk factors.

This study will reveal important insights regarding post-surgical pain management for a common procedure, and determine if certain patient or treatment characteristics increase the risk of chronic opioid use in this setting. We expect that this research will provide evidence for the need to improve clinical practice towards optimized pain management and reduced chronic opioid use in women after hysterectomy through identification the specific opioid, drug type (short-acting or long-acting), and doses that significantly associate with chronic opioid use in women after hysterectomy. Our study will provide sufficient evidence to draw conclusion about optimal strategies for initiation and titration of post-surgery opioid therapy, and enhance evidence-based medicine for opioid use in Rhode Island, and also the United State.

This award would enhance my capabilities on handling interdisciplinary studies and further help me to develop my own research projects and seek for external funding.

Innovation

Our proposed observational studies based on the administrative claims data will allow for investigation of time trends and geographic variation of drug use in large population and address for well-characterized clinical conditions. Our approach will employ state of the art, innovative pharmacoepidemiologic study designs and statistical models, to improve the precision of outcome definition and minimize measured and unmeasured confounding and bias in our estimation of significant predictors for chronic opioid use after hysterectomy.

The unintended outcomes from adverse drug effects make the prospective trials unethical. In this circumstance, a well-designed, retrospective observational study with sufficient sample size offers an efficient design to determine if there is an adequate signal for impropriate opioid prescribing to women post obstetrics and Gynecology surgery.

The generalizability of study results are guaranteed due to the nationwide large health plan data that the analyses are based upon.

C. Approach

Data Sources

Study data will be derived from the national OptumInsight Clinformatics Data Mart^TM, a research database spanning January 01, 2010 through December 31, 2013. The Optum Research Database includes about 23 million beneficiaries from the nationwide commercial health insurer, United Healthcare. The data contains health care utilization with transactional reimbursement data from outpatient pharmacy dispensing, inpatient and outpatient services.³⁰ It links administrative enrollment data with the important medical codes including the national drug code (NDC) for pharmacy dispensing, the Current Procedural Terminology (CPT) code for medical procedure, and International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code for diagnosis. The claims data in this database have been adjudicated, ascertained, and deidentified for research purpose. The University of Rhode Island (URI) and OptumInsight Inc have approved utilization of these data. URI already licenses this database that it is freely available for faculties to use.

Study Population

This study will include adult women who receive the hysterectomy surgery between July 01 2010 and June 31 2013 and have at least 6 months of continuous prior coverage under this health insurance plan. This continuous 6-month coverage provides a baseline for defining new drug users, as well as baseline clinical characteristics, such as comorbidities and co-medications. The hysterectomy procedures will be derived from the ICD-9-CM diagnostic codes 68.3x-68.7x, and 68.9x, where 68.3x indicates a subtotal abdominal hysterectomy, 68.4x indicates a total abdominal hysterectomy, 68.5x indicates a vaginal hysterectomy, 68.6x indicates a radical abdominal hysterectomy, 68.7x indicates a radical vaginal hysterectomy, and 68.9x indicates other and unspecified hysterectomy.^13,31 Patients with pelvic evisceration (ICD-9-CM: 68.8x) will be excluded. We also will exclude patients with any cancer diagnoses, including malignant neoplasm of the female genitourinary organs-cancer (ICD-9-CM: 179-184), and carcinoma in situ of female genitourinary system (ICD-9-CM: 233).

Exposure Assessment

Opioids in this study include the following medications: codeine, fentanyl, hydrocodone, hydromorphone, meperidine, methadone, morphine, oxycodone, and oxymorphone. Tramadol and propoxyphene, considered as weak opioids or already off market, will not be included.^11,15 The exposure group involves adult women who receive the hysterectomy surgery and fill more than one prescription for opioids on the day of hospital discharge or within 7 days after hysterectomy surgery, provided that they have no exposure to opioids for at least 6 months prior to hysterectomy.¹¹ The 7 days of window is based on the assumption that a filled prescription during this period would likely to treat acute postoperative pain caused by hysterectomy. A comparison group, defined as the patients with no opioids use on the day of hospital discharge or within 7 days after hysterectomy surgery, will serve as a control to make a comparison of risks for chronic use of opioids. The control patients also have no exposure to opioids for at least 6 months prior to hysterectomy. The exposure of opioids will be identified using NDC codes from pharmacy claims data.

In this study, exposure assessment includes all characteristics of the initial filled opioid prescription at the day of hospital discharge from hysterectomy or during 7 days post-hysterectomy. The different type of opioids (classified as long or short acting), number of supply days, and oral morphine equivalent daily dose (milligrams) will be assessed and analyzed.

Outcome Assessment

Since the study focuses on the risk of chronic opioid use after initial exposure to opioids following hysterectomy, patients were followed from the 8^th day after hysterectomy to the first day of outcome occurrence. The outcome, chronic use of opioids after hysterectomy, will be defined using trajectory models.¹¹ in which patients with similar patterns of medication filling during follow-up will be grouped together. The trajectory model was initially generated for the purpose of the assessment of medication adherence.^32,33

In order to classify the trajectory groups for opioid use during the defined follow-up window, we first generate 6 dichotomous variables to indicate if a study participant fills a prescription of an opioid medication during each of 6 consecutive 30 day follow-up periods.¹¹ we then model these 6 binary indicators of using opioids in each 30 day follow-up period as a longitudinal response in a logistic group-based trajectory.^34,35 With a trajectory model, we will estimate the probability of membership of patients in each group, and the probability of the certain opioid exposure over time as a smooth function of time. We will fit the model using 2 to 4 opioid exposure groups with comparison of the Bayesian Information Criterion.³⁶ The number of groups will be chosen based on the value of the Bayesian Information Criterion. In each group, a third-order polynomial (including linear, squared, and cubic terms) of time will be used to model the probability of being exposed to opioids. Patients will be assigned to different trajectory groups, which are generated from these models and have highest probability of the membership. Based on the model results, the group of patients with the highest probability of filling opioids over time will be defined as the chronic use. Other trajectory groups were classified as nonchronic users. The trajectory models provide new advanced approaches to utilize the observed data to determine distinct filling patterns of opioids in our study population during the year after hysterectomy surgery. It classifies patients into groups with similar opioid exposure patterns without relying on a priori and subjective cutoff line for the definition of chronic opioid use. The trajectory models will be conducted using SAS “Proc Traj” (SAS, version 9.4, SAS Institute Inc., Cary, NC, USA).

Patterns of Chronic Opioid Use

After chronic opioid use is defined using trajectory models, we will compare the frequency of the incidence of chronic opioid use over time from 2010 to 2013. The secular trend will be analyzed using Joinpoint regression program (National Cancer Institute, Calverton, MD) to identify the joint points and slopes. The rates of chronic opioid use in women post hysterectomy will also be compared with varied age, comorbidities, co-medications, as well as characteristics of initial opioids exposure, including opioid types and doses.

The geographic variation of chronic opioid use will be mapped and analyzed using ArcGIS 10.5 geographic information systems software (ESRI, Redlands, CA). The opioid prescription policies and socioeconomic status in different states will be compared with the geographic variation of chronic opioid use.

Predictors of Chronic Opioid Use

We will identify the predictors of chronic opioid use at baseline or the time of the initial opioid prescription. A previous study demonstrated that hysterectomy, older age, and higher levels of pain-related dysfunction were associated with opioid prescription.¹⁵ Potential predictors in this study would include age, year of surgery, smoking, illicit substance abuse, alcohol abuse, hysterectomy type, psychotropic medication use, pre-operative depression, pre-operative pain condition, primary indication for hysterectomy, co-medications, and characteristics of the initial opioid exposure. Although the benefits of minimally invasive surgical procedures have been documented,³⁷ hysterectomies performed for benign indications in 2010 in the US still consisted of 40.1% total or subtotal abdominal hysterectomies, 30.5% laparoscopic hysterectomy, 19.9% vaginal hysterectomy, and 9.5% robotic-assisted hysterectomy.^21,25 Hysterectomy type has been related to the initial opioid prescription,¹⁵ however, the results from randomized clinical trials showed that abdominal hysterectomies didn’t increase postoperative chronic pain.^38,39

The primary indication for hysterectomy is defined based on the following algorithm: (i) precancerous conditions will be assigned the primary indication if endometrial cystic hyperplasia or carcinoma in situ of female genitourinary system is listed as an indication associated with hysterectomy surgery; (ii) whichever is listed first of endometriosis, uterine prolapse, or uterine leiomyoma will be assigned the primary diagnosis associated with hysterectomy if cancer or a precancerous condition is not listed; (iii) endometrial cystic hyperplasia is combined with in situ under precancerous and that pain and bleeding were included as reported reasons for hysterectomy, and (iv) other was assigned the leading for the remaining diagnoses.¹⁹

Over 60% of patients have reported pain as pre-operative symptoms for hysterectomy.³⁹ After hysterectomy, the prevalent pain was reduced to 4.7% to 31.9% during 1-2 years after surgery, while new onset pain at follow-up was reported in 1-14.9% patients and increased pain was in 2.9-5% of patients with pre-operative pelvic pain.³⁹ Therefore, in our study, we plan to assess pre- and post-operative pain, as well as the pain conditions, including pelvic pain, back pain, fibromyalgia, and migraine or other headache syndromes using ICD-9-CM codes.

Patients with pharmacy claims for opioids or diagnoses for opioid abuse (ICD-9-CM: 304.xx) during the 6-month baseline period will be identified as opioid users prior to hysterectomy.

Preoperative depression that has been found significantly associated with postoperative chronic pain,⁴⁰ will be identified using diagnoses, as well as medication use at baseline. Antipsychotropic medication use will be obtained using prescription information. The psychiatric comorbidity, will be assessed using prescription information instead of ICD-9-CM codes for greater specificity.¹¹ Co-medications refer to the use of other medications on postoperative pain management. The postoperative use of Nonsteroidal anti-inflammatory drugs (NSAIDs) or Acetaminophen with opioids has been found safe and effective on pain relieve with reduced opioid dose.⁴¹

All potential predictors, including characteristics of initial opioid use, will be adjusted in multivariate analyses to identify the significant predictors for chronic use of opioids in patients post hysterectomy.

Statistical Analysis

Statistical tests and modeling will be conducted to identify potential significant risk factors for chronic use of prescribed opioids in women post-hysterectomy. Categorical variables will be examined and compared using Chi square or the Fisher exact test. Continuous variables will be compared using the student t test. The covariates will be selected from all demographic and clinical variables using statistical modeling. A univariate analysis will be performed with the initial opioid exposure variable and then with each covariate added one at a time to monitor the change of the estimate of the major exposure variable. The covariates that modify the estimate of opioid exposure (variable >5%) or significantly predict the chronic use of opioids will be selected for further multivariate analysis. To account for many confounding factors, we will also utilize propensity score matching method to combine all confounding factors into a propensity score. The patients with initial opioid exposure will be 1:5 matched with the patients without initial opioid exposure using the propensity score with a caliper of 0.01. A multivariate conditional logistic regression model will be utilized to measure the adjusted odds ratio of initial opioid exposure on chronic opioid use in two matched groups.

A probability of type 1 error (alpha) = 0.05 will be considered to be the threshold of statistical significance. Tukey-Kramer method will be applied to correct the inflated p value due to multiple comparisons. Statistical analysis will be performed using SAS software (version 9.4, Cary, NC).

Sensitivity Studies

In order to address unmeasured confounding factors, sensitivity analyses will be conducted in two sections: subgroup analyses and instrumental variables.

Subgroup analysis

In order to investigate the effects of initial opioid exposure in women post hysterectomy, multivariate analysis will be conducted in the following subgroups: women with postoperative chronic pain, women without postoperative chronic pain, women with postoperative NSAIDs use, women without postoperative NSAIDs use, women with postoperative Acetaminophen use, and women without postoperative Acetaminophen use. Interactions between the drug exposure and the important factors will be examined in the multivariate analyses. Subgroup analyses will likewise be conducted if the interaction is statistically significant.

Instrumental Variable 

We plan to use propensity score matching to address all measurable confounding factors and generate two balanced comparison groups. However, the unmeasurable confounding factors, like confounding by indication, may still impact the study validity. Instrumental variable is an advanced methodology that has been utilized in pharmocoepidemiological studies to control for the unmeasurable confounding factors, such as confounding by indication. In the sensitivity study, we will use the physician’s preference as an instrumental variable and reexamine the initial opioid exposure on chronic use of opioids post hysterectomy.

D. Timeline

Table. Study Timeline of the Study.

Time Period	Study Progress
Before 07/01/2017	Obtain IRB approval and Optum data use permission
07/01/2017 – 08/01/2017	Complete data cleaning, manipulating, and variable editing.
08/01/2017 – 10/01/2017	Complete analyses for demographic and clinical characteristics
10/01/2017 – 01/31/2018	Complete trajectory modeling to determine chronic opioid use. Analyze the time and geographic trends.
02/01/2018 – 02/28/2018	Submit an abstract to the annual meeting of International Society of Pharmacoepidemiology (ISPE)
03/01/2018 – 06/30/2018	Complete analyses for specific aim 1
07/01/2018 – 01/31/2019	Complete predictive modeling as described in specific aim 2
02/01/2019 – 02/28/2019	Submit an R21 to NIH, and an abstract to the annual meeting of ISPE
03/01/2019 – 06/30/2019	Compete analyses for specific aim 2, including all sensitivity studies. Submit a final paper to a high impact journal
09/01/2017 – 12/15/2017	Complete course PHP 2090
01/01/2018 – 05/15/2018	Complete course PHP 2470