Opioid-related overdose and chronic use following an initial prescription of hydrocodone versus oxycodone.

BACKGROUND: Hydrocodone and oxycodone are prescribed commonly to treat pain. However, differences in risk of opioid-related adverse outcomes after an initial prescription are unknown. This study aims to determine the risk of opioid-related adverse events, defined as either chronic use or opioid overdose, following a first prescription of hydrocodone or oxycodone to opioid naïve patients.
METHODS: A retrospective analysis of multiple linked public health datasets in the state of Oregon. Adult patients ages 18 and older who a) received an initial prescription for oxycodone or hydrocodone between 2015-2017 and b) had no opioid prescriptions or opioid-related hospitalizations or emergency department visits in the year preceding the prescription were followed through the end of 2018. First-year chronic opioid use was defined as ≥6 opioid prescriptions (including index) and average ≤30 days uncovered between prescriptions. Fatal or non-fatal opioid overdose was indicated from insurance claims, hospital discharge data or vital records.
RESULTS: After index prescription, 2.8% (n = 14,458) of individuals developed chronic use and 0.3% (n = 1,480) experienced overdose. After adjustment for patient and index prescription characteristics, patients receiving oxycodone had lower odds of developing chronic use relative to patients receiving hydrocodone (adjusted odds ratio = 0.95, 95% confidence interval (CI) 0.91-1.00) but a higher risk of overdose (adjusted hazard ratio (aHR) = 1.65, 95% CI 1.45-1.87). Oxycodone monotherapy appears to greatly increase the hazard of opioid overdose (aHR 2.18, 95% CI 1.86-2.57) compared with hydrocodone with acetaminophen. Oxycodone combined with acetaminophen also shows a significant increase (aHR 1.26, 95% CI 1.06-1.50), but not to the same extent.
CONCLUSIONS: Among previously opioid-naïve patients, the risk of developing chronic use was slightly higher with hydrocodone, whereas the risk of overdose was higher after oxycodone, in combination with acetaminophen or monotherapy. With a goal of reducing overdose-related deaths, hydrocodone may be the favorable agent.

Introduction

Hydrocodone or oxycodone? When prescribing opioids for pain, prescribers in the U.S. commonly decide between these two medications. Although the total number of opioid prescriptions has declined in recent years, hydrocodone remains the most frequently prescribed opioid with 83.6 million prescriptions dispensed in 2017 [1,2]. Oxycodone is prescribed less frequently (55.2 million in 2017), but still represents roughly half of the opioids consumed in the U.S. when calculated as total morphine milligram equivalents (MMEs) [3,4]. Although national data on opioid prescriptions to opioid-naïve patients is limited, our prior study based on the Ohio prescription drug monitoring program determined that 50.8% of prescriptions to opioid-naïve patients were for hydrocodone and 25.1% were for oxycodone [5]. A national sample of 19 emergency departments (EDs) discovered that the most common opioids prescribed in that setting were oxycodone (52.3%) and hydrocodone (40.9%) [6].

Both oxycodone and hydrocodone are indicated for treatment of moderate to severe pain. For acute pain, the two drugs have similar efficacy [7-9]. Since the drugs have the same mechanism, the question remains if their risk profiles differ. Hydrocodone was once considered a safer opioid, classified as a Drug Enforcement Administration schedule III drug, similar to some formulations of codeine and other less potent opioids. However, hydrocodone was reclassified in 2014 to schedule II, the category containing oxycodone; schedule II is reserved for prescribed medications with the highest misuse potential [10,11]. Despite being in the same category, it is unclear if the two drugs are similar in their potential for misuse. Oxycodone is theorized to have greater misuse liability due to its higher likability scores and a lack of negative subjective effects when compared to hydrocodone and morphine [12]. The Ohio study determined that, compared to codeine, first-time prescriptions of hydrocodone were associated with an odds ratio of 1.33 (95% confidence interval (CI) 1.31–1.35) for developing chronic use, and those of oxycodone had an odds ratio of 1.64 (95% CI 1.61–1.67) [5]. Another study of a sample of commercially insured patients determined that probability of long-term use after a first prescription was 5.1% for short-acting hydrocodone and 4.7% for short-acting oxycodone [13]. Neither of these studies controlled for dose, duration of therapy or patient level factors. In other studies of people who misuse drugs and in opioid naïve patients, however, no differences were seen in the pharmacodynamic effects or misuse liability of either drug [14,15].

Chronic opioid use is a negative outcome given the risks of chronic opioid therapy, including development of dependence and other medical complications such as myocardial infarction and fractures [16]. Many of the risks are dose dependent, increasing with higher morphine milligram equivalent use [17]. Furthermore, tapering from stable but higher dose (≥50 MME/day) chronic opioid therapy is associated with increased risk of emergency department or hospital encounters for drug overdose or withdrawal and mental health crisis [18]. Due to these and other complications, avoidance of chronic opioid therapy for pain not related to malignancy is recommended by the Centers for Disease Control and Prevention’s 2016 opioid guidelines, although an updated version is currently forthcoming [19,20].

Potentially the most dangerous outcome after initiation of opioid therapy is overdose, either of prescribed opioids or after transition to non-prescribed opioids, such as heroin or illicitly manufactured fentanyl. In our prior work, we found that 3 out of every 1,000 opioid naïve patients prescribed an opioid had a subsequent opioid overdose, and the risks were higher for certain groups, such as the elderly and those with multiple medical comorbidities or certain psychiatric comorbidities [21]. In Oregon, the state where this study took place, at least 300 people have died from opioid-related overdose each year since 2015 [22]. Deaths associated with prescribed opioids have remained stable, but there has been a rapid increase in deaths from heroin and synthetic opioids since the start of the COVID-19 pandemic [23].

Although current rising trends are not driven by prescriptions, there is still a need to ensure safe prescription opioid practices. Because oxycodone and hydrocodone appear to be used interchangeably for acute pain, this research aims to determine if there is variability in risk-related outcomes between patients given a first prescription of oxycodone versus hydrocodone. Using a merged public health database from Oregon, we compared the rates of chronic use or opioid overdose after an opioid-naïve patient is prescribed either of these medications. Because oxycodone, but not hydrocodone, is commonly prescribed in two different formulations, we also compared risks for oxycodone prescribed alone or as a combination pill along with acetaminophen. We also evaluated the cumulative MME that patients received in the first 6 months following a first prescription and the rate of switching between different opioids after the first prescription as secondary outcomes.

Methods

Study design and data sources

This was a retrospective cohort study using the Oregon Comprehensive Opioid Risk Registry, a dataset formed by merging several previously disparate public health datasets in that state and described in detail elsewhere [24]. In brief, 2014–2018 patient claims data from the Voluntary Oregon All Payer All Claims Database (APCD) were linked to several Oregon public health datasets including the Prescription Drug Monitoring Program (PDMP), Hospital Discharge Database (HDD), and Vital Records. Using the R package “fastLink”, datasets were probabilistically linked within the Oregon Health Authority Public Health Division using first name, last name, and date of birth [25]. The terms of the data use agreement prohibit public sharing of the source data used in this study but may be obtained via request to the public health entities in Oregon that oversee the individual source datasets.

The Oregon APCD includes claims for all medical services and prescriptions paid for by participating health plans. The dataset includes 100% of the Medicaid population, 85% of the Medicare Advantage population, 81% of the fully insured commercial population, and 24% of the self-insured commercial population, in total representing approximately 80% of all insured Oregonians. In the U.S., Medicaid is a state-run safety net insurance for low-income individuals of any age, Medicare Advantage is a federal government-administered insurance program primarily for individuals ≥65 years old (as well as younger disabled people and dialysis patients) which is managed by a commercial insurance company, commercial insurance is a privately managed plan generally serving employed individuals and their families, and dual Medicaid/Medicare Advantage enrollees qualify for both plans, typically due to a combination of age, disability, medical comorbidities and low income. Patients in the Medicare Fee-for-Service population, which is the traditional public insurance plan for individuals ≥65 years old, were not included due to data availability limitations. The APCD served as the primary source for patient characteristics and opioid-related ED visits. The PDMP, HDD, and Vital Records datasets provided data on dispensed opioid prescriptions, opioid-related inpatient hospitalizations, and opioid poisoning deaths for all Oregon residents, regardless of payer.

Inclusion criteria

The sample included opioid-naïve adults who received a first prescription (termed the index prescription) of short-acting hydrocodone or short-acting oxycodone between 2015–2017. To be considered opioid-naïve, patients must not have had any opioid prescriptions, opioid-related inpatient hospitalizations, or opioid-related ED visits in the 12 months preceding the index prescription. Patients were excluded if they did not have continuous insurance enrollment for the year that they received the index prescription (gaps less than 90 days were allowed). Patients who would have otherwise met continuous enrollment criteria but died were not excluded.

Patient characteristics

Patient characteristics originated in the APCD and included age, race, ethnicity, sex, insurance type, and urbanization of the county of residence. Age was defined as of January 1, 2014 and divided into 7 categories. Because the Oregon APCD enrollment files are missing race and ethnicity for about half of patients, the Bayesian Imputed Surname Geocoding (BISG) algorithm was used to impute missing race and ethnicity when adequate information was available to do so [26]. The imputation used the following categories: White, Black, Hispanic, Asian-Pacific Islander, and other. American Indian and Alaska Native race was not imputed due to inadequate reliability for BISG estimation in this race group [27]. Imputation was implemented using the R package “wru” [28].

The type of insurance, either Medicaid, Medicare Advantage, commercial, or dual Medicaid/Medicare Advantage, was based on the insurance that the patient was enrolled in for at least 6 months of the calendar year in which they received the index prescription. The 2013 National Center for Health Statistics urban-rural classification scheme was used to classify urbanization based on patients’ county of residence for the year they received the index prescription: large central metro, large fringe metro, medium metro, small metro, micropolitan, and noncore counties [29].

Prescription characteristics

Food and Drug Administration National Drug Codes were used to identify hydrocodone, oxycodone, and other opioids in the PDMP dataset based on pharmaceutical class and drug names [30]. The PDMP data were cleaned to remove duplicate entries, those missing a prescriber ID, quantities less than 4 or greater than or equal to the 99th percentile (n = 268), and days’ supply less than zero or greater than 90 days, similar to prior work [5]. Long-acting formulations were excluded due to the small percentage of index prescriptions that were long-acting in our sample (e.g., 0.05% of hydrocodone or oxycodone index prescriptions in 2015). The total morphine milligram equivalent (MME) for the index prescription was calculated using Centers for Disease Control and Prevention conversion reference tables [31]. MME was placed into 5 categories based on the following sample percentiles: 25th, 50th, 75th, and 90th. Days’ supply, as provided to the PDMP by the pharmacy at time of prescription fill, was categorized as follows: 3 or less days’ supply, 4–6 days’ supply, 7+ days’ supply.

Primary outcomes

First-year chronic opioid use

Chronic use was defined similarly to a previously published definition [5]. In brief, the following criteria had to have been met in the 365 days after the index prescription: a) the patient must have filled six prescriptions of any opioid type, including the index prescription, and b) periods uncovered by an opioid prescription, on average, must have been 30 days or less. The uncovered period between refills was calculated by taking the difference between the predicted date of the refill based on days’ supply and the actual refill date. For the last prescription of the year, the date 365 days from the index prescription was used instead of a refill date to calculate uncovered days. The average period uncovered by an opioid prescription (criteria b) was calculated by dividing the total uncovered days by the number of gaps between refills plus one (to reflect the uncovered gap from the last prescription of the year to the end of the year).

To test our criteria, we also calculated an alternative definition of chronic use that defines long-term use as episodes exceeding 90 days with either a total days’ supply greater than or equal to 120 or 10+ prescriptions filled [32]. For this analysis, the onset of the index prescription was considered the onset of the episode, and the last date of an opioid fill in the 365 days after the index prescription was considered the end of the episode. Also, our chronic use definition did not exclude people who died within a year of the initial dispensation, so not all individuals had the same follow-up period to be classified as chronic users. Therefore, we performed an additional sensitivity analysis that excluded individuals who died in the year following their index prescription.

Opioid overdose

This outcome was the time from the index prescription to the first incident fatal or non-fatal opioid overdose from either illicit or prescription opioids. Survival days were calculated as the number of days from the index prescription to either: loss of continuous insurance enrollment, the first nonfatal or fatal opioid overdose, death from a non-opioid-related cause, or the end of the study follow-up period on December 31, 2018. To meet data use agreement restrictions limiting use of enrollment dates, continuous enrollment was determined in one-year periods, so loss of continuous enrollment at any time during a calendar year, after the index year, resulted in a censor date on December 31 of the preceding year. Fatal opioid overdoses were identified from Vital Records using ICD-10 underlying cause of death X40-X44, X60-X64, X85, Y10-Y14 with multiple cause-of-death codes (T40.0, T40.1, T40.2, T40.3, T40.4, T40.6), following guidance from the Substance Abuse and Mental Health Services Administration [33]. Literal text fields in the death record were also searched for terms associated with opioid overdose deaths using a tool developed by the Council of State and Territorial Epidemiologists Overdose Subcommittee [34]. Nonfatal opioid overdoses were derived from diagnosis codes in hospital discharge records and ED claims in the APCD (see S1 Table in the supplemental materials for the list of codes used).

Secondary outcomes

Cumulative MME

To explore whether the drug type of the initial index prescription is associated with subsequent dosages, we calculated a cumulative MME for all opioid fills in the 6 months after the index prescription, excluding the index prescription.

Drug switching

Drug switching was defined as a replacement of the index drug (oxycodone or hydrocodone) with any other opioid within the first six months after the initial index prescription was filled. We created a flag that indicated if the patient switched drugs in the follow-up period.

Statistical analyses

We computed frequencies for patient characteristics (age, gender, ethnicity/race, insurance type in the index year, urbanization in county of residence in the index year) and prescription characteristics MME category, days’ supply) for the full cohort and by index prescription drug. Standardized differences between index drugs were calculated using Cohen’s d effect size index [35]. To obtain adjusted odds ratios for first-year chronic use and drug switching, we conducted multilevel logistic regression models, including fixed effects for patient and index prescription characteristics and a random effect for prescriber. We also conducted a sensitivity analysis to examine findings using an alternative, commonly used chronic use definition. Cox proportional hazard models were used to calculate adjusted hazard ratios for the risk of overdose during the follow-up period. Computer memory limitations did not allow us to add a random effect for prescriber to the Cox models for the full sample, so we conducted a subgroup analysis on the subset of prescribers with prescriptions written to at least 200 patients. A zero-inflated negative binomial model assessed the probability of having zero cumulative MME in the first 6 months after the index prescription (logistic portion) and predictors of the 6-month cumulative MME (count portion), conditioned on being non-zero. Analyses were conducted in SAS version 9.4 (SAS Institute Inc., Cary, NC, USA).

Combination medication subgroup analysis

In our data, prescriptions of immediate-release hydrocodone were always combined with other active ingredients (e.g., acetaminophen and ibuprofen), whereas oxycodone was equally commonly prescribed either alone or in combination with other the other active ingredients. Given this difference in typical formulations for oxycodone or hydrocodone, we could not be certain whether our findings were reflective of differing opioid risk profiles or due to differences in risk for oxycodone-acetaminophen combination therapy compared to opioid monotherapy. Previous studies comparing oxycodone-acetaminophen combination therapy with oxycodone monotherapy suggest that combination therapy provides similar pain relief with fewer side effects [36,37]. Oxycodone combined with acetaminophen may have a synergistic mechanism of action leading to overall decreased doses to treat moderate or severe pain, which may reduce the risk for opioid-related adverse events such as chronic use or overdose [38]. To examine this possibility, we conducted a sub-analysis of individuals who received oxycodone monotherapy, a combination oxycodone-acetaminophen index drug, or a combination hydrocodone-acetaminophen index drug with strength-per-unit of 5 mg.

Compliance

This was a retrospective study of administrative public health datasets. After dataset linkage was performed on identifiable data, all identifiers were destroyed and subsequent data analysis was performed on strictly deidentified data. Study activities were approved by the Mass General Brigham Human Research Committee, which waived the requirement for informed consent as the research was minimal risk and could not reasonably be carried out without the waiver. The Oregon Health Authority Public Health Division ceded oversight to the approving entity.

Results

Patients

Characteristics of the 519,066 opioid-naïve patients who met the inclusion criteria are shown in Table 1 (see S1 Fig in the supplemental materials for a participant flow diagram). Most were between the ages of 25 and 64 (72.7%), female (55.2%), White (73.5%) and lived in either a large or medium-sized metro area (59.1%). The most common insurance type was commercial (44.8%). Standardized differences for patient characteristics between those receiving hydrocodone or oxycodone treatment were generally small or close to zero.

Table 1

Patient and index prescription characteristics, by index prescription drug type.

	Full Cohort		Hydrocodone SA Treatment		Oxycodone SA Treatment		Standardized Differences
	N	%	N	%	N	%
Age
18–24	66,110	12.74	46,361	13.14	19,749	11.89	0.038
25–34	106,369	20.49	70,898	20.09	35,471	21.35	0.031
35–44	90,574	17.45	61,222	17.35	29,352	17.67	0.009
45–54	90,152	17.37	61,827	17.52	28,325	17.05	0.012
55–64	90,450	17.43	61,304	17.37	29,146	17.55	0.005
65–74	48,212	9.29	32,150	9.11	16,062	9.67	0.019
75+	27,199	5.24	19,191	5.44	8,008	4.82	0.028
Gender
F	286,488	55.19	190,687	54.03	95,801	57.67	0.073
M	232,578	44.81	162,266	45.97	70,312	42.33	0.073
Race/Ethnicity
White	381,580	73.51	258,135	73.14	123,445	74.31	0.027
Black	16,383	3.16	10,910	3.09	5,473	3.29	0.012
Hispanic	47,350	9.12	33,323	9.44	14,027	8.44	0.035
Asian-Pacific Islander	12,523	2.41	8,564	2.43	3,959	2.38	0.003
Other	7,908	1.52	5,598	1.59	2,310	1.39	0.016
Unknown	53,322	10.27	36,423	10.32	16,899	10.17	0.005
Insurance Plan in Index Year
Commercial	232,408	44.77	155,868	44.16	76,540	46.08	0.039
Medicaid	196,972	37.95	136,672	38.72	60,300	36.30	0.050
Medicare	67,031	12.91	45,421	12.87	21,610	13.01	0.004
Dual	22,453	4.33	14,854	4.21	7,599	4.57	0.018
Unknown	202	0.04	138	0.04	64	0.04	0.000
Urbanization in Index Year
Large central metro	93,689	18.05	60,065	17.02	33,624	20.24	0.083
Large fringe metro	114,626	22.08	73,610	20.86	41,016	24.69	0.092
Medium metro	98,412	18.96	68,742	19.48	29,670	17.86	0.041
Small metro	73,619	14.18	52,546	14.89	21,073	12.69	0.064
Micropolitan	48,618	9.37	35,776	10.14	12,842	7.73	0.084
Noncore	7,379	1.42	5,385	1.53	1,994	1.20	0.028
Unknown	82,723	15.94	56,829	16.10	25,894	15.59	0.014
Year of Index Prescription
2015	202,070	38.93	140,357	39.77	61,713	37.15	0.054
2016	177,365	34.17	120,426	34.12	56,939	34.28	0.003
2017	139,631	26.90	92,170	26.11	47,461	28.57	0.055
Index Prescription MME
MME < = 75	152,349	29.35	138,274	39.18	14,075	8.47	0.773
MME 76–100	108,985	21.00	100,861	28.58	8,124	4.89	0.669
MME 101–200	127,982	24.66	72,381	20.51	55,601	33.47	0.295
MME 201–300	85,245	16.42	31,293	8.87	53,952	32.48	0.610
MME >300	44,505	8.57	10,144	2.87	34,361	20.69	0.575
Index Prescription Days’ Supply
< = 3 Days	292,031	56.26	213,278	60.43	78,753	47.41	0.263
4–6 Days	151,032	29.10	92,024	26.07	59,008	35.52	0.206
7+ Days	76,003	14.64	47,651	13.50	28,352	17.07	0.099

SA, short-acting; MME, morphine milligram equivalents.

Index prescription characteristics

A total of 18,758 unique prescribers prescribed index opioid prescriptions of hydrocodone or oxycodone, with a median of 8 patients per prescriber (interquartile range (IQR) = 2–29). About two-thirds (68.0%) of patients filled an index prescription of hydrocodone and 32.0% filled oxycodone. The majority of patients (56.3%) filled a days’ supply less than or equal to 3 days. Standardized differences in dosages between patients receiving hydrocodone and those receiving oxycodone were large (Table 1). Most individuals who filled oxycodone received more than 100 MME (86.6%), while only 32.3% of those filling hydrocodone received more than 100 MME. Patients given oxycodone also received more days’ supply, with 52.6% of patients receiving a days’ supply greater than 3, compared to 39.6% of those with hydrocodone.

First-year chronic use

In the year after receiving the index prescription, 2.8% (n = 14,458) of patients met the criteria for chronic use. The median total days’ supply for individuals meeting the chronic use definition was 190 (IQR = 115–300), compared to 5 (IQR = 3–10) for people not meeting the definition. The raw percentage of patients receiving oxycodone that developed chronic use was 3.3%, compared to 2.5% of those with hydrocodone. After adjustment for patient and index prescription characteristics, however, patients receiving an index prescription of oxycodone had lower odds of developing chronic use relative to patients receiving hydrocodone (aOR = 0.95, 95% CI 0.91–1.00; Table 2). Further exploration of model results indicated that the change in relationship between drug type and chronic use after controlling for patient and index prescription characteristics was due to the addition of dosage to the model. When index prescription MME is excluded from the model, patients filling an oxycodone index prescription had significantly higher odds of developing chronic use relative to patients receiving hydrocodone (aOR = 1.36, 95% CI 1.29–1.42).

Table 2

Patient and index prescription characteristics associated with first-year chronic use.

	No Chronic Use (n)	Chronic Use (n)	Chronic Use Row %	Adjusted Odds Ratio (95% Confidence Interval)	p value
Total Cohort	504,608	14,458	2.79
Index Prescription Drug
Hydrocodone SA	344,029	8,924	2.53	ref	ref
Oxycodone SA	160,579	5,534	3.33	0.95 (0.91–1.00)	0.03
Age
18–24	65,558	552	0.83	Ref	ref
25–34	104,571	1,798	1.69	1.81 (1.64–2.00)	< .0001
35–44	88,277	2,297	2.54	2.71 (2.46–2.98)	< .0001
45–54	86,966	3,186	3.53	3.57 (3.25–3.92)	< .0001
55–64	86,932	3,518	3.89	3.82 (3.47–4.20)	< .0001
65–74	46,433	1,779	3.69	2.96 (2.62–3.33)	< .0001
75+	25,871	1,328	4.88	3.06 (2.70–3.47)	< .0001
Gender
F	278,826	7,662	2.67	0.98 (0.95–1.02)	0.27
M	225,782	6,796	2.92	ref	ref
Race/Ethnicity
White	370,164	11,416	2.99	ref	ref
Black	15,847	536	3.27	1.16 (1.05–1.28)	0.00
Hispanic	46,515	835	1.76	0.60 (0.56–0.65)	< .0001
Asian-Pacific Islander	12,393	130	1.04	0.37 (0.31–0.44)	< .0001
Other	7,603	305	3.86	1.09 (0.96–1.24)	0.17
Unknown	52,086	1,236	2.32	0.90 (0.84–0.97)	0.005
Insurance Plan in Index Year
Commercial	228,823	3,585	1.54	ref	ref
Medicaid	190,262	6,710	3.41	2.86 (2.73–3.00)	< .0001
Medicare	64,750	2,281	3.40	1.56 (1.43–1.69)	< .0001
Dual	20,575	1,878	8.36	3.75 (3.48–4.04)	< .0001
Unknown	198	4	1.98	1.36 (0.47–3.91)	0.57
Urbanization in Index Year
Large central metro	91,300	2,389	2.55	0.88 (0.75–1.03)	0.11
Large fringe metro	111,842	2,784	2.43	0.94 (0.80–1.10)	0.43
Medium metro	95,858	2,554	2.60	0.92 (0.79–1.08)	0.30
Small metro	71,519	2,100	2.85	0.92 (0.79–1.08)	0.30
Micropolitan	46,976	1,642	3.38	0.98 (0.84–1.15)	0.81
Noncore	7,118	261	3.54	ref	ref
Unknown	79,995	2,728	3.30	1.02 (0.88–1.19)	0.78
Year of Index Prescription
2015	195,521	6,549	3.24	ref	ref
2016	172,987	4,378	2.47	0.84 (0.81–0.88)	< .0001
2017	136,100	3,531	2.53	0.88 (0.84–0.92)	< .0001
Index Prescription MME
MME < = 75	149,757	2,592	1.70	ref	ref
MME 76–100	107,207	1,778	1.63	1.03 (0.96–1.10)	0.39
MME 101–200	124,579	3,403	2.66	1.48 (1.38–1.58)	< .0001
MME 201–300	82,382	2,863	3.36	1.84 (1.70–1.99)	< .0001
MME >300	40,683	3,822	8.59	3.89 (3.57–4.24)	< .0001
Index Prescription Days’ Supply
< = 3 Days	287,321	4,710	1.61	ref	ref
4–6 Days	147,755	3,277	2.17	1.08 (1.03–1.14)	0.004
7+ Days	69,532	6,471	8.51	2.49 (2.34–2.64)	< .0001

SA, short-acting; MME, morphine milligram equivalents.

The sensitivity analysis, using an alternative version of the chronic use definition, identified slightly more patients who meet criteria for chronic use (3.1%; n = 16,297). The median total days’ supply for people meeting the alternative chronic use definition was 174 (IQR = 116–282). There was significant overlap between the definitions: 85.3% of those who met the alternative definition also met our study definition, while 96.2% of those who met our study definition also met the alternative definition. S2 Table in the supplemental materials shows odds of first-year chronic use by index prescription and patient characteristics using the alternative definition. Consistent with the analyses using our study definition, individuals receiving an oxycodone index prescription had lower odds of chronic use (aOR = 0.91, 95% CI 0.87–0.95). Likewise, to address not including individuals who died in the year following their index prescription, we ascertained that 1.1% (n = 3,750) patients who received an index prescription of hydrocodone SA died in the following year compared with 1.7% (n = 2,868) who received oxycodone SA. Excluding the individuals who died during the follow-up year did not substantially change the chronic user results (oxycodone OR 0.94 (95% CI 0.89–0.98) vs. OR 0.95 (95% CI 0.91–1.0) compared with hydrocodone).

Opioid overdose was experienced by 0.3% (n = 1,480) of patients prior to censoring. The percentage of patients who filled oxycodone experiencing overdose was 0.38%, compared to 0.24% of those receiving hydrocodone. Table 3 shows adjusted hazard ratios for index prescription characteristics associated with subsequent opioid overdose, adjusting for patient and prescription characteristics. The adjusted hazard ratio (aHR) of having an overdose for individuals filling an oxycodone index prescription was 1.65 (95% CI 1.45–1.87) relative to those filling hydrocodone. To account for random variation among prescribers, we repeated the analysis above, but included a random prescriber effect on a sub-group of patients from prescribers with at least 200 patients who filled an index prescription of hydrocodone or oxycodone. The sample included 101,559 patients and 294 prescribers. Patients in the sub-sample experienced 287 overdoses during the study period. Results of this sub-analysis were similar to the overall sample findings, with a greater hazard of overdose for individuals filling an index prescription of oxycodone (aHR = 1.75, 95% CI 1.29–2.37) (S3 Table).

Table 3

Patient and index prescription characteristics associated with opioid overdose (fatal or non-fatal).

	No Overdose	Overdose	Overdose Row %	Adjusted Hazard Ratio (95% Confidence Interval)	p value
Total Cohort	517,586	1,480	0.29
Index Prescription Drug
Hydrocodone SA	352,096	857	0.24	ref	ref
Oxycodone SA	165,490	623	0.38	1.65 (1.45–1.87)	<0.01
Age
18–24	65,862	248	0.38	ref	ref
25–34	106,069	300	0.28	0.71 (0.60–0.84)	< .0001
35–44	90,358	216	0.24	0.66 (0.55–0.80)	< .0001
45–54	89,909	243	0.27	0.76 (0.64–0.91)	0.00
55–64	90,235	215	0.24	0.77 (0.64–0.93)	0.01
65–74	48,087	125	0.26	0.76 (0.56–1.03)	0.08
75+	27,066	133	0.49	1.45 (1.08–1.96)	0.02
Gender
F	285,770	718	0.25	0.68 (0.62–0.76)	<0.01
M	231,816	762	0.33	ref	ref
Race/Ethnicity
White	380,402	1,178	0.31	ref	ref
Black	16,312	71	0.43	0.96 (0.75–1.23)	0.76
Hispanic	47,258	92	0.19	0.50 (0.40–0.62)	<0.01
Asian-Pacific Islander	12,512	11	0.09	0.24 (0.13–0.43)	<0.01
Other	7,867	41	0.52	1.13 (0.83–1.55)	0.44
Unknown	53,235	87	0.16	0.93 (0.74–1.18)	0.57
Insurance Plan in Index Year
Commercial	232,197	211	0.09	ref	ref
Medicaid	196,050	922	0.47	5.39 (4.61–6.30)	<0.01
Medicare	66,855	176	0.26	2.12 (1.58–2.84)	<0.01
Dual	22,283	170	0.76	7.15 (5.65–9.04)	<0.01
Unknown	201	1	0.50	5.38 (0.75–38.45)	0.09
Urbanization in Index Year
Large central metro	93,337	352	0.38	ref	ref
Large fringe metro	114,361	265	0.23	0.74 (0.63–0.87)	<0.01
Medium metro	98,115	297	0.30	0.83 (0.71–0.98)	0.02
Small metro	73,421	198	0.27	0.69 (0.58–0.82)	<0.01
Micropolitan	48,495	123	0.25	0.60 (0.49–0.74)	<0.01
Noncore	7,365	14	0.19	0.46 (0.27–0.79)	0.01
Unknown	82,492	231	0.28	0.72 (0.60–0.86)	<0.01
Year of Index Prescription
2015	201,359	711	0.35	ref	ref
2016	176,885	480	0.27	1.07 (0.94–1.20)	0.31
2017	139,342	289	0.21	1.27 (1.10–1.47)	<0.01
Index Prescription MME
MME < = 75	151,909	440	0.29	ref	ref
MME 76–100	108,737	248	0.23	0.90 (0.77–1.06)	0.21
MME 101–200	127,599	383	0.30	0.95 (0.81–1.11)	0.53
MME 201–300	85,019	226	0.27	0.83 (0.68–1.01)	0.06
MME >300	44,322	183	0.41	1.04 (0.82–1.31)	0.78
Index Prescription Days Supply
< = 3 Days	291,199	832	0.28	ref	ref
4–6 Days	150,693	339	0.22	0.80 (0.70–0.92)	<0.01
7+ Days	75,694	309	0.41	1.35 (1.14–1.59)	<0.01

SA, short-acting; MME, morphine milligram equivalents.

After controlling for patient and prescription characteristics, patients receiving oxycodone were less likely than those with hydrocodone to have zero cumulative MME in the 6 months after the index prescription (aOR = 0.88, 95% CI 0.87–0.89), meaning that patients with oxycodone were more likely to have filled at least one additional opioid prescription after the index prescription fill. Cumulative 6-month MME for patients receiving oxycodone was 1.20 times greater than for those with hydrocodone (95% CI 1.18–1.22). The mean cumulative MME for patients with oxycodone was 393.85 (standard deviation (SD) = 2,013.34) and for those with hydrocodone was 213.37 (SD = 1,151.75).

Among the 215,726 patients with at least one additional prescription in the 6 months after the index prescription (representing 41.6% of the total cohort), more than a third (37.8%, n = 81,495) switched to another opioid. Of those patients who started on oxycodone, 41.5% switched to another opioid within 6 months (n = 29,977), while 35.9% of those started on hydrocodone switched (n = 51,518). After adjusting for patient and index prescription characteristics, patients whose index prescription was oxycodone were 1.24 times more likely to be switched to another opioid than those receiving hydrocodone (95% CI 1.21–1.27).

Combination product subgroup analysis

There were 478,133 patients in the sample who filled an index prescription for 5 mg strength immediate release oxycodone or hydrocodone. Of these, 319,942 filled hydrocodone-acetaminophen therapy (66.9%), 75,810 filled oxycodone-acetaminophen therapy (15.9%), and 82,381 filled oxycodone monotherapy (17.2%).

First-year chronic use was experienced by 2.5% (n = 12,004) of individuals in this sub-group analysis. Of those who filled an oxycodone monotherapy index prescription, 3.5% (n = 2,903) developed chronic use within the first year, compared to 2.4% (n = 1,822) among people filling an oxycodone-acetaminophen index prescription and 2.3% (n = 7,279) among people filling a hydrocodone-acetaminophen index prescription. Consistent with our previous modeling results, when dosage of the index prescription is not considered in the model, both oxycodone-acetaminophen (aOR = 1.22, 95% CI 1.15–1.30) and oxycodone monotherapy (aOR = 1.69, 95% CI 1.60–1.78) were associated with increased odds of first-year chronic use relative to hydrocodone-acetaminophen. However, when controlling for both patient and index prescription characteristics, patients who received oxycodone-acetaminophen were significantly less likely to experience first-year chronic use compared to those receiving hydrocodone-acetaminophen therapy (aOR 0.88, 95% CI 0.82–0.94). The likelihood of chronic use for those receiving oxycodone monotherapy was not significantly different from those receiving hydrocodone-acetaminophen (aOR 1.06, 95% CI 1.00–1.13) (S4 Table).

The total number of opioid overdoses in this subset of the sample was 1,346 (0.28%). Among individuals filling an oxycodone monotherapy index prescription, 0.44% experienced an opioid overdose in the study period, compared to 0.27% of those who received oxycodone-acetaminophen and 0.24% of those who received hydrocodone-acetaminophen. Relative to those receiving hydrocodone-acetaminophen, patients who received oxycodone monotherapy (aHR = 2.18, 95% CI 1.86–2.57) were at the greatest risk for opioid overdose, followed by those receiving oxycodone-acetaminophen combination therapy (aHR = 1.26, 95% CI 1.06–1.50), (S5 Table).

Discussion

Using a comprehensive database, and after adjusting for multiple patient- and prescription-level factors, our study found that the risk of first-year chronic use after an initial opioid prescription was lower with oxycodone compared to hydrocodone (aOR 0.95, 95% CI 0.91–1.0) but the risk of opioid overdose was markedly higher following oxycodone (aHR 1.65, 95% CI 1.45–1.87). This finding should be notable for providers who write these prescriptions. The rates of chronic use for hydrocodone and oxycodone (2.53–3.33% respectively) and overdose (0.24–0.38% respectively) have significant implications on the lives of the patients affected by them. These findings vary from prior literature which used limited datasets such as only PDMP or only commercial insured data [5,13], and likely reflect this study’s ability to capture prescriptions regardless of payer, control for prescription characteristics like MME, and control for patient level-factors including age, gender and comorbidities. The linkage with hospital discharge data allows an unprecedented description of overdose incidence after an initial prescription at the level of an entire state as compared to prior work that was limited only to certain populations.

An unexpected finding was the role of opioids in combination with acetaminophen. Whereas short-acting hydrocodone is always combined with other non-opioid active ingredients, prescribers have a choice when prescribing short-acting oxycodone. Prescribing an opioid medication in combination form allows clinicians to provide convenient multi-modal therapy, whereas prescribing the active ingredients separately allows clinicians to maximize the dose of the non-opioid first before adding an opioid for acute pain. For the outcome of chronic use, combination oxycodone-acetaminophen was associated with lower odds of chronic use (aOR 0.9 (95% CI 0.8–0.9)), whereas the outcome with oxycodone monotherapy was similar to hydrocodone/acetaminophen (aOR 1.1 (95% CI (1.0–1.1)). Conversely, both forms of oxycodone were associated with a higher risk of overdose when compared with hydrocodone-acetaminophen, but oxycodone monotherapy was associated with the highest risk of overdose (aHR 2.2 (95% CI 1.9–2.6)). Because we are unable to detect if patients on monotherapy concurrently used over-the-counter acetaminophen, we cannot describe the protective effect with certainty, nor can we ascertain if the opioid prescription and/or formulation was based on perceived risk by the prescriber or patient request. We were also unable to detect if acetaminophen overdose was higher when combination products were prescribed.

For our primary outcomes, socioeconomic and demographic factors did influence results. In the adjusted analysis, older age, Black race, Medicaid alone or in dual enrollment with Medicare Advantage and living outside urban areas were associated with chronic use. These factors may be important to prescribers who begin treatment for painful conditions that may become chronic, such as back pain. For overdose risk, however, another pattern emerged. Compared to the reference group of age 18–24, only the most elderly group (75+) was more likely to experience overdose, highlighting the elevated risk of injury or death in this population [39]. Men were more likely to experience overdose than women and White patients were more likely to overdose compared with Hispanic and Asian-Pacific Islander individuals. Contrasting with chronic use, those in metro areas were more likely to experience overdose.

We identified an interesting pattern for the MME of the index prescription. Regardless of the agent, higher MME for the index prescription was strongly associated with development of chronic use, a finding consistent with other studies [13,40]. An index prescription MME of >300 (equivalent to forty 5 mg tablets of oxycodone or hydrocodone) had an aOR for chronic use of 3.9 (95% CI 3.6–4.2) compared to a prescription for < = 75 MME. When evaluating overdose, however, the adjusted analysis showed no difference in odds of overdose based on the MME of the initial prescription. Days’ supply was associated with a higher risk only for prescriptions ≥7 days in duration (aHR 1.35 (95% CI 1.14–1.59)), although that variable may be unreliable given that it is calculated at the pharmacy level and may not be accurate for as-needed prescriptions. Although overdose was a rare event, the stronger association of oxycodone with overdose may indicate that the choice of agent is more important than the quantity prescribed, consistent with prior findings that the “likeability” of oxycodone may increase its harm potential [12]. Our finding that patients prescribed oxycodone were more likely to have an additional prescription beyond the first one further supports this hypothesis, although further work is also needed to determine if prescribers vary their prescription choice based on patient factors or indication. Our secondary analysis including a random prescriber effect for those with at least 200 patients prescribed an opioid yielded similar results, indicating that prescriber characteristics were unlikely to influence the findings.

Limitations

Data analysis was retrospective and based upon a dataset used for administrative purposes; clinical details that might affect prescribing decisions were unavailable, including pain severity or indications for prescribing. Probabilistic matching in the merged dataset may have failed in some cases. Black patients comprised about 3% of the cohort and about half of the race/ethnicity data had to be imputed, rendering findings about race and ethnicity potentially less accurate. Although we controlled for many sociodemographic factors, because of data use limitations we could not control for indication of the first prescription, such as major surgery vs. injury vs. a chronic painful illness (e.g. cancer). Therefore, it is possible that the results may differ if specific indications, including severity and chronicity of pain, were considered independently. Due to data availability, we included the Medicare Advantage population but did not include traditional Medicare fee-for-service insurance, meaning that the findings may not apply to the entire elderly population. ICD-10 codes do not perfectly delineate between overdoses from prescribed opioids vs. illicit opioids, so we could not define which type of opioid caused the overdose outcome. After the initial prescription, 41.6% of individuals received at least a second prescription, and 37.8% switched to another opioid, so the outcomes may not be directly related to the initial opioid dispensed. The chronic use analysis did not exclude people who died within a year of the initial dispensation, but our sensitivity analysis revealed that including them did not substantially change the results. Oregon’s PDMP reports opioid prescriptions which were dispensed, but not if the patient consumed some or all of the prescription. Likewise, the PDMP does not capture prescriber specialty, so it was not possible to evaluate prescriber-related factors, which may have influenced the results. We included a random prescriber effect sub-analysis to partially explore this limitation and saw no differences in results. Finally, this study used data from Oregon and the findings may not apply to other settings.

In conclusion, among patients who received an index prescription of hydrocodone or oxycodone, hydrocodone is slightly more associated with subsequent chronic use when compared with oxycodone, but oxycodone is much more likely to be associated with future overdose. Opioid formulations in combination with acetaminophen may be relatively protective against these adverse outcomes. Additionally, certain patient characteristics that are more likely to be associated with both chronic use and overdose should be considered by prescribers. Given the risks, opioids should be prescribed to previously naïve patients only when felt it is absolutely necessary. When that decision is made, hydrocodone may be the favorable agent.

Selection and inclusion of prescriptions and patients in the study.

(TIF)

Click here for additional data file.

ICD-9 or 10 diagnosis codes used to identify opioid-related overdose/poisoning for ED visits and inpatient hospitalization.

(DOCX)

Click here for additional data file.

Sensitivity analysis: Patient and index prescription characteristics associated with first-year chronic use, using an alternative chronic use definition.

(DOCX)

Click here for additional data file.

Multilevel cox regression sub-analysis: Patient and index prescription characteristics associated with opioid overdose.

(DOCX)

Click here for additional data file.

Combination drug sub-analysis: Patient and index prescription characteristics associated with first-year chronic use.

(DOCX)

Click here for additional data file.

Combination drug sub-analysis: Patient and index prescription characteristics associated with opioid overdose.

(DOCX)

Click here for additional data file.

14 Dec 2021

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Tara Gomes

Academic Editor

PLOS ONE

Journal requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf  and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Please provide additional details regarding participant consent. If you are reporting a retrospective study of medical records, archived samples or third party data, please ensure that you have discussed whether all data were fully anonymized before you accessed them and/or whether the IRB or ethics committee waived the requirement for informed consent. If patients provided informed written consent to have data from their medical records used in research, please include this information.

Once you have amended this/these statement(s) in the Methods section of the manuscript, please add the same text to the “Ethics Statement” field of the submission form (via “Edit Submission”).

3. Thank you for stating the following in the Acknowledgments Section of your manuscript:

“This work was funded by the National Institutes of Health, NIH 5-R01-DA044167 and 5-R01-HS026753.”

We note that you have provided information within the Acknowledgements Section. Please note that funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form.

Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

“National Institute on Drug Abuse 5-R01-DA044167 (SGW, MAH, SEI, GAR, SEH, CH, RDW, EWB, DF, PWK) and Agency for Healthcare Research and Quality 5-R01-HS026753 (SGW). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.”

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors conducted a study assessing predictors of chronic use and opioid-related overdose after an index prescription in a large linked retrospective cohort. While several other studies on this topic have been conducted, the study specifically focuses on the formulation dispensed in index dispensation (hydrocodone vs. oxycodone - two of the most commonly used opioids in the US). The large analysis (N=519,066) allowed for investigation into what are relatively rare outcomes (chronic use, opioid-related overdose). The analysis found that oxycodone is associated with a reduced risk of chronic use (even though the crude rate was higher for oxycodone), but increased risk of overdose. The study is well-conducted, although I have a couple of major comments related to adjustment of confounders in the model, need for further exploration into the change in direction of association for oxycodone and chronic use, and the chronic use definition. Also, a I think a key limitations is the lack of adjustment for indication, type of pain (e.g., acute, chronic), etc., which is touched upon briefly a couple times in Discussion but could be expanded upon.

1) Major comments

- Adjustment for dose of index dispensation is critical in the multivariable models, and adjustment for this confounder may be responsible for the change in direction of association between oxycodone and chronic use between the crude and adjusted analyses. Also, the finding of an association between increasing dose and chronic use, but not overdose risk, is one of the more interesting findings in the analysis. However, for the MME of the index dispensation, would it not make more sense to calculate and adjust for the average daily dose in MME across the index dispensation (cumulative MME divided by days supply) rather than the cumulative MME, particularly given that the authors also adjust for days’ supply in the model? I would imagine the cumulative MME and days’ supply would be highly correlated/collinear. Also, average daily dose may be a more meaningful value for physicians.

- As mentioned above, the crude chronic use rate was higher with Oxycodone but this relationship inverted after adjustment for other variables. I think it is important to investigate what confounder was responsible for changing the direction of association after adjustment. I am guessing that it is the dose but this would be good to fully unravel.

- Although touched upon briefly in the Discussion a couple times, it would be good to elaborate more on how the inability to adjust for severity of pain, indication, and chronic vs. acute pain may affect the analysis and the findings. I think this is a key limitation, as other studies assessing predictors of long-term use are able to adjust for indication or often focus on a specific acute event (e.g., surgery).

- Definitions for chronic/long-term use vary widely in literature and the chronic use definition in this analysis differs from one of the more commonly used definitions (see Von Korff et al - DOI: 10.1097/AJP.0b013e318169d03b). Please also see a review of long-term definitions and limitations of definitions used (https://onlinelibrary.wiley.com/doi/abs/10.1002/pds.4929). While this doesn't mean that the definition used in this paper is not valid, it might be good to address and discuss some of the limitations with the definition used. In particular, for people meeting the chronic use definition, it would be useful to know the average total days supply in the 365 days after index prescription. If my understanding is correct, it would be possible for some people to have only a very small number of days supply in these 365 days but still meet the chronic use definition (e.g., 6 dispensations with only 1 days drug supply in each, as an extreme example). I like the inclusion of the secondary outcome of cumulative MME because it overcomes some of the limitations of the chronic use definition.

2) Minor comments

Introduction

- The Introduction might want to include a greater focus on some of the negative consequences of long-term chronic opioid use (e.g., dependence, overdose, transitions to illicit drugs), given that this is one of the outcomes of the study. Transitions to illicit drugs is particularly important to mention given that the opioid overdose outcome used in this analysis is based on ICD codes that do not differentiate between prescription vs. illicit opioids. Also, since reducing overdose-related deaths is a key public health goal, it would also be good to discuss trends in opioid-related overdoses in Oregon within the Introduction (prescription vs. illicit, if possible).

- The Shah et al reference (#25) also compares oxycodone and hydrocodone (see paper's Supplementary material), and might want to be referenced in the Introduction.

- Lines 117-118 - Similar efficacy for acute or chronic pain, or both? It seems that the cited studies are mostly for acute pain.

- Lines 127-128 – It would be good to add 95% CIs for these odds ratios to provide insight into whether there is a statistically significant difference between the two.

- It might be worth discussing in Introduction or Discussion why some studies suggest higher risk of risk of chronic use with Oxycodone while others do not. Further, given that reference #5 was also performed by the study team, the Introduction should elaborate on how the current analysis builds upon this previous analysis.

Methods

- For the International audience, it would be good to have some additional information on the different health plans (Medicaid vs. medicare vs. commercial, etc), such as who is eligible etc. This is particularly important given the strong associations between insurance plan and the outcomes in this analysis.

- Lines 160-161 - Please provide rationale for why long-acting formulations were excluded, as it is likely that most of the studies cited in the Introduction (e.g., number of prescriptions dispensed in 2017, efficacy for pain, development of chronic use) included long-acting formulations?

- Opioid overdose definition – please reference other literature that has also used this definition. It would be good to mention here and in the Limitations that these codes capture overdoses from both illicit and prescription opioids.

- With regards to switching, is it possible hat an individual may have been dispensed two different types of opioid on the same day (oxycodone and another), and therefore they are not identifying an actual switch?

- Did the analysis with chronic use as the outcome exclude people who died within 365 days of the initial dispensation? If not, the limitations of not doing so should be noted (not all individuals had same follow-up to be classified as a chronic user), particularly given that people dispensed oxycodone were more likely to die from overdose (and potentially other causes).

Discussion

- Lines 300-301 – it would be good to add 95% CIs here and/or note that the findings were statistically significant.

In first paragraph of discussion, can compare to existing literature (references 5, 13 and 14, as noted in Introduction) and discuss why contrasting findings exist and how this study fills a gap in literature.

- Lines 306-307 – the paragraph starts by saying there are many potential explanations for these findings and then states that “the first” is related to combination formulations. However, there does not seem to be a second or third explanation within this paragraph or elsewhere in Discussion.

- Lines 311-314 – Can the authors comment on whether this explanation (ie. that combination formulations allow for lower dose thereby reducing overdose risk) would not be accounted for by adjusting for MME in the analysis?

- The authors suggest that combination formulations may be protective against overdose because it limits the dose of the opioid, but then also say that “prescribing the active ingredients separately allows clinicians to maximize the dose of the non-opioid before adding an opioid”. These statements seem to suggest that both strategies (ie. combination formulations and prescribing medications separately) could serve to limit the opioid dose.

- Somewhere in discussion, it would be good to compare the rate of transition to chronic use in their analysis (2.8%) to other studies assessing this transition rate.

- Lines 330-331 say that MME and days supply were not associated with overdose, but this isn’t true. Table 3 shows that initial days supply of 7+ is associated with higher risk of overdose.

- It would be good to note in the Discussion/Limitations that there was a high rate of switching and therefore chronic use or overdose might not be directly related to the initial opioid dispensed.

Reviewer #2: The authors present a retrospective cohort analysis of patients residing in Oregon, USA, who initiated hydrocodone or oxycodone as their first opioid prescription. The analysis uses novel linkages of databases including the Oregon prescription drug monitoring program (PDMP) and Oregon All Payer All Claims (APAC) databases. Primary outcomes included chronic opioid use and opioid overdose. The current analysis is rigorous, and the manuscript is well-written; however, I believe analyses related to health services factors and combination products could enhance the paper prior to publication.

Major considerations

1. My major concern is that the paper focuses on patient outcomes for those dispensed hydrocodone versus oxycodone, yet some of the outcomes are likely driven by unaccounted health services factors. Mainly, prescriber is not considered in the analysis. The choice of opioid is likely largely dependent on the prescriber. Prescriber preference is a well-established factor for exposures when two in-class drugs are shown to have clinical equipoise; importantly, prescriber may also be related to the clinical outcomes studied (e.g., patients seeing an oncologist are likely to become chronic opioid users vs. those prescribed a short opioid course after dental surgery). Based on my understanding, the NPI is a required entry the PDMP and should be available in the dataset. Is this the case?

a. If this information is available, I recommend a secondary analysis that examines the outcomes using the same multivariable models, but adds prescriber as a clustering variable (so, a multi-level model with NPI as a clustering variable).

b. At minimum, the authors should recognize lack of prescriber information as a limitation in their Discussion section and hypothesize how accounting for this information might affect results.

2. The authors briefly mention the potential that the use of combination agents for hydrocodone may in part explain results. To strengthen conclusions, I suggest to describe the proportion of initial prescriptions that were for a combination product, and, ideally, conduct a secondary analysis examining outcomes stratified by combination vs. single-agent product for the initial prescription (understanding that over the counter analgesic use will not be captured). The current analysis, which uses National Drug Codes, should be able to accommodate this.

Minor considerations

1. Given the large sample size, P-values are not particularly useful to examine differences in patient characteristics between exposure groups. Instead, I suggest the authors use standardized mean differences (Austin 2009 Stat Med [PMCID: PMC3472075] is helpful).

2. Please mention the dates of the study within the abstract for readers.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: Yes: Kaleen Hayes

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

9 Feb 2022

PONE-D-21-33002

Opioid-Related Overdose and Chronic Use Following an Initial Prescription of Hydrocodone versus Oxycodone

PLOS ONE

Dear Dr. Weiner,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Response: Thank you. We are pleased to provide this revised version which addresses the majority of the Reviewers’ requests.

Specifically, we would like you to consider several suggestions for analysis that the reviewers have outlined including:

1. Justifying your chronic use definition and considering applying a more commonly applied definition as a sensitivity analysis.

Response: We conducted a sensitivity analysis to examine findings from the alternative definition (derived from Van Korff et al. as suggested by the reviewer below) and added it to the manuscript. The alternative definition was slightly more inclusive than our definition, but there was significant overlap between the definitions – 96% of those who met our study definition also met the alternative definition, while 85.3% of those who met the alternative definition also met our study definition.

2. Speaking more to the potential impact of having no indication data in your data. If this data is available, I would strongly urge the authors to incorporate it into the analysis in some way.

Response: Unfortunately, due to data use limitations, we were not permitted to link diagnosis to opioid prescriptions in the PDMP. Therefore, we do not have indication data. We have stated this fact, and its implications, more clearly in the Limitations.

3. Consider a stratified analysis where you consider combination products separately from single-agent opioids.

Response: We had also wondered about this question and had prepared a separate manuscript addressing this request. However, we have decided to merge that analysis into this paper to address this suggestion. The results of subgroup analyses that address this issue have been added to the manuscript. The results are interesting, and show that individuals who received oxycodone in either form (either alone or in combination with acetaminophen) were more likely to experience overdose than individuals receiving hydrocodone-acetaminophen, and those receiving oxycodone monotherapy were at greatest risk of overdose. However, for chronic use, individuals receiving oxycodone-acetaminophen were significantly less likely to experience chronic use than those receiving hydrocodone-acetaminophen or oxycodone alone.

4. Incorporate prescriber-related factors as suggested by Reviewer 2.

Response: We were able to update the chronic use models to include a random effect for prescriber that will help adjust the model for unmeasured prescriber factors that may affect prescribing and outcomes. Due to the memory limitations of our software, were not able to conduct a multilevel Cox regression on the entire sample on the overdose outcome. We were, however, able to run the multilevel Cox regression on a subset of prescribers with at least 200 patients who filled their opioid prescriptions. The results we obtained from this model were very similar to the results from the Cox regression without the prescriber random effect. We added this sub-analysis to the manuscript and a table showing the results is in the supplemental materials.

Note, that Oregon’s PDMP does not capture prescriber characteristics, such as specialty. The database does include DEA/NPI numbers and it would be theoretically possible to look these up on an individual level. Unfortunately, due to data use protections, the final linked database did not include that information and we are unable to go back and obtain it.

Furthermore, in revising the manuscript, we suggest the authors also consider the following:

1. Refer to recommendations on language to reduce stigma and apply this to the revised text (e.g. avoiding use of the word 'abuse'. See https://www.drugabuse.gov/nidamed-medical-health-professionals/health-professions-education/words-matter-terms-to-use-avoid-when-talking-about-addiction for example)

Response: We are very attuned to the issue of stigma. The only places we used the word “abuse” were in reference to medications and not individuals (i.e. “abuse potential” or “abuse liability”), which are commonly used in the medical literature. However, we have changed this to “misuse” as it captures the same concept and hopefully will not be perceived as stigmatizing.

2. Add a limitation related to the fact that you don't know if people actually took the medication dispensed.

Response: Done.

3. Temper comments such as " In an era when opioid overdose deaths remain alarmingly high" - recent reporting from the CDC demonstrates how the illicit drug supply (ie. synthetic opioids like fentanyl) is now the predominant driver of the overdose crisis. As such, this strong wording should be softened.

Response: Done. We have removed this phrase.

Please submit your revised manuscript by Jan 28 2022 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Tara Gomes

Academic Editor

PLOS ONE

Journal requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

Response: Done

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Response: Done

2. Please provide additional details regarding participant consent. If you are reporting a retrospective study of medical records, archived samples or third party data, please ensure that you have discussed whether all data were fully anonymized before you accessed them and/or whether the IRB or ethics committee waived the requirement for informed consent. If patients provided informed written consent to have data from their medical records used in research, please include this information.

Response: Done

Once you have amended this/these statement(s) in the Methods section of the manuscript, please add the same text to the “Ethics Statement” field of the submission form (via “Edit Submission”).

Response: Done

3. Thank you for stating the following in the Acknowledgments Section of your manuscript:

“This work was funded by the National Institutes of Health, NIH 5-R01-DA044167 and 5-R01-HS026753.”

We note that you have provided information within the Acknowledgements Section. Please note that funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form.

Response: Done

Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

“National Institute on Drug Abuse 5-R01-DA044167 (SGW, MAH, SEI, GAR, SEH, CH, RDW, EWB, DF, PWK) and Agency for Healthcare Research and Quality 5-R01-HS026753 (SGW). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.”

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

Response: Thank you. We have included this in the cover letter. There is one small edit (addition of a middle initial):

“National Institute on Drug Abuse 5-R01-DA044167 (SGW, MAH, SEI, GAR, SEH, CH, RDW, EWB, DPF, PWK) and Agency for Healthcare Research and Quality 5-R01-HS026753 (SGW). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.”

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: No

Response: The terms of the data use agreement prohibit public sharing of the source data used in this study but may be obtained via request to the public health entities in Oregon that oversee the individual source datasets. We have indicated this in the text.

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The authors conducted a study assessing predictors of chronic use and opioid-related overdose after an index prescription in a large linked retrospective cohort. While several other studies on this topic have been conducted, the study specifically focuses on the formulation dispensed in index dispensation (hydrocodone vs. oxycodone - two of the most commonly used opioids in the US). The large analysis (N=519,066) allowed for investigation into what are relatively rare outcomes (chronic use, opioid-related overdose). The analysis found that oxycodone is associated with a reduced risk of chronic use (even though the crude rate was higher for oxycodone), but increased risk of overdose. The study is well-conducted, although I have a couple of major comments related to adjustment of confounders in the model, need for further exploration into the change in direction of association for oxycodone and chronic use, and the chronic use definition. Also, a I think a key limitations is the lack of adjustment for indication, type of pain (e.g., acute, chronic), etc., which is touched upon briefly a couple times in Discussion but could be expanded upon.

1) Major comments

- Adjustment for dose of index dispensation is critical in the multivariable models, and adjustment for this confounder may be responsible for the change in direction of association between oxycodone and chronic use between the crude and adjusted analyses. Also, the finding of an association between increasing dose and chronic use, but not overdose risk, is one of the more interesting findings in the analysis. However, for the MME of the index dispensation, would it not make more sense to calculate and adjust for the average daily dose in MME across the index dispensation (cumulative MME divided by days supply) rather than the cumulative MME, particularly given that the authors also adjust for days’ supply in the model? I would imagine the cumulative MME and days’ supply would be highly correlated/collinear. Also, average daily dose may be a more meaningful value for physicians.

Response: Thank you. We agree with the interesting finding about increasing dose and chronic use (but not overdose risk). We were hesitant to include days’ supply in the model because it can be problematic in PDMP-based opioid-related research because, for acute opioid prescriptions, prescribers don’t write an intended number of days for “as needed” prescriptions. Days are calculated at the pharmacy or PDMP level based on if a patient were to take the maximum number of pills. As an example, a prescription for 2 tablets, every six hours as needed for severe pain, 24 tablets would be recorded as a 3 days’ supply even if the prescriber didn’t intend the patient to take them all in 3 days. That reality also discourages us from calculating cumulate MME/days’ supply. However, day’s supply and MME are not multi-collinear, so they are both included in the model.

- As mentioned above, the crude chronic use rate was higher with oxycodone but this relationship inverted after adjustment for other variables. I think it is important to investigate what confounder was responsible for changing the direction of association after adjustment. I am guessing that it is the dose but this would be good to fully unravel.

Response: We investigated this within the modeling and found that inclusion of MME in the model was responsible for the change in relationship after confounder adjustments. These findings have been added to the manuscript.

- Although touched upon briefly in the Discussion a couple times, it would be good to elaborate more on how the inability to adjust for severity of pain, indication, and chronic vs. acute pain may affect the analysis and the findings. I think this is a key limitation, as other studies assessing predictors of long-term use are able to adjust for indication or often focus on a specific acute event (e.g., surgery).

Response: Done. We have added additional language to this Limitations section.

- Definitions for chronic/long-term use vary widely in literature and the chronic use definition in this analysis differs from one of the more commonly used definitions (see Von Korff et al - DOI: 10.1097/AJP.0b013e318169d03b). Please also see a review of long-term definitions and limitations of definitions used (https://onlinelibrary.wiley.com/doi/abs/10.1002/pds.4929). While this doesn't mean that the definition used in this paper is not valid, it might be good to address and discuss some of the limitations with the definition used. In particular, for people meeting the chronic use definition, it would be useful to know the average total days supply in the 365 days after index prescription. If my understanding is correct, it would be possible for some people to have only a very small number of days supply in these 365 days but still meet the chronic use definition (e.g., 6 dispensations with only 1 days drug supply in each, as an extreme example). I like the inclusion of the secondary outcome of cumulative MME because it overcomes some of the limitations of the chronic use definition.

Response: We conducted a sensitivity analysis to examine findings from the alternative definition (derived from Van Korff et al) and added it to the manuscript. We also added the median total days’ supply for individuals meeting each definition to the text. The alternative definition was slightly more inclusive than our definition, but there was significant overlap between the definitions – 96% of those who met our study definition also met the alternative definition, while 85.3% of those who met the alternative definition also met our study definition.

2) Minor comments

Introduction

- The Introduction might want to include a greater focus on some of the negative consequences of long-term chronic opioid use (e.g., dependence, overdose, transitions to illicit drugs), given that this is one of the outcomes of the study. Transitions to illicit drugs is particularly important to mention given that the opioid overdose outcome used in this analysis is based on ICD codes that do not differentiate between prescription vs. illicit opioids. Also, since reducing overdose-related deaths is a key public health goal, it would also be good to discuss trends in opioid-related overdoses in Oregon within the Introduction (prescription vs. illicit, if possible).

Response: As suggested, we have added two additional paragraphs to the Introduction discussing the negative consequences of opioid use, transition to illicit opioids, and specifics about Oregon.

- The Shah et al reference (#25) also compares oxycodone and hydrocodone (see paper's Supplementary material), and might want to be referenced in the Introduction.

Response: Excellent suggestion – we were unaware of that supplementary material. We have now included this information in the Introduction.

- Lines 117-118 - Similar efficacy for acute or chronic pain, or both? It seems that the cited studies are mostly for acute pain.

Response: We split this sentence into two to make it clearer to the reader.

- Lines 127-128 – It would be good to add 95% CIs for these odds ratios to provide insight into whether there is a statistically significant difference between the two.

Response: Done. The CIs do not overlap.

- It might be worth discussing in Introduction or Discussion why some studies suggest higher risk of risk of chronic use with Oxycodone while others do not. Further, given that reference #5 was also performed by the study team, the Introduction should elaborate on how the current analysis builds upon this previous analysis.

Response: We have attempted to explain this concept better in this version of the manuscript. We believe prior studies have been subject to several limitations. The Ohio study was limited to PDMP data and did not control for any patient-level data. Other studies are limited to certain insurance datasets and do not include self-pay patients. We believe our study provides the most compelling evidence to date, and also includes the overdose outcome which has been incomplete in past work.

Methods

- For the International audience, it would be good to have some additional information on the different health plans (Medicaid vs. medicare vs. commercial, etc), such as who is eligible etc. This is particularly important given the strong associations between insurance plan and the outcomes in this analysis.

Response: Done. We have added a sentence briefly describing the different insurance types.

- Lines 160-161 - Please provide rationale for why long-acting formulations were excluded, as it is likely that most of the studies cited in the Introduction (e.g., number of prescriptions dispensed in 2017, efficacy for pain, development of chronic use) included long-acting formulations?

Response: This has been added. Long-acting formulations were excluded due to the small percentage of index prescriptions that were long-acting in our sample (e.g., 0.05% of hydrocodone or oxycodone index prescriptions in 2015).

- Opioid overdose definition – please reference other literature that has also used this definition. It would be good to mention here and in the Limitations that these codes capture overdoses from both illicit and prescription opioids.

Response: We have added to the methods that we are capturing both illicit and prescription opioids in our definition of overdose. We have referenced SAMHSA’s guidance for defining opioid overdose. In the limitations, we now state: “ICD-10 codes do not perfectly delineate between overdoses from prescribed opioids vs. illicit opioids, so we could not define which type of opioid caused the overdose outcome.”

- With regards to switching, is it possible that an individual may have been dispensed two different types of opioid on the same day (oxycodone and another), and therefore they are not identifying an actual switch?

Response: This is not possible because, as stated in the Methods, individuals dispensed more than one opioid on their index prescription date were excluded from the analysis.

- Did the analysis with chronic use as the outcome exclude people who died within 365 days of the initial dispensation? If not, the limitations of not doing so should be noted (not all individuals had same follow-up to be classified as a chronic user), particularly given that people dispensed oxycodone were more likely to die from overdose (and potentially other causes).

Response: Excellent point. No, we did not exclude people who died within 365 days of the initial dispensation, and have now listed this in the limitations as suggested.

Discussion

- Lines 300-301 – it would be good to add 95% CIs here and/or note that the findings were statistically significant.

Response: Done.

In first paragraph of discussion, can compare to existing literature (references 5, 13 and 14, as noted in Introduction) and discuss why contrasting findings exist and how this study fills a gap in literature.

Response: We are happy to make this suggested change, which highlights the novelty and rigor of our study.

- Lines 306-307 – the paragraph starts by saying there are many potential explanations for these findings and then states that “the first” is related to combination formulations. However, there does not seem to be a second or third explanation within this paragraph or elsewhere in Discussion.

Response: The subsequent reasons are in the following paragraphs. To make this clearer, we have modified the first sentence of the following paragraphs.

- Lines 311-314 – Can the authors comment on whether this explanation (ie. that combination formulations allow for lower dose thereby reducing overdose risk) would not be accounted for by adjusting for MME in the analysis?

Response: This question has been addressed by the new secondary analysis of combination vs. solo formulations of oxycodone.

- The authors suggest that combination formulations may be protective against overdose because it limits the dose of the opioid, but then also say that “prescribing the active ingredients separately allows clinicians to maximize the dose of the non-opioid before adding an opioid”. These statements seem to suggest that both strategies (ie. combination formulations and prescribing medications separately) could serve to limit the opioid dose.

Response: Here, we intend that a prescriber can fully separate out the agents. That is, they can maximize use of non-opioids, and then only add an opioid when the non-opioids are not working. With combination pills, the prescriber starts immediately with an opioid. Therefore, the strategy of separating the agents allows for no opioid use vs. some opioid use. We added the word “first” to this sentence which we believe makes the point easier to understand.

- Somewhere in discussion, it would be good to compare the rate of transition to chronic use in their analysis (2.8%) to other studies assessing this transition rate.

Response: Done. We have added text in both the Introduction and Discussion to address this suggestion.

- Lines 330-331 say that MME and days supply were not associated with overdose, but this isn’t true. Table 3 shows that initial days supply of 7+ is associated with higher risk of overdose.

Response: The Reviewer is correct, and we have made this change.

- It would be good to note in the Discussion/Limitations that there was a high rate of switching and therefore chronic use or overdose might not be directly related to the initial opioid dispensed.

Response: Excellent point. We have added this to the Limitations.

Reviewer #2: The authors present a retrospective cohort analysis of patients residing in Oregon, USA, who initiated hydrocodone or oxycodone as their first opioid prescription. The analysis uses novel linkages of databases including the Oregon prescription drug monitoring program (PDMP) and Oregon All Payer All Claims (APAC) databases. Primary outcomes included chronic opioid use and opioid overdose. The current analysis is rigorous, and the manuscript is well-written; however, I believe analyses related to health services factors and combination products could enhance the paper prior to publication.

Response: Thank you for these comments. We address the suggestions below.

Major considerations

1. My major concern is that the paper focuses on patient outcomes for those dispensed hydrocodone versus oxycodone, yet some of the outcomes are likely driven by unaccounted health services factors. Mainly, prescriber is not considered in the analysis. The choice of opioid is likely largely dependent on the prescriber. Prescriber preference is a well-established factor for exposures when two in-class drugs are shown to have clinical equipoise; importantly, prescriber may also be related to the clinical outcomes studied (e.g., patients seeing an oncologist are likely to become chronic opioid users vs. those prescribed a short opioid course after dental surgery). Based on my understanding, the NPI is a required entry the PDMP and should be available in the dataset. Is this the case?

a. If this information is available, I recommend a secondary analysis that examines the outcomes using the same multivariable models, but adds prescriber as a clustering variable (so, a multi-level model with NPI as a clustering variable).

b. At minimum, the authors should recognize lack of prescriber information as a limitation in their Discussion section and hypothesize how accounting for this information might affect results.

Response: See our comment above regarding this issue. We do not have much in the way of prescriber factors, but we were able to add a random effect for prescriber to most of our models. For the Cox regression, we had to run a model on a subset of the sample due to memory limitations, but results were similar to the original models that did not include a prescriber random effect.

2. The authors briefly mention the potential that the use of combination agents for hydrocodone may in part explain results. To strengthen conclusions, I suggest to describe the proportion of initial prescriptions that were for a combination product, and, ideally, conduct a secondary analysis examining outcomes stratified by combination vs. single-agent product for the initial prescription (understanding that over the counter analgesic use will not be captured). The current analysis, which uses National Drug Codes, should be able to accommodate this.

Response: We have added an additional sub-analysis analysis that compares chronic use and overdose risk for combination products compared to monotherapy.

Minor considerations

1. Given the large sample size, P-values are not particularly useful to examine differences in patient characteristics between exposure groups. Instead, I suggest the authors use standardized mean differences (Austin 2009 Stat Med [PMCID: PMC3472075] is helpful).

Response: This has been updated.

2. Please mention the dates of the study within the abstract for readers.

Response: Done.

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: Yes: Kaleen Hayes

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Submitted filename: Response to Review 20220126.docx

Click here for additional data file.

11 Mar 2022

Opioid-Related Overdose and Chronic Use Following an Initial Prescription of Hydrocodone versus Oxycodone

PLOS ONE

Dear Dr. Weiner,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

1. The changes to the introduction have strengthened the paper, however the rationale in your final paragraph which states "Given these hazards associated with opioid use, and because oxycodone and hydrocodone appear to be used interchangeably for acute pain..." doesn't follow from the paragraph above where you outline that rises in opioid-related deaths are being driven by the illicit supply. I would suggest changing this sentence to address the fact that current rising trends are not being driven by prescriptions, but that despite this, there is a need to ensure safe prescription opioid practices.

2. You may want to consider including citations to the revised CDC guidelines that are now released for comment (https://www.regulations.gov/document/CDC-2022-0024-0002) as they provide more up-to-date guidance related to opioid use for CNCP.

3. When describing the alternative definition of chronic use, consider rephrasing for clarity as "...with either a total days’ supply greater than or equal to 120 or 10+ prescriptions filled."

4. As one of the reviewers pointed out originally, there are limitations to the decision not to exclude those who died in the 365 days of follow-up as they had a different timeline to meet the chronic use criteria. While you added this as a limitation in the revision, sufficient details weren't provided to understand the impact of this limitation. Can you please provide further detail as to how deaths were accounted for in your chronic use definition (ie when defining periods uncovered by opioid prescriptions being <30 days, what if someone died 60 days before the 1-year follow-up period ended? Would that be counted as a 60 day period 'uncovered' or did you only consider the period of time when they were alive to identify gaps?). Can you also provide the N(%) of people who died during the first year by exposure group to allow an assessment of whether the prevalence of death differed by group (which could therefore bias the findings), or whether it was fairly rare/similar between groups (in which case the decision not to exclude these individuals likely is less problematic).

Please submit your revised manuscript by Apr 25 2022 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Tara Gomes

Academic Editor

PLOS ONE

Journal Requirements:

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

17 Mar 2022

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Response: Thank you for the additional review. We are happy to address these remaining concerns.

1. The changes to the introduction have strengthened the paper, however the rationale in your final paragraph which states "Given these hazards associated with opioid use, and because oxycodone and hydrocodone appear to be used interchangeably for acute pain..." doesn't follow from the paragraph above where you outline that rises in opioid-related deaths are being driven by the illicit supply. I would suggest changing this sentence to address the fact that current rising trends are not being driven by prescriptions, but that despite this, there is a need to ensure safe prescription opioid practices.

Response: We have made the adjustment to this paragraph as suggested.

2. You may want to consider including citations to the revised CDC guidelines that are now released for comment (https://www.regulations.gov/document/CDC-2022-0024-0002) as they provide more up-to-date guidance related to opioid use for CNCP.

Response: This is a delicate balance as we don’t know when the new guidelines will come out nor how they will differ from the version which is open to public comment presently (until April 11). We therefore modified the sentence about the guidelines to read: “Due to these and other complications, avoidance of chronic opioid therapy for pain not related to malignancy is recommended by the Centers for Disease Control and Prevention’s 2016 opioid guidelines, although an updated version is currently forthcoming [19, 20].” Hopefully, this edit adequately addresses the suggestion.

3. When describing the alternative definition of chronic use, consider rephrasing for clarity as "...with either a total days’ supply greater than or equal to 120 or 10+ prescriptions filled."

Response: We have made the suggested edit.

4. As one of the reviewers pointed out originally, there are limitations to the decision not to exclude those who died in the 365 days of follow-up as they had a different timeline to meet the chronic use criteria. While you added this as a limitation in the revision, sufficient details weren't provided to understand the impact of this limitation. Can you please provide further detail as to how deaths were accounted for in your chronic use definition (ie when defining periods uncovered by opioid prescriptions being <30 days, what if someone died 60 days before the 1-year follow-up period ended? Would that be counted as a 60 day period 'uncovered' or did you only consider the period of time when they were alive to identify gaps?). Can you also provide the N(%) of people who died during the first year by exposure group to allow an assessment of whether the prevalence of death differed by group (which could therefore bias the findings), or whether it was fairly rare/similar between groups (in which case the decision not to exclude these individuals likely is less problematic).

Response: For people who died, we did not limit the follow-up period to the time they were alive. Among those receiving an index prescription of hydrocodone SA, 1.06% (n = 3,750) died during the follow-up year. Among those receiving oxycodone SA, 1.73% (n = 2,868) died during the follow-up year. We also reran the chronic use analysis after excluding individuals that died during the follow-up year. The OR for oxycodone SA was 0.94 (95% CI 0.89 – 0.98; reference = hydrocodone SA). Comparing this to the estimate for oxycodone SA in the original model that did not exclude individuals who died during the follow-up (OR 0.95, 95% CI 0.91-1.00) demonstrated very little change in the results. We have added these findings to the Limitations section.

21 Mar 2022

21 Mar 2022

Thank you for the additional suggestion, which we are happy to address.

As requested, we have moved the sensitivity analysis from the Limitations to the Methods and Results, leaving one sentence in Limitations. Hopefully, the article now meets the journal's formatting requirements.

23 Mar 2022

Opioid-Related Overdose and Chronic Use Following an Initial Prescription of Hydrocodone versus Oxycodone

PONE-D-21-33002R3

Dear Dr. Weiner,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Tara Gomes

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

28 Mar 2022

PONE-D-21-33002R3

Opioid-related overdose and chronic use following an initial prescription of hydrocodone versus oxycodone

Dear Dr. Weiner:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Tara Gomes

Academic Editor

PLOS ONE