Cancer survivorship and its association with perioperative opioid use for minor non-cancer surgery

Abstract

Purpose:

Reducing high-risk prescription opioid use after surgery has become a key strategy in mitigating the opioid crisis. Yet, despite their vulnerabilities, we know little about how cancer survivors use opioids for non-cancer perioperative pain compared to those with no history of cancer. The purpose was to examine the association of cancer survivorship with the likelihood of receiving perioperative opioid therapy for non-cancer minor surgery.

Methods:

Using 2007–2014 SEER-Medicare data for breast, colorectal, prostate, and non-cancer populations, we conducted retrospective cohort study of opioid-naïve Medicare beneficiaries who underwent one of 6 common minor non-cancer surgeries. Modified Poisson regression estimated the relative risk of receiving a perioperative opioid prescription associated with cancer survivorship compared to no history of cancer. Stabilized inverse probability of treatment weights were used to balance measurable covariates between cohorts.

Results:

We included 1,486 opioid-naïve older adult cancer survivors and 3,682 opioid-naïve non-cancer controls. Cancer survivorship was associated with a 5% lower risk of receiving a perioperative opioid prescription (95% confidence interval: 0.89, 1.00; p=0.06) compared to no history of cancer. Cancer survivorship was not associated with the extent of perioperative opioid exposure.

Conclusion:

Cancer survivors were slightly less likely to receive opioid therapy for non-cancer perioperative pain than those without a history of cancer. It is unclear if this reflects a reduced risk of opioid-related harms for cancer survivors or avoidance of appropriate perioperative pain therapy. Further examination of cancer survivors’ experiences with and attitudes about opioids may inform improvements to non-cancer pain management for cancer survivors.

INTRODUCTION

There is increasing concern amid the opioid crisis in America about healthcare providers inadvertently contributing to opioid misuse, addiction, and overdose in patients newly prescribed opioid therapy. Recent studies identify a strong association between new opioid exposure after surgery and the development of long-term opioid use.[1–6] Perioperative opioid prescribing patterns are also highly variable even within the same procedure, and overprescribing opioids for perioperative pain has been implicated in infusing large quantities of unused opioids into communities.[7–10] As such, opioid prescribing, particularly for perioperative pain, has come under intense scrutiny and spawned concerted efforts to limit unnecessary perioperative opioid exposure.

However, the evidence available to understand and improve perioperative opioid prescribing has generally excluded cancer patients and survivors because of the fundamental role opioids have retained in cancer-related pain management. The underrepresentation of cancer survivors, in particular, overlooks a large and growing population[11] with increasing survival rates[12] and numerous risk factors for adverse opioid use outcomes.[13–19] It is necessary to understand opioid-related risks associated with cancer survivorship to optimize opioid prescribing policies in this complex population.

In 2017, Lee, et al. found that 10.4% of opioid-naïve patients undergoing curative intent cancer surgery developed new persistent opioid use.[20] Another 2017 study found a 22% higher opioid prescribing rate among long-term cancer survivors in Canada compared to non-cancer controls.[21] But no studies examined how cancer survivorship affects opioid use patterns specifically for non-cancer pain conditions, including common non-cancer minor surgical procedures. The purpose of this study was to examine the association of cancer survivorship among older adult Medicare beneficiaries with the likelihood of receiving perioperative opioid therapy from a non-cancer minor surgery. Findings from this study are necessary to inform targeted improvements to perioperative pain management for this large, high-risk patient population.

METHODS

Data

We conducted a retrospective cohort study using Surveillance, Epidemiology, and End Result (SEER) data linked with Medicare claims data (SEER-Medicare) for years 2007–2014.[22] SEER-Medicare contains detailed clinical and diagnostic data for cancer cases among Medicare beneficiaries residing in one of 19 SEER tumor registry regions across the country. SEER data is linked with fee-for-service administrative Medicare data, which contains Medicare enrollment and medical and Part D prescription claims data for individuals in the SEER tumor registry data. For this study, we used SEER-Medicare data for breast, colorectal, and prostate cancer cases and a 5% random sample of Medicare beneficiaries residing in a SEER region with no history of cancer as controls.

Cohort

We included older adult Medicare beneficiaries in the SEER-Medicare breast, colorectal, prostate, and non-cancer control populations who underwent one of the following minor surgical procedures between 7/1/2007 and 12/17/2014 at age 66 or older: laparoscopic cholecystectomy, laparoscopic appendectomy, hemorrhoidectomy, parathyroidectomy, thyroidectomy, or varicose vein removal (Supplemental Table 1). These procedures were selected because they were unlikely to be related to a subject’s cancer history or an existing pain condition, but were likely to have some variation in perioperative opioid prescribing because of their low invasiveness.

For subjects in the cancer survivor cohort, we required their breast, colorectal, or prostate cancer diagnosis to be their first and only cancer diagnosis and to have occurred at least 2 years prior to their index surgery date to ensure completion of active cancer treatment. Because our data spanned 2007 to 2014, the maximum difference between a patient’s cancer diagnosis date and minor non-cancer surgery date was 7 years. All subjects were required to have continuous enrollment in Medicare Parts A, B, and D from 6 months before through 14 days after the index surgery date to enable observation of baseline measures and the perioperative opioid use outcome. We required subjects to be opioid-naïve prior to the perioperative period using a washout period of 6 months to 30 days prior to the index surgery date. This helped ensure that any opioid use observed during the perioperative period was indicated for surgical pain and not influenced by pre-existing pain conditions or opioid use patterns. We also excluded patients who underwent any other surgical procedure during the six-month baseline period. Subjects could contribute multiple index dates for minor surgeries of interest so long as each procedure was more than 194 days apart (180 days for baseline observation + 14 days of post-surgery perioperative observation). There were 27 cases and 83 controls who underwent more than one surgery type. Finally, we excluded any patients who received hospice care during the six months prior to their minor procedure date.

Measures

The primary outcome measure was a binary indicator of a perioperative prescription opioid fill, defined as the presence of a Medicare Part D claim for a prescription opioid from 30 days prior through 14 days following the minor surgery index date.[1] We observed perioperative prescription fills, because clinicians often will provide patients with an opioid prescription for surgical pain shortly before their scheduled surgery with instructions to initiate the prescription after discharge.[23] Secondary outcome measures included total days’ supply of prescription opioids received during the perioperative period, as well as average prescribed daily dose of perioperative prescription opioid fills, measured as oral milligram morphine equivalents (MME).[24]

The primary independent variable of cancer survivorship was defined as a binary indicator of a historical SEER-confirmed case of either breast, colorectal, or prostate cancer 2 or more years before their index minor surgery. We also assessed multiple patient-level covariates including age at index surgery, sex, race, 6-month Charlson Comorbidity Index score (excluding malignancies)[25], baseline anxiety and depression diagnoses[26], SEER region, calendar year of index surgery, index surgery type, and measures of baseline healthcare utilization, including number of outpatient visits, emergency department (ED) visits, and inpatient days during the 6-month baseline period.

Propensity weighting

To address confounding in the association between cancer survivorship and perioperative opioid use for non-cancer minor surgery, we applied stabilized inverse probability of treatment weights (IPTW) in our analyses.[27] IPTW is a method of employing propensity scores to balance measurable confounders between the cancer survivor and non-cancer control groups and approximate random assignment to a prior diagnosis of either breast, colorectal, or prostate cancer versus no cancer. We estimated the propensity of a prior breast, colorectal, or prostate cancer diagnosis for each cancer survivor and non-cancer control in our cohort using a logistic regression model regressing a history of cancer on all patient-level covariates described previously. Data for each cancer survivor was weighted by the inverse of their propensity score, while non-cancer controls were weighted by the inverse of one minus their propensity score. To improve the precision of estimated associations of historical cancer on study outcomes in our analyses, we stabilized the propensity weights by multiplying the IPTW by the probability of having a prior cancer diagnosis for subjects in the cancer survivor cohort, and by the probability of having no history of cancer for subjects in the non-cancer control cohort. The stabilized IPTW approach estimates precise population average treatment effects that are generalizable to the entire study population.

Analysis

We described study cohort characteristics and reported standardized mean differences to assess the balance of measured cohort characteristics after propensity weighting.[28] Weighted bivariate statistics were used to compare study outcomes between the cancer survivor and control cohorts, both overall and by minor surgery type.

For the primary analysis, we conducted a modified Poisson regression model[29], using the stabilized IPTW, to estimate the relative risk of a perioperative prescription opioid fill associated with cancer survivorship status. We controlled for previously-described patient-level covariates in the model. The inclusion of these covariates in both the propensity model and the modified Poisson model provided a doubly-robust estimate, which requires only either the propensity score model or the final weighted analytic model to be correctly specified to obtain an unbiased estimator.[30] We also conducted the primary analytic model stratified by cancer type. Stabilized IPTWs were recalculated for each individual cancer subset combined with the full non-cancer control cohort. These findings are presented in the supplemental materials.

For each of the two secondary outcome measures—total opioid days supplied in the perioperative period and mean daily MME of perioperative opioid prescriptions filled—we performed weighted linear regression models on the log-transformed outcome to estimate the percent change in the outcome associated with cancer survivorship status, controlling for patient covariates, among subjects with any perioperative opioid fill. All analyses were conducted in SAS 9.4 (SAS Institute, Cary NC) and Stata 15 (StataCorp LLC, College Station TX). This research was approved by the Institutional Review Board of the primary author’s home institution.

RESULTS

The study cohort included 1,486 opioid-naïve older adult cancer survivors [breast (n=574); prostate (n=677); colorectal (n=235)] who were a mean of 3.5 years post-diagnosis, and 3,682 opioid-naïve non-cancer controls [Supplemental Figures 1 and 2]. Table 1 presents bivariate comparisons of unweighted and weighted cohort characteristics between the cancer and non-cancer control groups. All standardized mean differences of measured cohort characteristics between the study groups after propensity weighting with stabilized IPTW were <0.1, indicating successful balance of measurable characteristics between study groups.[28]

Table 1:

Cohort characteristics before and after propensity score weighting

	Unweighted			Weighted
	Cancer survivors (n=1486)	Non-cancer controls (n=3682)	Std. Diff.	Cancer survivors (n=1486)	Non-cancer controls (n=3682)	Std. Diff.
Age, mean	74.0	74.3	−0.045	74.1	74.2	−0.018
Female, %	48%	65%	−0.351	58%	60%	−0.042
White, %	86%	82%	0.098	84%	83%	0.011
Charlson score, %
0	46%	44%	0.050	44%	45%	−0.016
1	28%	27%	0.022	28%	27%	0.030
≥2	26%	30%	−0.077	28%	29%	−0.012
Depression, %	8%	8%	0.008	8%	8%	0.010
Anxiety, %	7%	6%	0.022	7%	7%	0.009
Baseline utilization, mean
Outpatient visits	2.5	2.0	0.172	2.3	2.4	−0.039
ED visits	0.2	0.2	−0.028	0.2	0.2	−0.016
Inpatient days	0.2	0.6	−0.058	0.3	0.5	−0.031
Minor surgery type, %
Laparoscopic cholecystectomy	50%	56%	−0.120	52%	54%	−0.041
Appendectomy	8%	8%	0.009	9%	8%	0.006
Parathyroidectomy	7%	6%	0.027	7%	6%	0.014
Hemorrhoidectomy	30%	26%	0.106	28%	27%	0.033
Varicose vein removal	2%	2%	0.001	1%	2%	−0.011
Thyroidectomy	3%	3%	0.023	4%	3%	0.021
Cancer type, % of cancer cohort
Breast	39%			46%
Colorectal	16%			18%
Prostate	46%			37%
Years since cancer diagnosis, mean	3.5			3.2

Note: Weighted analyses used stabilized inverse probability of treatment weights. The propensity score model include all characteristics reported in the table, plus SEER registry region and year of index minor surgery. A standardized mean difference with an absolute value less than 0.1 was considered to represent covariate balance between study groups.

Std. Diff = Standardized mean difference; ED=Emergency department

The study cohort had an average age of 74 years and was 60% female and over 80% White [Table 1]. Almost 30% had a Charlson comorbidity score of ≥2 during the 6-month baseline period before their index minor surgery, but the prevalence of pre-existing depression (8%) and anxiety disorder (7%) was low. Baseline healthcare utilization was comparable between groups. The most common index minor non-cancer surgery was laparoscopic cholecystectomy (52%–54%), followed by hemorrhoidectomy (27%–28%), laparoscopic appendectomy (8%–9%), parathyroidectomy (6%–7%), thyroidectomy (3%–4%), and varicose vein removal (1%–2%).

Across all minor surgery types, 53% of cancer survivors filled a perioperative opioid prescription compared to 57% of non-cancer controls (p=0.03) [Table 2]. There was no difference between cancer survivors and non-cancer controls in the prevalence of a perioperative opioid fill for those undergoing laparoscopic cholecystectomy, thyroidectomy, and parathyroidectomy. Cancer survivors had a lower prevalence of perioperative opioid fills compared to non-cancer controls in the context of hemorrhoidectomy [cancer survivors 23%; non-cancer controls 31% (p<0.01)], laparoscopic appendectomy [cancer survivors 49%; non-cancer controls 61% (p=0.07)], and varicose vein removal [cancer survivors 24%; non-cancer controls 49% (p=0.05)]. The weighted mean days’ supply of prescription opioids received in the perioperative period, as well as the mean prescribed opioid dose, were comparable between cancer survivors and non-cancer controls.

Table 2:

Weighted bivariate comparisons of non-cancer perioperative opioid use outcomes between cancer survivors and non-cancer controls

	Total (n=5168)	Cancer survivors (n=1486)	Non-cancer controls (n=3682)	p-value
Any perioperative opioid fill, (%)	56%	53%	57%	0.03
Any perioperative opioid fill by minor surgery type, (%)
Laparoscopic cholecystectomy (n=2811)	69%	70%	69%	0.63
Appendectomy (n=405)	58%	49%	61%	0.07
Parathyroidectomy (n=327)	52%	51%	53%	0.72
Hemorrhoidectomy (n=1397)	29%	23%	31%	<0.01
Varicose vein removal (n=83)	43%	24%	49%	0.05
Thyroidectomy (n=159)	67%	69%	66%	0.71
Among those with a perioperative opioid fill (n=2868)
Perioperative opioid days supplied, mean (SD)	6.5 (0.1)	6.5 (0.2)	6.6 (0.1)	0.68
Average daily MME of perioperative opioid fills, mean (SD)	45.6 (0.6)	44.6 (1.6)	46.0 (0.6)	0.42

Note: Proportions are weighted by stabilized inverse probability of treatment weights. The denominator for reported proportions for each minor surgery type is the total number of subjects receiving that specific minor surgery. A total of 27 cases and 83 controls were included in multiple surgery type subsets.

SD=Standard deviation; MME=Milligram morphine equivalents

In the primary analysis reported in Table 3, older adult cancer survivors had an adjusted 5% lower risk of a perioperative opioid fill for a minor non-cancer surgery, compared to older adults without a history of cancer [Risk ratio (RR)=0.95; 95% confidence interval (CI) (0.89, 1.00); p=0.06]. Decreasing age, in years [RR=0.98; 95% CI (0.98, 0.99); p<0.01] and higher baseline comorbidity burden [RR=0.94; 95% CI (0.89, 1.00); p<0.06] were also associated with a lower risk of perioperative opioid fills. Women had a 6% higher risk of perioperative opioid fills compared to men [RR=1.06; 95% CI (1.01, 1.12); p=0.02]. The risk of perioperative opioid fills also increased as baseline outpatient care utilization increased [RR=1.01; 95% CI (1.01, 1.02); p<0.01].

Table 3:

Modified Poisson regression estimates of the adjusted relative risk of a perioperative opioid prescription fill for a non-cancer minor surgery associated with cancer survivorship

	Risk Ratio	95% CI	p-value
Cancer survivorship	0.95	(0.89, 1.00)	0.06
Age in years	0.98	(0.98, 0.99)	<0.01
Female	1.06	(1.01, 1.12)	0.02
Race
White	1.00	(0.92, 1.07)	0.92
Non-White		Comparator
Charlson comorbidity score
0		Comparator
1	1.01	(0.96, 1.07)	0.69
≥2	0.94	(0.89, 1.00)	0.06
Depression	0.98	(0.89, 1.08)	0.72
Anxiety	0.98	(0.89, 1.07)	0.60
Baseline utilization, mean
Outpatient visits	1.01	(1.01, 1.02)	<0.01
ED visits	1.02	(0.98, 1.07)	0.37
Inpatient days	1.00	(0.99, 1.00)	0.09
Minor surgery type, %
Laparoscopic cholecystectomy		Comparator
Appendectomy	0.83	(0.76, 0.91)	<0.01
Parathyroidectomy	0.75	(0.67, 0.84)	<0.01
Hemorrhoidectomy	0.45	(0.41, 0.49)	<0.01
Varicose vein removal	0.62	(0.47, 0.81)	<0.01
Thyroidectomy	0.94	(0.83, 1.06)	0.31

Note: n=5168 (1486 with history of cancer; 3682 noncancer controls). Adjusted risk ratio estimates were obtained by modified Poisson regression weighted with stabilized inverse probability of treatment weights calculated based on the observed characteristics reported in the table, as well as SEER registry region and year of index minor surgery; the modified Poisson model also controlled for these covariates.

CI=Confidence interval; ED=Emergency department

Among subjects who filled any perioperative opioid prescriptions for a minor non-cancer surgery, cancer survivorship was not associated with the total opioid days supplied or the mean daily MME of perioperative opioid fills [Table 4]. White subjects received an estimated 30% higher mean daily MME of perioperative opioid prescriptions compared to non-White subjects [95% CI (0.24, 0.37); p<0.01]. A high baseline comorbidity burden was associated with 12% increase in opioid days supplied [95% CI (0.06, 0.18); p<0.01], but a 10% lower mean daily MME [95% CI (−0.16, −0.04); p<0.01].

Table 4:

Weighted linear regression estimates of percent change in perioperative days supplied and average daily dose of perioperative opioids associated with cancer survivorship

	Total perioperative opioid days supplied			Mean daily MME of perioperative opioid fills
	Estimated % change	95% CI	p-value	Estimated % change	95% CI	p-value
Cancer survivorship	3.1%	(−0.02, 0.09)	0.26	0.1%	(−0.05, 0.06)	0.97
Age in years	1.6%	(0.01, 0.02)	<0.01	3.4%	(0.03, 0.04)	<0.01
Female	0.2%	(−0.05, 0.05)	0.94	3.2%	(−0.01, 0.08)	0.17
Race
White	4.0%	(−0.02, 0.10)	0.23	30.2%	(0. 24, 0.37)	<0.01
Non-White	Comparator			Comparator
Charlson comorbidity score
0	Comparator			Comparator
1	6.2%	(0.01, 0.12)	0.03	−0.8%	(−0.06, 0.05)	0.79
≥2	12.0%	(0.06, 0.18)	<0.01	−9.9%	(−0.16, −0.04)	<0.01
Depression	0.4%	(−0.09, 0.10)	0.94	−1.2%	(−0.11, 0.08)	0.81
Anxiety	−1.7%	(−0.10, 0.07)	0.70	−1.8%	(−0.10, 0.07)	0.67
Baseline utilization, mean
Outpatient visits	0.7%	(0.00, 0.01)	0.01	1.0%	(0.01, 0.01)	<0.01
ED visits	0.0%	(−0.05, 0.05)	0.99	−9.1%	(−0.13, −0.05)	<0.01
Inpatient days	−0.1%	(−0.01, 0.01)	0.76	−0.3%	(−0.01, 0.00)	0.43
Minor surgery type, %
Laparoscopic cholecystectomy	Comparator			Comparator
Appendectomy	−1.2%	(−0.10, 0.07)	0.78	9.2%	(0.01, 0.17)	0.02
Parathyroidectomy	−4.2%	(−0.14, 0.06)	0.42	−14.7%	(−0.25, −0.04)	<0.01
Hemorrhoidectomy	21.1%	(0.14, 0.28)	<0.01	1.0%	(−0.06, 0.08)	0.78
Varicose vein removal	−2.0%	(−0.24, 0.20)	0.86	−12.1%	(−0.35, 0.11)	0.30
Thyroidectomy	5.5%	(−0.06, 0.17)	0.35	0.4%	(−0.13, 0.14)	0.96

Note: n=2868 (767 with history of cancer; 2101 noncancer controls). Adjusted percent change estimates were obtained by linear regression of the log-transformed outcome measure, weighted with stabilized inverse probability of treatment weights calculated based on the observed characteristics reported in the table, as well as SEER registry region and year of index minor surgery; the models also controlled for these covariates.

CI=Confidence interval; ED=Emergency department

Supplemental table 2 in the online supplement reports primary analyses by cancer subtype. Of note, prostate cancer survivorship was associated with a 10% lower risk of non-cancer-related perioperative opioid fill compared to non-cancer controls [RR=0.90, 95% CI (0.83, 0.98); p=0.02]. Older adults with a history of breast cancer also appeared to be less likely to receive perioperative opioids [RR=0.77, 95% CI (0.57, 1.03); p=0.08]. Subjects with a history of colorectal cancer were estimated to have a non-significant 7% lower risk of perioperative opioid use than non-cancer controls.

DISCUSSION

This is the first known study to examine the association of cancer survivorship on perioperative prescription opioid use for common non-cancer minor procedures. This study builds on recent calls for more deliberate recognition of cancer survivorship status in opioid research[31] and policy[32] by providing early evidence that cancer survivorship affects downstream opioid use even for non-cancer-related pain conditions.

In our cohort of opioid-naïve older adult Medicare beneficiaries, survivors of breast, colorectal, or prostate cancer were 5% less likely to fill a perioperative opioid prescription for a non-cancer minor surgery compared to non-cancer controls. Cancer survivorship status was not associated with the dose or duration of perioperative opioid therapy. At face value, these findings may be reassuring in the context of growing national concerns about inadvertently kickstarting opioid dependence and addiction among opioid-naïve patients undergoing surgery. Given the complex history and risk factors among cancer survivors, surgeons may have been more conservative in their perioperative pain management approach for these patients. However, the lower risk of perioperative opioid fills among opioid-naïve older adult cancer survivors may also reflect a constellation of patient-driven factors. This notion is supported by a body of evidence demonstrating strong associations between patient psychological factors and their perception and management of their cancer pain.[33]

Cancer survivors may have been more likely to opt for non-opioid pain management therapies compared to patients without a history of cancer because of differences in perceived self-efficacy at managing perioperative pain. Major surgeries are a mainstay of curative treatment for breast, colorectal, and prostate cancer in older adults. As such, cancer survivors who underwent major procedures, such as mastectomy or radical prostatectomy, may be drawing on past experiences with significant perioperative pain during subsequent minor surgeries. After battling cancer, perhaps survivors view these surgeries as too minor to require opioid use. Our findings lend some credence to this idea if we consider the overall extent of perioperative opioid use for a specific procedure as a crude proxy for the intensity of its postsurgical pain. The two surgeries with the highest overall prevalence of any perioperative opioid use, laparoscopic cholecystectomy (69%) and thyroidectomy (67%), showed no bivariate difference in perioperative opioid use between cancer survivors and non-cancer controls. However, the two procedures with the lowest overall prevalence of perioperative opioid use, hemorrhoidectomy (29%) and varicose vein removal (43%), showed that cancer survivors had a significantly lower frequency of perioperative opioid use than non-cancer controls. This may indicate that cancer survivors were more willing or better equipped to manage perioperative pain without opioids as the invasiveness and postsurgical pain of the procedure lessened.

A more concerning explanation for the lower observed risk of perioperative opioid fills among cancer survivors may be that these patients were more likely to access stockpiled opioid supplies prescribed for prior surgeries. Excessive opioid prescribing is common for major surgical procedures.[8–10] Moreover, prior research has shown that patients often retain excessive supply of opioids after the initial pain indication subsides, often for the purpose of having opioids readily available to cost-effectively self-medicate potential future pain events.[34–39]

Another concerning factor could be that some cancer survivors in our cohort may have purposefully avoided appropriate perioperative opioid therapy because of past adverse experiences with or perceptions of opioid use stemming from prior cancer pain management. There is longstanding evidence that patients undergoing cancer treatment frequently report fear of addiction and other adverse physical effects of narcotic analgesics.[40–42] It is also possible that the cancer survivors in our study internalized societal stigma around opioid use and addiction during their cancer treatment. A recent pilot study of 97 patients undergoing active cancer treatment who were prescribed opioids for cancer-related pain found that more than 60% of patients reported experiencing some form of opioid-related stigma.[43] This included worrying about being perceived as drug-seeking, difficulty obtaining opioid prescriptions, uncomfortable conversations with healthcare providers, and feeling judged by providers, family, and friends. More than one-quarter of these patients altered their opioid use because of stigma.

Additional research is needed to clarify which provider- and patient-level factors explain the observed protective association of cancer survivorship on initiation of opioid therapy for non-cancer perioperative pain. Insights from future work can inform ongoing strategies to improve patient-centered perioperative pain management and outcomes while mitigating unnecessary exposure to prescription opioids in healthcare settings. For example, if opioid stockpiling behaviors are found to be prevalent among cancer patients, this can inform efforts to minimize opioid overprescribing and increase use of community-based safe opioid disposal methods. Alternatively, if cancer survivors are withholding indicated opioid therapy because of stigma or fear, targeted strategies can be developed to improve patient-provider communication about experience with and expectations for perioperative pain management.

Our study is subject to multiple limitations. First, we used Medicare Part D prescription claims data, which precluded us from observing opioid prescriptions purchased entirely out-of-pocket, instances where a provider prescribed perioperative opioid therapy but a patient did not fill it, use of any unused opioids accumulated from prior healthcare encounters, or the extent of opioid therapy administered directly to the patient during their surgical stay. As with all retrospective claims-based research, our findings may be subject to bias by unmeasured confounders, including health system and prescribing characteristics. We also cannot generalize our findings to patient populations outside of the older adult Medicare population. It is possible our perioperative opioid use measure inadvertently misclassified opioid fills obtained in the perioperative period for pain indications unrelated to their minor surgery as perioperative. However, our approach to defining perioperative opioid use was consistent with prior literature[44], and the use of an opioid washout exclusion prior to the perioperative period further ensured that any opioid use observed 30 days prior through 14 days following the index surgery was related to perioperative pain.

In conclusion, older adult cancer survivors were less likely to fill perioperative opioid prescriptions for a non-cancer minor surgery compared to those without a history of cancer. While it appears encouraging to observe a protective association of cancer survivorship on opioid initiation for this specific non-cancer pain condition, further work is needed to determine the extent to which this signifies reduced risk of opioid-related harms or under use of appropriate perioperative pain management. Regardless, these findings can aid surgeons in their careful assessment of their patients’ experiences and expectations surrounding opioid therapy when prescribing opioids perioperatively, as well as motivate the broader research and clinical communities to more deliberately incorporate cancer survivors into the opioid crisis response effort.