A harm reduction model to assess the impact of new treatments for pain over standard of care among patients with osteoarthritis

BACKGROUND: Osteoarthritis (OA) affects millions of adults in the United States and can result in substantial pain, functional impairment, and significant clinical and economic burden. To manage chronic pain associated with OA, treatment guidelines recommend a variety of pharmacologic treatments, including traditional oral nonsteroidal anti-inflammatory drugs (NSAIDs), cyclooxygenase-2 inhibitors (COX-2s), and opioids. While these drug treatments can be effective at pain management, they are also associated with significant clinical and economic burden. New treatments for chronic pain among patients with OA of the hip and/or knee have the potential to reduce the occurrence of such negative clinical outcomes, including cardiovascular events, renal events, and opioid use disorder (OUD), thereby reducing health care resource use (HRU) and medical costs. OBJECTIVE: To develop a harm reduction model (HRM) to assess potential reductions of negative clinical outcomes, HRU, and medical costs associated with the use of new treatments in place of oral NSAIDs, tramadol, and non-tramadol opioids among patients with OA of the hip and/or knee in the United States. METHODS: The HRM model integrated findings from the literature and inputs from a variety of sources, along with assumptions regarding potential ability of new treatments to replace existing treatments and market penetration into a unified framework to estimate outcomes and costs. The model outputs included estimated per-patient and population-level reductions in negative clinical outcomes associated with prescribing new treatments in place of oral NSAIDs or opioids along with number needed to treat (NNT) associated with new treatments. The model assumed that new treatments will primarily be used in place of non-tramadol opioids, but more modest adoption in place of oral NSAIDs and tramadol. RESULTS: Among patients with OA of the hip and/or knee who were prescribed oral NSAIDs, tramadol, or non-tramadol opioids for chronic use (≥ 90 days), the HRM estimated total cost savings of $3.8 billion, $5.1 billion, and $29.9 billion, respectively, from prescribing new treatments for OA pain over a 36-month period. The reduced economic burden was driven by significant reductions in the incidence of negative clinical outcomes. Estimates of the NNT to avoid a negative clinical event related to NSAID and opioid treatment initiation were low for most outcomes. Estimates of NNT associated with NSAID use ranged from 4 to 17 patients, depending on outcome, and estimates of NNT associated with opioid use was 35 non-tramadol and 134 tramadol patients for OUD and ranged from 6 to 21 patients for the other clinical outcomes, depending on treatment and outcome. CONCLUSIONS: Results from the HRM suggest that prescribing new treatments in place of oral NSAIDs and/or opioids for OA pain results in a potentially substantial reduction in patients experiencing negative clinical outcomes and reductions in all-cause HRU and costs.


METHODS:
The HRM model integrated findings from the literature and inputs from a variety of sources, along with assumptions regarding potential ability of new treatments to replace existing treatments and market penetration into a unified framework to estimate outcomes and costs. The model outputs included estimated per-patient and population-level reductions in negative clinical outcomes associated with prescribing new treatments in place of oral NSAIDs or opioids along with number needed to treat (NNT) associated with new treatments. The model assumed that new treatments will primarily be used in place of non-tramadol opioids, but more modest adoption in place of oral NSAIDs and tramadol.

RESULTS:
Among patients with OA of the hip and/or knee who were prescribed oral NSAIDs, tramadol, or non-tramadol opioids What is already known about this subject • Nonsteroidal anti-inflammatory drugs (NSAIDs) and opioids are often recommended to manage pain among patients with osteoarthritis (OA) of hip and/or knee.
• Chronic use of NSAIDs and opioids are associated with negative clinical outcomes and increased health care resource use and medical costs.
• Negative clinical outcomes associated with NSAID use include gastrointestinal issues, renal toxicity, and cardiovascular events; negative clinical outcomes associated with opioids include opioid use disorder, constipation, nausea, fatigue, falls, and fractures.

What this study adds
• This study developed a harm reduction model that used findings/inputs from a variety of sources for an integrated analytical framework, with the aim to assess potential reductions of negative clinical and economic outcomes associated with use of new treatment alternatives in place of NSAIDs and/or opioids among patients with OA of the hip and/or knee, in the United States.
• The study results may help researchers, health care practitioners, and payers gain a better understanding of the potential to reduce harms associated with NSAID and opioid use by using new drug treatments for OA pain of the hip and/or knee.
Osteoarthritis (OA) is described as a disease that affects 1 or more whole joints, involving structural alterations in the articular cartilage, subchondral bone, ligaments, capsule, synovial membrane, and periarticular muscles. 1,2 Consequently, OA is associated with substantial pain and functional impairment that significantly decreases quality of life. 3 OA is estimated to affect 32.5 million adults in the United States, and although it can affect several types of joints, hip and knee are among the most affected joints. 4 To manage pain associated with OA, treatment strategies routinely include a variety of physical, physiological, and mind-body approaches, including exercise, weight loss, and assistive devices, as well as pharmacologic treatments. 1 For example, among the pharmacologic therapies listed by the American College of Rheumatology as strongly and conditionally recommended for the management of OA are oral nonsteroidal anti-inflammatory drugs (NSAIDs) and tramadol, while non-tramadol opioids are not recommended. 5,6 Similarly, the European Society for Clinical and Economic Aspects of Osteoporosis, Osteoarthritis and Musculoskeletal Diseases recommends the use of traditional oral NSAIDs, cyclooxygenase-2 inhibitors (COX-2s), and opioids (including tramadol) as potential treatment approaches. 7 The Osteoarthritis Research Society International lists traditional oral NSAIDs and COX-2s among its treatment recommendations; however, use of tramadol or other opioids is not recommended. 7 Although these treatments can be effective at managing pain, they are associated with known negative clinical outcomes. Use of traditional oral NSAIDs, in particular, is associated with increased risk of numerous negative clinical outcomes, including gastrointestinal issues such as ulcers and bleeding; renal toxicity; and cardiovascular events, such as heart failure, myocardial infarction, and stroke. 8-10 The use of COX-2s has also been associated with increased rates of cardiovascular complications compared with traditional oral NSAIDs. 9 In addition, there are numerous studies that have demonstrated that long-term use of opioids for the management of OA pain may lead to increases in negative clinical outcomes. These negative clinical outcomes include an increased risk of opioid use disorder (OUD), constipation, fatigue, falls, fractures, and nausea, which have all been shown to result in significant humanistic and economic burden. [11][12][13] Given the clinical and economic burden associated with existing pain treatments for OA, there are new drug treatments currently in development that are anticipated to be alternatives to existing treatments. For example, on the short-term horizon, 2 anti-nerve growth factor antibodies, tanezumab and fasinumab, are in active development. [14][15][16] Also, several clinical trials are testing other mechanisms of action for the treatment of chronic pain, including cytokine inhibition, selective opioid receptor agonists, long-acting bisphosphonates (ie, zoledronate), and intraarticular capsaicin. 17 Although there is uncertainty about the future landscape of available treatments for chronic pain management among patients with OA, it is hoped that forthcoming new treatment alternatives may lead to clinical benefits relative to the current standard of care. Because regulatory approval of any new drug treatment requires demonstration of improved safety and efficacy relative to existing treatments, it is anticipated that forthcoming treatment alternatives will lead to clinical benefits and harm reduction over the current standard of care.
The objective of this research was to develop a harm reduction model (HRM) to assess the impact of potential reductions in negative clinical outcomes associated with NSAID (hereafter refers to traditional oral NSAIDs and COX-2s) and opioid (hereafter refers to tramadol and nontramadol opioids) use among patients with OA of the hip and/or knee in the United States. The model integrates the latest available data and findings from a variety of sources, along with assumptions regarding the ability of new treatments to replace ("replacement rate") existing treatments and market penetration into a unified framework to estimate potential harm reduction associated with the introduction of new treatments in place of NSAIDs and/ or opioids.
Given these inputs and assumptions, the model outputs included estimated per-patient and population-level reductions in negative clinical outcomes associated with for chronic use (≥ 90 days), the HRM estimated total cost savings of $3.8 billion, $5.1 billion, and $29.9 billion, respectively, from prescribing new treatments for OA pain over a 36-month period. The reduced economic burden was driven by significant reductions in the incidence of negative clinical outcomes. Estimates of the NNT to avoid a negative clinical event related to NSAID and opioid treatment initiation were low for most outcomes. Estimates of NNT associated with NSAID use ranged from 4 to 17 patients, depending on outcome, and estimates of NNT associated with opioid use was 35 nontramadol and 134 tramadol patients for OUD and ranged from 6 to 21 patients for the other clinical outcomes, depending on treatment and outcome.

CONCLUSIONS:
Results from the HRM suggest that prescribing new treatments in place of oral NSAIDs and/or opioids for OA pain results in a potentially substantial reduction in patients experiencing negative clinical outcomes and reductions in all-cause HRU and costs.
where adoption of a hypothetical new treatment (ie, the rate of use of the new treatment in place of current treatment alternatives) for moderate to severe OA pain was 75% among patients using non-tramadol opioids, 25% among patients using tramadol, and 5% among patients using NSAIDs. This assumption was based on the 2019 American College of Rheumatology guidelines, which strongly recommend oral NSAIDs, conditionally recommend tramadol, and conditionally recommend against non-tramadol opioids to patients with OA of the hip and/or knee. 5,6 The ability of new treatments to replace ("replacement rate") existing treatments (ie, NSAIDs, tramadol, and non-tramadol opioids) was assumed to be 75%; this assumption is consistent with recent clinical trial evidence of one potential new treatment. 21 Thus, the negative clinical outcomes, health care resource use (HRU), and costs associated with existing treatments were avoided only for those patients who remained adherent to the new treatments so that the use of current pharmacological alternatives was not needed. Possible adverse events or other associated harms possible from new treatments were not explicitly considered outside of inclusion in the replacement rate, given that rates from specific products were not available.
Negative Clinical Outcomes, All-Cause HRU, and Medical Costs Associated With NSAIDs and Opioids. Outcomes associated with NSAIDs and opioids were based on a Silverman et al study, which examined outcomes among patients with OA of the hip and/or knee in the United States. 20 Specifically, this study provided recent data on a robust set of outcomes that included rates of negative clinical outcomes and per-patient rates of all-cause HRU and medical costs, separately, for patients using NSAIDs, tramadol, and non-tramadol opioids over a 36-month period. For NSAIDs, clinical outcomes included gastroinstestinal issues, renal toxicity, and cardiovascular events. For patients using tramadol and non-tramadol opioids, clinical outcomes measured included the rates of OUD, falls, fractures, nausea, constipation, and fatigue. HRU and medical costs in these studies were broken out into categories such as inpatient, emergency department, outpatient/physician office, prescription drug, and other medical services (costs in 2017 US dollars, as reported in the literature).

MODEL OUTPUTS
The HRM presented 3 categories of outputs: (1) estimated per-patient reductions in negative clinical outcomes, HRU, and costs; (2) population-level estimated reductions in negative clinical outcomes, HRU, and costs; and (3) the number needed to treat (NNT). For the per-patient estimates, the model reflected the reduction in negative outcomes prescribing new treatments in place of NSAIDs or opioids.
The results of this model may help researchers, health care practitioners, and payers better understand the potential for reducing harms associated with NSAIDs and opioids by using new pharmacological treatments, once they are available.

Methods
The model framework was adapted from a previously published HRM that was developed to assess the potential reduction in rates of negative outcomes associated with existing patient treatments by prescribing an alternative treatment, including treatments not yet approved by the US Food and Drug Administration (FDA). 18 For this study, the Microsoft Excel-based model relied on numerous inputs/ findings that are detailed in the next section. The inputs were based on existing relevant literature where possible and on clinical assumptions when relevant literature was unavailable or where future projections were necessary.
These clinical assumptions primarily involved the potential side effect profile of new treatments relative to NSAIDs and opioids and the replacement and adoption rates of the new treatments. Accordingly, the results of the HRM relied on clinical assumptions that could not be validated until a new treatment has been approved by the FDA and is available on the market. The HRM did not account for potential other unwanted outcomes (eg, differential side effect profile) that may be associated with a new treatment; therefore, the model reflected harm reduction potential from decreased use of NSAIDs and opioids rather than actual harm.

MODEL INPUTS
OA Population Size. The estimated relevant OA patient population size in the United States was based on Zhao et al, which provided a national estimate of 25.6 million (an overall prevalence of 10.5% of US noninstitutionalized adults). 13 This estimate of the total population with OA was scaled by 54.6% to reflect the percentage of US patients with moderate to severe OA pain expected to be prescribed drug treatments for OA-related pain. 19 New Treatment Adoption and Ability to Replace Existing Treatments. Data from a recent publication on the first-line use of NSAIDs and opioids among patients with OA of the hip and/or knee was used to determine the shares of patients on each treatment type (ie, NSAIDs, tramadol, and nontramadol opioids) 20 -55.0% were prescribed only NSAIDs; 32.4% were prescribed only opioids; and 12.6% were prescribed NSAIDs and opioids. The model assumed a scenario costs, summing across inpatient, outpatient/physician office, emergency department, other medical services, and prescription drug costs. The NNT estimates took the per-patient reductions to illustrate outcomes in terms of the number of patients that the new treatment would need to treat in order to eliminate each negative clinical event (eg, 1 fewer OUD outcome or hospitalization).

SENSITIVITY ANALYSES
To test the effect of adjusting underlying model assumptions, we estimated a scenario where we adjusted replacement and adoption rates from the base scenario just described. The replacement rate was assumed to be lower, at 50% rather than 75%, but the adoption rate of the new treatment was increased among patients currently receiving NSAIDs (10% vs 5%) and tramadol (35% vs 25%). These changes were applied in recognition of the considerable uncertainty regarding the hypothetical new treatment's replacement and adoption rates. For example, a new treatment with comparatively low efficacy but without known serious side effects, such as nonpharmacological shockwave therapies being investigated for treatment of OA-related pain, 22,23 represent 1 category of possible new treatments that may be tried by a larger number of patients (eg, due to limited risk of adverse events) but also more frequently discontinued (eg, due to failure to provide adequate analgesia).

CLINICAL OUTCOMES
Estimated Per-Patient Reduction in Clinical Outcomes. Table 1 presents the HRM results at the per-patient level, estimating for each patient the harm reduction from the new treatment being prescribed over NSAIDs The national projection multiplied the per-patient estimates by the number of patients prescribed and adhering to the new treatment. These estimates also presented aggregate total medical per-patient receiving new treatments in place of NSAIDs or opioids. The perpatient estimates relied on neither the estimated OA population size nor the assumed treatment adoption rates.  Consistent with per-patient estimates, population-level estimates of harm reduction based on an assumed 25% adoption rate of the new treatment over tramadol use were smaller than the reductions observed for non-tramadol opioid use but remained substantial: the population-level estimates suggested a reduction of 2,123 OUD events, 21,442 constipation events, 50,102 fatigue events, 30,571 nausea events, 13,799 fewer falls, and 25,900 fewer fractures.
Number Needed to Treat Results. Estimates of harm reduction in terms of NNT are presented in the final column of Table 1. Estimates of the NNT to avoid a negative clinical event related to NSAID treatment initiation were low. The HRM estimated that 17 patients would need to be treated with the new treatment in order to avoid 1 negative gastrointestinal outcome; 4 patients would need to be treated with the new treatment to avoid a negative cardiovascular outcome; and 16 patients would need to be treated to avoid a renal toxicity outcome.
The NNT for a new treatment to avoid OUDs differed substantially for the tramadol and non-tramadol cohorts (35 non-tramadol patients vs 134 tramadol patients). However, NNT estimates for the new treatment to avoid other opioid-related events, including constipation, fatigue, nausea, falls, or fractures, were comparable across the 2 opioid cohorts. NNT estimates for these other opioidrelated events were also lower than for those estimated to avoid OUDs for each cohort (NNT ranging from 6 patients to 21 patients across treatment cohort and specific outcome).

HRU AND MEDICAL COST OUTCOMES
The HRM estimated substantial reductions in inpatient, outpatient, and other visits per patient associated with use of new treatments. Limited effect on emergency department use was observed in per-patient and population-level estimates. Figure 1 presents estimated population-level reductions in all-cause HRU by setting and treatment cohort.
Per-patient estimates of reductions in all-cause costs vary by type of cost and treatment cohort. Overall, the HRM estimated cost savings to payers following adoption of a new treatment of $7,758 per patient previously using NSAIDs, $13,643 per patient using non-tramadol opioids, and $18,010 per patient previously using tramadol opioids. Figure 2 presents estimated population-level all-cause cost savings for each treatment cohort in total and by cost type over the 36-month period. In total, across the NSAID, tramadol, and non-tramadol initiating populations of patients with OA, the HRM calculated total medical cost savings of $3.8 billion, $5.1 billion, and $29.9 billion, or opioids over a 36-month period. When used in place of NSAIDs, the model estimated harm reduction in terms of gastrointestinal outcomes (4.4% risk reduction), cardiovascular outcomes (22.5% risk reduction), and renal toxicity outcomes (4.9% risk reduction). When used in place of opioids, the per-patient estimated reduction in risk of OUD was 0.8% and 2.9% when used in place of tramadol and non-tramadol opioids, respectively. Reductions in rates of constipation, fatigue, and nausea ranged from 7.6% to 19.9% per patient, depending on the opioid treatment cohort and the specific outcome. Per-patient estimates of reductions for falls (4.9% and 5.1%, respectively, for tramadol and nontramadol users) and fractures (9.2% and 10.2%, respectively, for tramadol and non-tramadol users) were also substantial.

Population-Level Estimated Reduction in Clinical
Outcomes. Extrapolated per-patient estimates of reductions in negative clinical outcomes to population-level estimates of clinical events avoided are also reported in Table 1. Under the scenario of 5% adoption rate of the new treatment among current users of NSAIDs, estimates suggested reductions of approximately 29,276 fewer gastrointestinal outcomes, 148,861 fewer cardiovascular outcomes, and 32,253 fewer renal toxicity outcomes. An assumed scenario of 75% adoption rate of the new treatment over non-tramadol opioid use was associated with meaningful estimated reductions at the population level, including decreases in clinical outcomes related to OUDs

FIGURE 1
Population-Level Estimated Reduction in All-Cause Health Care Resource Use on literature-based estimates of the prevalence of OA along with recently published studies based on findings that measured the changes in negative clinical outcomes associated with use of NSAIDs and opioids. The underlying claims analysis in the Silverman et al study found that patients with OA of the hip and/or knee who were prescribed NSAIDs or opioids experienced significant increases in a number of negative clinical outcomes and a statistically significant increase in total all-cause HRU and medical costs in the period following initiation of treatment. 20 The HRM estimated substantially reduced negative clinical outcomes and all-cause HRU and costs based on assumptions about the potential ability of new treatments to reduce the negative clinical outcomes associated with NSAID or opioid use and the adoption of new treatments. Notably, overall estimates of economic burden estimated in the HRM were comparable to another recent study, which used the validated Osteoarthritis Policy Model (OPM) of clinical outcomes in OA of the knee. 24 The HRM base case estimated $3.8 billion, $5.1 billion, and $29.9 billion for the NSAID, tramadol, and nontramadol opioid cohorts of patients with OA of the hip and/or knee over a 36-month period. The OPM base case estimated total lifetime costs of nontramadol opioids to be $14.0 billion in a smaller patient population with OA of the knee initiated on tramadol, then prescribed non-tramadol opioids, and with a Western Ontario and McMaster Universities Osteoarthritis (WOMAC) pain index score greater than 40. 24 The comparable total magnitude of the OPM estimates for non-tramadol opioids supports the HRM's finding of substantial harm reduction possible from reduced use of non-tramadol opioids. greater reduction in negative clinical outcomes for NSAIDs compared with the base case and less of a reduction in negative clinical outcomes (due to lower replacement rate estimates) for tramadol and non-tramadol opioids ( Figure 3). Harm reduction for using the new treatment over NSAIDs, tramadol, and non-tramadol opioids was associated with population-level allcause medical costs decreasing by $5.1 billion, $4.8 billion, and $19.9 billion, respectively, compared to $3.8 billion, $5.1 billion, and $29.9 billion in the base case.

Discussion
To our knowledge, this study is among the first to provide estimates of harm reduction (ie, reduced negative clinical outcomes) and associated reductions in all-cause HRU and medical costs arising from the adoption of new treatments as an alternative to NSAIDs and opioids for the treatment of chronic pain among adult patients with OA of the hip and/or knee that lack these harms. 24 The HRM relied respectively, from reduced initiation of standard of care treatments (ie, NSAIDs, non-tramadol opioids, and tramadol) due to adoption of a new treatment (2017 US dollars).
Harm reduction for substitution of NSAIDs was associated with population-level all-cause inpatient costs decreasing by $1 billion, outpatient costs decreasing by $200 million, and drug costs decreasing by $2.5 billion. Harm reduction due to substitution of tramadol was similarly associated with population-level all-cause inpatient costs decreasing by $2.3 billion, outpatient costs decreasing by $700 million, and drug costs decreasing by $1.9 billion due to adoption of a new treatment. Population-level estimates of cost savings following substitution from use of non-tramadol opioids were largest, with all-cause inpatient costs decreasing by $3.5 billion, outpatient costs decreasing by $4.7 billion, and drug costs decreasing by $20.0 billion.
Output of sensitivity analyses reflecting lower overall replacement (ie, less adherence to the hypothetical new treatment) but increased adoption rates vs NSAIDs and tramadol led to a  These findings, consistent with the substantial evidence of clinical and economic burden of standard of care therapies for treatment of OA, [11][12][13]24 suggest several takeaways and possible areas for further research. Patients initiated on NSAIDs and opioids as part of their OA treatments may experience severe and costly negative health outcomes, particularly in the presence of comorbid conditions in an elderly population, such as cardiovascular disease, hypertension, and kidney disease. 25,26 Patients may then be switched to other treatments, such as acetaminophen, duloxetine, or intra-articular corticosteroid injections; however, such alternatives have their own associated side effects. This suggests that there is a in harm outcomes will depend on realized ability of new treatments to replace existing ones and market penetration. Input values for this assumption were selected based on limited available clinical evidence, with sensitivity analyses substantially increasing the modeled treatment discontinuation to half of all patients. The replacement rate may further be considered to encompass a broad range of individual patient outcomes: nonadherence and return to standard of care treatments because of inefficacy or other reasons, continued use of these treatments in combination with the new treatment, or a lack of harm reduction from the hypothetical new treatment because of its corresponding safety profile relative to standard of care treatments. As additional clinical trial data become available, this will further inform the validity of input assumptions used in the HRM.

Conclusions
The HRM provides health plans and other stakeholders with a tool to help quantify the potential reduction in negative clinical outcomes and all-cause HRU and costs associated with the use of new treatments as an alternative to NSAIDs and opioids. Results from the HRM suggest that prescribing new treatments over NSAIDs and/or opioids for chronic pain associated with OA results in a potentially substantial reduction in patients experiencing negative clinical outcomes and thus commercial payers realizing reductions in all-cause HRU and costs.
In addition, the HRM took payer and patient perspectives by focusing on harm reduction that may be associated with specific negative clinical outcomes and reductions in all-cause HRU and costs. However, employers, policymakers, and other stakeholders may also be interested in assessing the indirect cost offset from patients with OA substituting new treatments for NSAIDs and opioids. 13,30 Employers have a particular interest in understanding the indirect cost burden of OA, in terms of the work loss experienced by employed patients and any incremental medical costs and work loss incurred by caregivers and/or family members. 30

LIMITATIONS
This study has some limitations that need to be considered. Any actual harm reduction depends on the relative safety profile related to possible new medications. The results of the HRM rely on assumptions related to the relative side effect profile and costs of new treatments, neither of which are known until new treatments have been approved by the FDA and are available on the market.
The HRM did not account for potential other unwanted outcomes that may be associated with new treatments so could only reflect harm reduction potential from decreased use of NSAIDs and opioids rather than actual harm.
In addition, the ability of new medications to replace existing treatments will vary based on numerous factors, including safety profile, cost, and formulation (eg, whether the medication is a subcutaneous injection, a single intravenous infusion, requires repeated injections, or is a single articular injection). As such, the ability to replace the rate assumed in this study was intended to be illustrative of those eventually obtained by new treatments, but any actual reduction