Trends in chronic opioid therapy among survivors of head and neck cancer

Abstract

Background:

Survivors of head and neck cancer (HNC) have increased risk of opioid misuse.

Methods:

Using Surveillance, Epidemiology and End-Results-Medicare data, we matched adults ≥66 years diagnosed with HNC 2008–2015 with cancer-free controls. We computed odds ratios (OR) for receipt of chronic opioid therapy (COT, claims for ≥90 consecutive days) for HNC survivors compared to controls each year after matching through 2016.

Results:

The cohort of HNC survivors declined from 5107 in the first year after diagnosis to 604 in the sixth year after diagnosis. For 5 years, rates of COT among HNC survivors exceeded that of controls. Differences between survivors and controls declined each year (ORs: year 1, 4.36; year 2, 2.60; year 3, 2.18; year 4, 1.85; and year 5, 1.35; all P-values <.05).

Conclusions:

Among older HNC survivors, cancer-associated opioid use in the first years after diagnosis suggests that the benefit of opioids must balance the risk of opioid misuse.

1 |. INTRODUCTION

Advances in treatment for head and neck cancer (HNC) have improved survival, but they have led to an increase in toxicities, including acute and chronic pain.^1–9 Common treatment-related painful toxicities include chemotherapy-induced neuropathies, muscle fibrosis after surgery or radiation, and, less commonly, osteoradionecrosis of the jaw after radiation. With HNC survivors living longer (67% 5-year survival), pain may persist or arise long after treatment completion, diminishing function and quality of life.^10,11 HNC patients have been reported to have among the highest prevalence of pain when compared across multiple cancer types, with 42% of HNC survivors reporting pain.^12–14

Opioids are commonly used for short- and long-term cancer pain management in HNC.^15–17 Exposure to opioids, in combination with other risk factors that are common in survivors of HNC (eg, history of addiction, anxiety, and distress) may all contribute to opioid misuse. Other opioid-related harms, such as fatigue, xerostomia, and endocrinopathies, may also be especially damaging in the context of treatment-related toxicities and an older survivor population, who may be susceptible to polypharmacy, falls, and fractures.^18–20

There is limited evidence characterizing the risk of opioid dependence in the years after treatment for HNC.^17,21,22 Single institution retrospective reviews have found that 7% to 15% of patients treated for HNC continued to use opioids at >1 year out from curative intent radiation therapy or surgery.^23,24 These studies do not elucidate whether HNC survivors actually have higher rates of chronic opioid use than people without cancer. A recent study of survivors of other cancers have shown dramatic differences in chronic opioid use between cancer populations, limiting the ability to generalize across cancers.²⁵ We investigated rates of chronic opioid therapy (COT), a risk factor for opioid dependence and other harms, among older HNC survivors up to 6 years out from cancer diagnosis. We hypothesize that survivors of HNC have higher rates of chronic opioid use than matched cancer-free individuals.

2 |. MATERIALS AND METHODS

The methodology for this analysis is adapted from prior work in other cancer populations and is described in more detail elsewhere.²⁵ This study was deemed exempt by the Institutional Review Board of Memorial Sloan Kettering Cancer Center and was performed under a data use agreement with the National Cancer Institute.

2.1 |. Data source

The Surveillance, Epidemiology and End-Results (SEER) registry linked to Medicare claims was used to identify cases of primary HNCs. The SEER registry covers approximately 28% of the United States population and includes information regarding site and extent of disease, patient demographics, and survival for incident cancer cases.^26,27 Medicare is a federal insurance program that provides insurance to adults over age 65 and some younger people with certain medical conditions. Medicare data include information regarding prescription fills, outpatient visits, and hospice use.²⁸ SEER also provides demographic data on a subpopulation of Medicare beneficiaries that do not have a documented history of cancer, which can be linked to Medicare claims.

2.2 |. Identification of opioids and definition of chronic opioid use

Opioids were identified from a list of generic names identified in prior work²⁵ (Table S1). Chronic opioid use was defined by 90 consecutive days of opioid use, allowing for up to a 7-day gap between consecutive fills.^29,30

2.3 |. Cohort

Patients with HNC diagnosed between 2008 and 2015 as their first and only cancer (survivors) were identified. Survivors ≥66 years at diagnosis were included to enable ascertainment of comorbidities and opioid use prior to cancer diagnosis. We restricted analyses to survivors who were opioid naïve at diagnosis, defined as the absence of chronic opioid use (≥90 consecutive days of opioid claims) in the year prior to HNC diagnosis. We did not exclude patients who had more limited use prior to diagnosis, in order to include in our cohort any patients who had short-term opioid use for procedures that may be cancer-related (eg, a diagnostic procedure) or unrelated to cancer (eg, a dental procedure). Survivors were required to have complete coverage in Fee-for-Service Medicare parts A, B, and D (covering inpatient care, outpatient services, and prescriptions) from 1 year prior to diagnosis through the end of 2016 or death, whichever occurred sooner. We excluded survivors diagnosed with stage 0 or IVB (distant metastatic disease) or diagnosed at death, in order to minimize counting opioid use at end of life. Patients in the non-cancer sample were eligible to be considered as cancer-free controls if they met the same coverage criteria as survivors.

2.4 |. Matching

Survivors were matched at the date of diagnosis (index date) without replacement to three controls on sex, race (white, black, other), Charlson comorbidity score (0, 1, 2 +; excluding malignancies) in the year prior to the index date, region (Northeast, West, South, Midwest), and year of birth. Exact matching of each survivor to three controls was required for all characteristics except for year of birth (±2 years).

2.5 |. Analyses

Analyses were restricted to pairs with ≥90 days of follow-up (n = 1708 excluded). The presence of at least one episode of chronic opioid use was described in each year after the index date. Pairs were censored if either member died or entered hospice. If 90 days of chronic use spanned two analysis years, the episode was attributed to the second year.

We computed the odds ratios (OR) and 95% confidence intervals (CI) for chronic opioid use for survivors compared to matched controls (a) for each year of diagnosis (2008–2015) and (b) for each year after the index date (eg, post-diagnosis year 1, 2, etc.), using a hierarchical logistic regression model accounting for matching and repeated measurements across years.

2.6 |. Sensitivity analyses

Because COT may be an appropriate palliative approach at the end of life, the analyses were repeated with censoring opioid use 6, 9, and 12 months prior to deaths attributed to cancer.

Analyses were performed in SAS v9.4 (Cary, North Carolina), using the GLIMMIX procedure for hierarchical modeling.

3 |. RESULTS

Prior to matching, 8.1% of HNC survivors (N = 495) received COT in the year prior to diagnosis and were excluded from analysis. The analytic cohort included 5107 patients with HNC (Table 1) and 15 095 matched individuals without cancer. By the seventh year after diagnosis, only 354 HNC cases remained eligible for analysis (alive and not in hospice); therefore, analyses were performed for the first 6 years after the diagnosis date.

TABLE 1.

Cohort characteristics of patients with cancer (N = 5107 cases)

Characteristics	%
Head and neck site
Lip	4.9
Tongue	22.1
Salivary gland	10.4
Floor of mouth	2.9
Gum and other mouth	12.6
Nasopharynx	2.4
Tonsil	8.4
Oropharynx	2.2
Hypopharynx	3.9
Larynx	26.4
Nose, nasal cavity and middle ear	3.8
Year of diagnosis
2008	9.2
2009	10.0
2010	10.3
2011	11.1
2012	12.3
2013	13.8
2014	16.1
2015	17.1
Age (years) at diagnosis
66–69	24.2
70–74	27.7
75–79	20.2
80–84	13.7
85+	14.2
Stage
Stage I	29.4
Stage II	17.1
Stage III	17.2
Stage IV	36.3
Sex
Female	34.1
Male	65.9
Race
White	84.2
Black	6.9
Other	8.9
Charlson comorbidity index (year pre-dx)
0	48.0
1	24.8
2+	27.2
Geographic region
West	41.1
Northeast	20.3
Midwest	12.4
South	26.1

Notes: Cancer-free controls were matched on sex, race, region, comorbidity in the year prior to diagnosis, and birth year ±2 years. Therefore, percentages of controls with these characteristics are identical to percentages of HNC cases with these characteristics.

3.1 |. COT rates by diagnosis year

For cohorts of HNC survivors diagnosed in each year from 2008 to 2015, rates of COT ranged from 1.3% in 2008 to 2.9% in 2015 among cancer free controls, and from 4.4% in 2008 to 6.1% in 2015 among HNC survivors. The rates of COT among were statistically significantly higher among survivors than controls for each diagnosis year, with odds ratios declining from 3.85 (95% confidence interval [CI] 2.62, 5.64) for survivors diagnosed in 2008 and their matched controls to 1.56 (95% CI 1.32, 1.84) for survivors diagnosed in 2015 and their controls (all P-values <.05).

3.2 |. COT rates by year of follow-up

Each year after the index date, rates of COT among cancer-free controls increased as the cohort aged, starting from 1.4% in year 1 through 4.3% in year 6. Among HNC survivors, in the first year after diagnosis, 5.9% received COT. The rate of COT use increased to 6.1% in years 2 and 3 and dropped each subsequent year. For the first 5 years after the index date, the rates of COT among HNC survivors exceeded that of cancer-free controls, with the difference between survivors and controls declining each year after matching (OR 4.36, 95% CI 3.75, 5.07 for year 1; OR 2.60, 95% CI 2.23, 3.04 for year 2; OR 2.18, 95% CI 1.81, 2.63 for year 3; OR 1.85, 95% CI 1.47, 2.32 for year 4; and OR 1.35, 95% CI 1.03, 1.77 for year 5, all P-values <.05) (Figure 1). By year 6, the rates of COT did not differ significantly between HNC cases and controls (4.3% vs 4.5%).

FIGURE 1.

Chronic opioid therapy use by year after index date among head and neck cancer survivors and cancer-free controls [Color figure can be viewed at wileyonlinelibrary.com]

3.3 |. Sensitivity analyses

After censoring follow-up at 6, 9, and 12 months prior to cancer-attributed deaths, patterns of COT use were similar in magnitude and significance, although attenuated (Table S2).

4 |. DISCUSSION

In a large population-based study, we found that older survivors of HNC have significantly higher rates of COT than matched cancer-free controls for 5 years after cancer diagnosis. The greatest differences in COT use between survivors and controls was seen in the first year after diagnosis, with HNC survivors having 4.36 times the odds of COT compared to controls in the first year. The differences between survivors and controls subsequently declined each year.

In the first 6 years after diagnosis, the rates of COT among opioid-naïve HNC survivors ranged from 4.5% to 6.1%. These rates are lower than have been seen in single-institution studies following HNC patients receiving either radiation therapy or undergoing surgery (7%−15%).^23,24 Our findings likely reflect the heterogeneous histology in this nationwide registry, because patients with tumors at different sites have differing distributions of age, treatment modalities, and pain burden—all of which, in turn, may be associated with opioid use. The lower opioid rates in our study may reflect the restriction of our analysis to opioid-naïve, older survivors. The COT rates in our analysis are comparable to COT rates among opioid-naïve, older lung cancer survivors in SEER-Medicare, which ranged from 5.3% to 7.7% in the 6 years after diagnosis.²⁵ This similarity is unsurprising, given that both lung and HNC survivors populations experience a high burden of pain and prevalent risk factors for dependence.

The decline in difference in COT use between survivors in controls over time since diagnosis was driven by rising rates of COT among controls, rather than declining use of COT among HNC survivors. In fact, the elevated rates of use of COT among HNC survivors declined minimally in the years following diagnosis. This slight decline in rates may be due to reduced cancer-related pain over time since treatment completion, as well as longer survival for those with earlier stage disease. During this period of minimal decline in rates of COT use among HNC survivors, rates of COT use among matched controls increased over time, presumably due to aging and increasing pain from comorbidities in the general population.

The greater rates of COT among survivors compared to controls may be viewed as cancer-associated. Cancer-associated pain in HNC survivors is multifactorial, arising from the tumor itself, surgery, chemotherapy, radiation, or any combination of these. Comorbidities, which are common among HNC survivors, may contribute to or even exacerbate pain; although we matched survivors to controls on a claims-based measure of comorbidity, HNC survivors in our analysis may still experience more comorbid conditions and associated pain than their counterparts. Furthermore, pain arising from recurrent disease may be addressed with opioids; although we attempted to limit our analysis to disease-free survivors, excluding patients prior to hospice or death does not completely exclude patients with recurrent disease. However, the extent to which COT is appropriate pain management is unknown for HNC survivors. COT is associated with adverse outcomes, such as addiction, overdose, and fractures.³⁰ The evidence for the effectiveness of long-term opioid use is mixed, but professional guidelines advocate for the cautious use of opioids, in conjunction with other approaches, when other pain management approaches are insufficient.^10,31–33

Our study has limitations. COT among HNC survivors may be an appropriate response to manage chronic cancer-related pain, or it may reflect excessive use of opioids resulting from poorly coordinated survivorship care; neither scenario can be identified using claims data. We could not identify HNC recurrence in these data, and the higher rates of COT among HNC survivors may therefore reflect pain management for recurrent disease. However, when follow-up was censored prior to cancer-related death in sensitivity analyses, similar associations remained, suggesting that cancer-associated COT use among survivors is not driven by end-of-life care. Additionally, the HNC survivors in our older population are more likely to be associated with tobacco use than are younger survivors, many of whom have HPV-driven oropharyngeal disease. Therefore, these data may not accurately reflect rates of COT in younger HNC survivors. Comparing COT use between subgroups of HNC survivors (eg, by stage or age) is beyond the scope of this matched analysis; instead, an ongoing case-only analysis of HNC survivors addresses patient characteristics associated with COT use.

In our population-based study of older HNC survivors, cancer-associated opioid use in the first years after diagnosis should be viewed as a cautious reminder that while opioid use may be warranted, the risk for harms must be balanced among this population of survivors who may expect to live for years after treatment completion. Future research should quantify the actual risk of harms among HNC survivors using COT.