If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Address correspondence and reprint requests to: M. Boeri, RTI Health Solutions, 123B Forsyth House, Cromac Square, Belfast, Northern Ireland, BT2 8LA, United Kingdom. Tel.: 44-161-447-6016.
To quantify preferences for attributes of potential analgesic treatments for moderate-to-severe pain associated with osteoarthritis (OA) and/or chronic low back pain (CLBP) as relevant to injectable nerve growth factor (NGF)–inhibitors, nonsteroidal anti-inflammatory drugs (NSAIDs), and opioids.
Methods
We used a discrete-choice experiment (DCE) to elicit preferences for attributes of OA and CLBP pharmaceutical treatments, and a best-worst scaling (BWS) exercise to further characterize the relative importance of treatment-related side-effect risks. The survey was completed online by 602 US residents with self-reported chronic, moderate-to-severe OA pain and/or CLBP who had tried, had contraindications for, or were unwilling to take currently available pharmaceutical therapies. In the DCE, respondents repeatedly chose between two hypothetical treatments defined by six attributes (symptom control; treatment-related risks of (1) severe joint problems, (2) heart attack, and (3) physical dependence; mode/frequency of administration; and cost). In the BWS exercise, respondents evaluated ten side-effect risks. Random-parameters logit models were estimated; conditional relative attribute importance, maximum acceptable risks, and willingness to pay were calculated.
Results
The most important DCE attributes were improving symptom control (scaled conditional relative importance, 10.00) and reducing risk of physical dependence (6.99). The three most important BWS attributes were, in rank order, risks of stroke, physical dependence, and heart attack. Respondents were willing to accept a > 4% treatment-related risk of severe joint problems for even modest symptom improvement.
Conclusion
A pharmaceutical treatment with a risk of severe joint problems was viewed as an acceptable alternative to other treatments with comparable efficacy but risks associated with NSAIDs or opioids.
Osteoarthritis (OA) and chronic low back pain (CLBP) are common chronic pain conditions associated with substantial humanistic and socioeconomic burden
Incidence, prevalence, costs, and impact on disability of common conditions requiring rehabilitation in the United States: stroke, spinal cord injury, traumatic brain injury, multiple sclerosis, osteoarthritis, rheumatoid arthritis, limb loss, and back pain.
Clinical Guidelines Committee of the American College of Physicians. Noninvasive treatments for acute, subacute, and chronic low back pain: a clinical practice guideline from the American College of Physicians.
. Major risks and modest efficacy associated with NSAIDs and opioids limit their usefulness, particularly for long-term use and for patients with comorbidities
. Nerve growth factor (NGF)–inhibitors belong to an emerging class of non-opioid injectable treatments for chronic pain. Each of these therapies carries different risks: whereas NGF inhibitors are associated with a risk of rapidly progressive OA and abnormal peripheral sensations, NSAIDs are associated with cardiovascular and gastrointestinal risks, among others, and opioids carry many risks, including physical dependency.
Prior research has explored patient preferences for NSAIDs and opioids as treatments for chronic OA pain and CLBP
A discrete choice experiment on preferences of patients with low back pain about non-surgical treatments: identification, refinement and selection of attributes and levels.
. However, no prior studies have evaluated preferences for features that differentiate NSAIDs, opioids, and NGF inhibitors.
The objectives of this study were to elicit preferences for attributes of pharmaceutical treatments for moderate-to-severe pain among individuals in the United States (US) with OA and/or CLBP who have tried, have contraindications for, or are unwilling to take currently available therapies. A secondary objective was to explore heterogeneity in preferences for prespecified subgroups. The study protocol was reviewed by the Innovative Medicines Initiative's (IMI's) IMI-PREFER project, a public-private collaborative research initiative, and will contribute to their recommendations to inform guidelines for industry, regulatory authorities, and health technology assessment bodies on the incorporation of patient perspectives on benefits and risks of medicinal products
A stated-preference survey, involving both a discrete-choice experiment (DCE) to elicit trade-offs respondents were willing to make among a set of attributes of pharmaceutical treatments for chronic pain (i.e., non-opioid NGF inhibitors, NSAIDs, and opioids) and a best-worst scaling (BWS) exercise to quantify the relative importance of an additional set of treatment-related risks, was administered online.
Qualitative research informed development of the survey instrument. Four focus groups were conducted with a total of 30 individuals with self-reported moderate-to-severe hip or knee OA pain or CLBP to identify the treatment features important to them. Six categories of treatment attributes emerged from the focus groups as particularly important: efficacy, medication cost (out-of-pocket or covered by insurance), tolerability, risk of physical dependence/addiction, and frequency and mode of administration.
The attributes included in the DCE, the respondent-facing descriptions of these attributes, and the range of levels for the attributes were informed primarily by the qualitative focus groups. Attributes related to risks and mode of administration were also informed by input from clinical experts in rheumatology and chronic pain, reflecting their knowledge of classes of drugs and the clinical literature regarding the characteristics of NSAIDs, opioids, and NGF inhibitors for OA pain and/or CLBP
Subcutaneous tanezumab versus NSAID for the treatment of osteoarthritis: joint safety events in arandomized, double-blind, active-controlled, 80-week, phase-3 study.
in: American College of Rheumatology (ACR)/Association of Rheumatology Professionals (ARP) Annual Meeting. November 8–13, 2019
. The ranges of levels of the risk attributes were refined by assessing respondents' willingness to trade among attributes in pretests. The survey instruments also included questions about respondents’ demographic characteristics and experiences with pain treatments. Questions from the Multidimensional Health Locus of Control (MHLC) Scale–Form C
and seven questions to assess respondent comprehension of the treatment attributes and DCE questions were included to define samples for subgroup analyses. A scope test evaluated whether respondents were attentive to the levels of the cost attribute (see Supplemental Appendix (A)).
In the DCE portion of the survey, respondents chose between alternative hypothetical treatment profiles in a series of questions. Fig. 1 presents an example DCE question. Each treatment profile was defined by attributes with varying levels, presented in Table I. The D-optimal algorithm was used to construct a fractional factorial experimental design
. The experimental design included 48 sets of two treatments split into six blocks of eight choice questions; each respondent was randomly assigned to one block, and question order was randomized to avoid ordering effects. Respondents' choices were a function of the utility of each attribute level in the profiles presented in each choice question. Statistical analysis was used to estimate respondents’ likelihood of choosing a treatment profile as a function of specific attributes and levels of that profile.
Table IAttributes and levels included in the discrete-choice experiment
Attribute
Respondent-facing attribute label
Respondent-facing attribute levels
Symptom control (patient global assessment)
Symptom control: Symptom control while you are taking the medicine
Very good (no symptoms; no limitations on normal activities)
Good (mild symptoms; no limitations on normal activities)
Fair (moderate symptoms; limitations on some normal activities)
Poor (severe symptoms; unable to carry out most normal activities)
Incremental treatment-related risk of severe, rapidly progressive joint problems requiring total joint replacement
Additional risk of severe joint problems: Additional risk each year of having joint problems that are severe enough that you would need a total joint replacement while you are taking the medicine or within 6 months of stopping the medicine. These joint problems are so severe and get bad so quickly that, if they occur, you would need to have a total joint replacement. This type of severe joint problem can occur in any of the major joints in your body when you are taking the medicine, even if you don't have pain or stiffness in that joint when you start taking the medicine.
No additional risk (0%)
5 people out of 1,000 (0.5%)
40 people out of 1,000 (4%)
Incremental treatment-related risk of heart attack
Additional risk of heart attack: Additional risk each year of a heart attack while you are taking the medicine
No additional risk (0%)
2 people out of 1,000 (0.2%)
5 people out of 1,000 (0.5%)
Treatment-related risk of physical dependence
Risk of physical dependence: Risk each year of becoming physically dependent on the medicine
No risk (0%)
50 people out of 1,000 (5%)
250 people out of 1,000 (25%)
Mode and frequency of administration
How you take the medicine
Oral pills two or more times a day
Oral pills once a day
Injection every 4 weeks (about once a month)
Injection every 8 weeks (about once every 2 months)
Variable created for scope test (not an attribute visible to respondents)
To implement the scope test, respondents were assigned to one of the two cost-value ranges in the discrete-choice experiment questions: a narrow or wide range of cost values (see Supplemental Appendix (A)). The scope test tested whether respondents evaluated a given difference in cost similarly regardless of whether that change in cost occurred within a narrow range of costs or within a wide range of costs. If the respondents passed the scope test, the impact of, for example, a difference in cost between $0 and $30 should be the same whether respondents evaluated other costs ranging from $0 to $85 or other costs ranging from $0 to $110.
Narrow range
Wide range
Personal (out-of-pocket) cost per month
Cost: Personal cost of the medicine to you every month
$0 every month
$0 every month
$30 every month
$30 every month
$55 every month
$75 every month
$85 every month
$110 every month
∗ To implement the scope test, respondents were assigned to one of the two cost-value ranges in the discrete-choice experiment questions: a narrow or wide range of cost values (see Supplemental Appendix (A)). The scope test tested whether respondents evaluated a given difference in cost similarly regardless of whether that change in cost occurred within a narrow range of costs or within a wide range of costs. If the respondents passed the scope test, the impact of, for example, a difference in cost between $0 and $30 should be the same whether respondents evaluated other costs ranging from $0 to $85 or other costs ranging from $0 to $110.
to elicit respondents' assessments of the relative importance of additional side-effect attributes associated with pharmaceutical treatments for chronic pain that could not be included in the DCE because of the limit on the number of attributes in the DCE (see Table II). Each of the ten risks presented in the BWS included both an outcome and a probability. Three risks—severe joint problems in any major joint, heart attack, and physical dependence—were included in both the DCE and the BWS; the probability for each corresponded to the highest risk level presented in the DCE. Probabilities for the remaining risks were based on physicians' knowledge of classes of drugs and the clinical literature regarding adverse-event rates
Subcutaneous tanezumab versus NSAID for the treatment of osteoarthritis: joint safety events in arandomized, double-blind, active-controlled, 80-week, phase-3 study.
in: American College of Rheumatology (ACR)/Association of Rheumatology Professionals (ARP) Annual Meeting. November 8–13, 2019
. For each BWS question, respondents selected the most and least important risk to avoid from a subset of three risks (Fig. 2). The experimental design included 90 questions split into nine blocks of ten questions; each respondent was randomly assigned to one block, and question order was randomized to avoid ordering effects. The pattern of choices over the series of questions allowed estimation of a full ranking of the respondents’ preferences for all items included in the exercise and a measure of relative importance of avoiding each risk.
Table IIAttributes included in the best-worst scaling questions
10% (100 out of 1,000) risk of moderate-to-severe constipation while taking a medicine
10% (100 out of 1,000) risk of feeling foggy and drowsy while taking a medicine
1% (10 out of 1,000) risk of having a bleeding stomach ulcer when first starting a medicine
10% (100 out of 1,000) risk of mild-to-moderate nausea and vomiting while taking a medicine
0.5% (5 out of 1,000) risk each year of having a heart attack because of a medicine
4% (40 out of 1,000) risk each year of severe joint problems because of a medicine
25% (250 out of 1,000) risk each year of becoming physically dependent on a prescription pain medicine
6% (60 out of 1,000) risk of having a tingling or burning sensation in the fingers or toes while taking a medicine
5% (50 out of 1,000) risk of mild-to-moderate swelling in the ankles and feet while taking a medicine
0.6% (6 out of 1,000) risk each year of having a moderate stroke because of a medicine
Before the survey was administered to the study sample, the survey instrument was qualitatively pretested during 15 in-person interviews with individuals meeting the study eligibility criteria. Minor revisions were made to the survey instrument based on these interviews. This process confirmed that the survey questions were comprehensible and that the attributes and levels were comprehensive, relevant, and clearly described.
Study population
The health care market research company SSI, now known as Dynata, used a purposive convenience sampling approach to invite potential US respondents from SSI's maintained and partner panels of individuals who have opted to participate in health care research. Participants were recruited to be panel members via partnerships with loyalty programs and through Internet and television advertising. Panelists were invited by email to complete the survey instruments online. Potential respondents completed a 12-item screening questionnaire to determine study eligibility.
Eligible respondents were residents of the US and were aged 18 years or older with a self-reported physician diagnosis of hip or knee OA only, CLBP only, or concurrent OA and CLBP, each diagnosed at least 3 months before the date of the survey. Eligible respondents also had to have self-reported moderate-to-severe pain in the hip, knee, and/or lower back, each defined as a self-assessed rating of ≥five on average in the past week on an 11-point numeric rating scale (NRS) ranging from 0 (no pain) to 10 (worst possible pain). For respondents with concurrent OA pain and CLBP, a self-assessed pain rating of ≥five was required for either OA pain or CLBP. Respondents were required to be taking or to have tried (1) three or more classes of pain treatment in the past 2 years; (2) two prior classes of pain treatment, either excluding NSAIDs because of NSAID contraindication or excluding opioids because of opioid contraindication or the respondent's unwillingness to take opioids; or (3) one prior class of pain treatment excluding NSAIDs because of NSAID contraindication and excluding opioids because of opioid contraindication or the respondent's unwillingness to take opioids.
Respondents were also required to be able to read and understand English and provide informed consent. Individuals were excluded if they had a self-reported diagnosis of Alzheimer's disease, axial spondyloarthritis, fibromyalgia, major depressive disorder, migraine headaches, myopathy, neuropathic pain, psoriatic arthritis, radiculopathy or sciatica, rheumatoid arthritis, spinal stenosis, spondyloarthropathy, or pain as a result of having had surgery in the past 3 months. Respondents who completed the survey received a modest compensation (equivalent to $10) for their participation.
Recruitment quotas were imposed to achieve a balanced composition of respondents, such that (1) approximately one-third had OA only, one-third had CLBP only, and one-third had concurrent OA and CLBP; and (2) approximately half were taking or had taken an opioid to treat OA and/or CLBP within the past 2 years, and approximately half had not. These quotas were sequentially filled as respondents completed the survey. Respondents who showed no variability in their DCE responses or who did not respond to at least one DCE and one BWS question did not add information to the analysis and were excluded. Data collection was discontinued once recruitment quotas were reached.
For DCE studies, the minimum sample size required depends on several criteria, including the question format, the complexity of the choice task, the desired precision of the results, and the need to conduct subgroup analyses
, which is typically impossible. Empirical data indicate that an optimal sample size for a DCE with six to eight attributes and three levels per attribute is 250–350 respondents
. Generally, sample sizes of 150 for each subgroup are considered sufficient to support the analyses. To enable recruitment of at least 150 respondents for each condition (OA, CLBP, or both), the study used an initial recruitment target of 450 respondents. An additional 152 respondents then were recruited to implement a scope test on the cost attribute (see Supplemental Appendix (A)).
The study was reviewed and deemed exempt by the RTI International institutional review board. All survey respondents provided electronic informed consent.
Statistical analyses
The DCE data were analyzed using a random-parameters logit (RPL) model following good research practices
. The RPL model yielded a set of relative preference weights for the attribute levels included in the DCE, which were used to calculate the trade-offs respondents were willing to make between benefits, risks, and cost. Specifically, the RPL results from the DCE were used to calculate the maximum acceptable risks (MARs) of incremental annual treatment-related risk of severe joint problems (a potential risk associated with NGF inhibitors), incremental annual treatment-related risk of heart attack, and treatment-related risk of physical dependence that respondents were willing to accept in exchange for improvements in symptom control. In addition, the DCE data were used to calculate marginal willingness to pay (WTP) out-of-pocket for improvements in each attribute.
As a partial internal validity test, a scope test was conducted to assess whether respondents paid attention to the actual levels of the personal cost of the medicine each month when making treatment choices in the DCE
. The scope test posited that respondents would evaluate a given difference in cost similarly regardless of whether that change in cost occurred within a narrow or wide range of costs (see Supplemental Appendix (A)). A Wald test was used to evaluate jointly whether the two cost levels ($0 and $30 every month) that appeared in both the narrow and wide cost ranges had the same estimated preference weights when the results from the subsample that saw the narrow cost range were compared with those from the subsample that saw the wide cost range
Subgroup analyses of the DCE data were conducted by estimating interaction models for mutually exclusive pairs of subgroups in the sample and testing for systematic differences in preferences between each subgroup pair. Subgroups were defined by respondent condition (OA, CLBP, or concurrent OA and CLBP); opioid experience; age; time living with chronic pain; baseline pain level; score on the internal, chance, and powerful others locus of control questions on the MHLC Scale–Form C (Table (B)-1 in Supplemental Appendix (B)); comprehension questions; and experience with joint replacement. Table (C)-1 (Supplemental Appendix (C)) presents the parameters used for continuous and categorical subgroup variables. A chi-square test of the joint significance of the interaction terms indicated whether preferences between the groups were statistically significantly systematically different
All analyses were conducted using STATA 15 (Stata Corp, College Station, Texas).
Results
Respondent characteristics
To recruit 602 respondents, 316,241 potential panelists were invited; 25,352 (8.0%) accessed the link, 2,924 (11.5%) indicated that they met the eligibility criteria, and 2,558 (87.5%) consented to participate. A total of 209 respondents (8.2%) were not administered the survey because they exceeded a recruitment quota, 1,716 (73.1%) were excluded from the final sample because they did not show any variability in their answers to DCE or did not answer at least one DCE question, and 31 (1.3%) were excluded because they exceeded the opioid experience recruitment quota.
Respondents had a mean age of 64 years; 59% were female, and 94% were white (Table III). Thirty-two percent were employed full-time or part-time or were self-employed. The sample was, by design, balanced among respondents with OA (n = 201), with CLBP (n = 202), and with both OA and CLBP (n = 199). Respondents with only OA had a mean (standard deviation [SD]) rating of OA-related pain of 6.6 (1.2) on a 0-to-10 NRS. Respondents with only CLBP had a mean (SD) rating of CLBP of 6.6 (1.2), and those with both OA and CLBP had mean (SD) ratings of 6.1 (1.8) for OA-related pain and 6.6 (1.6) for CLBP.
Respondents could provide multiple responses to these questions. Therefore, the totals may exceed the total number of respondents.
White
189 (94.0)
188 (93.1)
189 (95.0)
566 (94.0)
Hispanic or Latino
6 (3.0)
4 (2.0)
2 (1.0)
12 (2.0)
Black or African American
5 (2.5)
7 (3.5)
10 (5.0)
22 (3.7)
Native American or American Indian
6 (3.0)
4 (2.0)
2 (1.0)
12 (2.0)
Asian/Pacific Islander
3 (1.5)
5 (2.5)
1 (0.5)
9 (1.5)
Other
2 (1.0)
1 (0.5)
0
3 (0.5)
Prefer not to say
0
0
0
0
Highest level of education, n (%)
Less than high school
0
0
0
0
Some high school
3 (1.5)
2 (1.0)
0
5 (0.8)
High school or equivalent (e.g., GED)
27 (13.4)
26 (12.9)
32 (16.1)
85 (14.1)
Some college but no degree
43 (21.4)
41 (20.3)
49 (24.6)
133 (22.1)
Technical school
7 (3.5)
21 (10.4)
13 (6.5)
41 (6.8)
Associate's degree (2-year college degree)
26 (12.9)
25 (12.4)
24 (12.1)
75 (12.5)
4-year college degree (e.g., B.A., B.S.)
52 (25.9)
46 (22.8)
42 (21.1)
140 (23.3)
Some graduate school but no degree
13 (6.5)
12 (5.9)
5 (2.5)
30 (5.0)
Graduate or professional degree (e.g., M.B.A., M.S., M.D., PhD.)
30 (14.9)
29 (14.4)
34 (17.1)
93 (15.4)
Prefer not to say
0
0
0
0
Employment status, n (%)
Employed full-time
25 (12.4)
56 (27.7)
31 (15.6)
112 (18.6)
Employed part-time
17 (8.5)
17 (8.4)
12 (6.0)
46 (7.6)
Self-employed
8 (4.0)
19 (9.4)
8 (4.0)
35 (5.8)
Homemaker
6 (3.0)
13 (6.4)
8 (4.0)
27 (4.5)
Student
2 (1.0)
0
1 (0.5)
3 (0.5)
Retired
126 (62.7)
84 (41.6)
117 (58.8)
327 (54.3)
Unable to work because of your [OA and/or CLBP]
5 (2.5)
0
6 (3.0)
11 (1.8)
Unable to work because of disability or other health problem
11 (5.5)
11 (5.4)
13 (6.5)
35 (5.8)
Unemployed and looking for work
0
2 (1.0)
3 (1.5)
5 (0.8)
Unemployed and not looking for work
1 (0.5)
0
0
1 (0.2)
Prefer not to say
0
0
0
0
Experiences with osteoarthritis and/or chronic low back pain and related treatments
[If diagnosed with OA] How would you rate the level of your pain caused by your OA in your knee(s) or hip(s) on average over the past week on a scale from 0 (no pain) to 10 (worst pain you can imagine)?
Respondents with OA and CLBP were eligible for the survey if their average pain level in the past week for OA or CLBP was ≥five on an 11-point numeric pain scale ranging from 0 (no pain) to 10 (worst possible pain). Thus, the minimum pain level shown for respondents with OA and CLBP may be < five based on the eligibility criteria. The minimum pain level selected by respondents with OA and CLBP is also reflected in the full sample summary results.
N
201
–
199
400
Mean pain level (SD)
6.6 (1.2)
–
6.1 (1.8)
6.4 (1.6)
Median pain level
7
–
7
7
Min pain level
5
–
0
0
Max pain level
10
–
10
10
[If diagnosed with CLBP] How would you rate the level of your pain caused by your CLBP on average over the past week on a scale from 0 (no pain) to 10 (worst pain you can imagine)?
Respondents with OA and CLBP were eligible for the survey if their average pain level in the past week for OA or CLBP was ≥five on an 11-point numeric pain scale ranging from 0 (no pain) to 10 (worst possible pain). Thus, the minimum pain level shown for respondents with OA and CLBP may be < five based on the eligibility criteria. The minimum pain level selected by respondents with OA and CLBP is also reflected in the full sample summary results.
N
–
202
199
401
Mean pain level (SD)
–
6.6 (1.2)
6.6 (1.6)
6.6 (1.4)
Median pain level
–
6.0
7.0
7.0
Min pain level
–
5
1
1
Max pain level
–
10
10
10
Respondent eligibility based on classes of pain treatment, n (%)
Respondents who met the minimum of three current or prior medications
173 (86.1)
175 (86.6)
162 (81.4)
510 (84.7)
Respondents who selected two current or prior medications but could not take NSAIDs
16 (8.0)
3 (1.5)
13 (6.5)
32 (5.3)
Respondents who selected two current or prior medications but could not take opioids or would never consider taking opioids
9 (4.5)
18 (8.9)
17 (8.5)
44 (7.3)
Selected two current or prior medications but could not take NSAIDs and could not take opioids or would never consider taking opioids
1 (0.5)
1 (0.5)
2 (1.0)
4 (0.7)
Respondents who selected one current or prior medication but could not take NSAIDs and could not take opioids or would never consider taking opioids
2 (1.0)
5 (2.5)
5 (2.5)
12 (2.0)
How satisfied are you with how well your current treatment works to control your [OA] [CLBP] [OA and CLBP] pain?
While this question was included in the survey to ensure that people who were satisfied with their current treatment and people who were dissatisfied with their current treatment were represented in the respondent sample, the question was not included as a stratification variable.
(Please check only one answer), n (%)
Very satisfied
12 (6.0)
13 (6.4)
10 (5.0)
35 (5.8)
Somewhat satisfied
79 (39.3)
83 (41.1)
75 (37.7)
237 (39.4)
Neither satisfied nor dissatisfied
43 (21.4)
64 (31.7)
49 (24.6)
156 (25.9)
Somewhat dissatisfied
46 (22.9)
32 (15.8)
47 (23.6)
125 (20.8)
Very dissatisfied
21 (10.4)
10 (5.0)
18 (9.0)
49 (8.1)
CLBP = chronic low back pain; ER = extended release; max = maximum; min = minimum; NSAID = nonsteroidal anti-inflammatory drug; OA = osteoarthritis; SD = standard deviation.
Note: The percentage totals may not sum exactly to 100% because of rounding.
∗ Respondents could provide multiple responses to these questions. Therefore, the totals may exceed the total number of respondents.
† Respondents with OA and CLBP were eligible for the survey if their average pain level in the past week for OA or CLBP was ≥five on an 11-point numeric pain scale ranging from 0 (no pain) to 10 (worst possible pain). Thus, the minimum pain level shown for respondents with OA and CLBP may be < five based on the eligibility criteria. The minimum pain level selected by respondents with OA and CLBP is also reflected in the full sample summary results.
‡ While this question was included in the survey to ensure that people who were satisfied with their current treatment and people who were dissatisfied with their current treatment were represented in the respondent sample, the question was not included as a stratification variable.
Estimated preference weights for the levels of symptom control, incremental annual risk of heart attack, risk of physical dependence, and monthly personal (out-of-pocket) cost were ordered as expected; that is, better levels were preferred to worse levels of these attributes (Fig. 3). An improvement in pain control from poor to fair yielded a larger utility gain than an improvement from fair to good, which in turn yielded a larger utility gain than an improvement from good to very good. Respondents preferred taking an oral pill once daily to oral pills taken two or more times daily and preferred oral pills taken two or more times daily to monthly injections; however, preference weights for injections every 4 weeks did not statistically significantly differ from injections every 8 weeks.
Fig. 3Discrete-Choice Experiment Preference Weights. Note: Attributes are presented in the order in which they appeared in the DCE questions. The vertical bars around each mean preference weight represent the 95% confidence interval around the point estimate. The change in utility associated with a change in the levels of each attribute is represented by the vertical distance between the preference weights for any two levels of that attribute. Larger differences between preference weights indicate that respondents viewed the change as having a relatively greater effect on overall utility. For example, improving symptom control from “Poor” to “Fair” yielded a change in utility of approximately 2.5 ([−0.035] − [−2.531] = 2.496). Improving symptom control from “Fair” to “Good” yielded a change in utility of approximately 1.1 (1.018 − [−0.035] = 1.053), indicating that an improvement from “Fair” symptom control to “Good” symptom control was less than half as important to patients as an improvement from “Poor” to “Fair” symptom control. Likewise, decreasing the incremental annual treatment-related risk of severe rapidly progressive joint problems from 4% to 0.5% yielded a change in utility of approximately 0.6 (0.253 − [−0.350] = 0.603). Therefore, improving symptom control from “Poor” to “Fair” was preferable to decreasing the incremental annual treatment-related risk of severe rapidly progressive joint problems from 4% to 0.5% because it had approximately 4.1 times (2.496 ÷ 0.603) more impact on utility than decreasing the incremental annual treatment-related risk of severe joint problems from 4% to 0.5%. In the final model, out-of-pocket cost was interacted with respondent income to adjust for the effect of income.
For the incremental annual risk of severe joint problems, the preference weight for a 0.5% annual risk was statistically significantly greater than that for a 4% annual risk (Fig. 3). However, the estimated mean preference weight for an incremental annual risk of severe joint problems of 0.5% was slightly higher but not statistically significantly different than the estimated mean preference weight for no incremental increase in risk, potentially indicating that respondents were indifferent between these two levels.
The difference between the most-preferred and least-preferred level of each attribute in Fig. 3 captures the importance of each attribute, relative to the other attributes in the DCE, conditional on the levels of each attribute. The two most important attributes were symptom control (scaled conditional relative importance, 10.0) and treatment-related risk of physical dependence (6.99). The three least important attributes were mode and frequency of administration (2.66), incremental annual treatment-related risk of heart attack (1.64), and incremental annual treatment-related risk of severe joint problems (1.10).
Maximum acceptable risks
Table IV presents the MARs of incremental annual risk of severe joint problems, incremental annual risk of heart attack, and treatment-related risk of physical dependence that respondents would accept in exchange for different levels of improvement in symptom control. In general, respondents were willing to accept more than a 4% increase in the annual risk of severe joint problems and more than a 0.5% increase in the annual risk of heart attack for any of the improvements in pain control included in the survey. Respondents were willing to accept a treatment-related risk of physical dependence of approximately 21% for an improvement in symptoms from poor to fair control—the only MAR that did not exceed the maximum level of this risk in the survey. Respondents were willing to accept more than a 25% risk of physical dependence to improve the level of pain control from poor to either good or very good.
Table IVDiscrete-choice experiment: MARs and willingness to pay
The MAR for each risk relative to each improvement in the levels of all other attributes provides the average percentage increase in treatment-related risk that yields a level of disutility equal to the utility generated by improving a treatment from one level to the other level included in table. A reduction in risk of severe joint problems from 0% to 0.5% did not cause respondents an increase in utility (this small variation on this risk was not ordered as expected but was not significantly different from zero). Therefore, the MAR was not reported for this variation.
It is possible to estimate a specific value for the MAR above the maximum level included in the survey instrument only by making the strong assumption that the disutility of each incremental increase in risk remains constant beyond the maximum level included in the survey. Rather than making this assumption, we only state that the MAR is greater than the maximum level included in the survey, and we did not calculate the actual MAR.
Negative values of the mean or 95% CI for MARs are possible in cases in which the MAR offsets a change in another attribute that results in a decrease in utility (i.e., negative mean) or when the mean MAR is not statistically significantly different from zero at the 95% level of confidence (i.e., the CIs overlap zero).
From level
To level
Mean (%)
95% CI (%)
MAR of severe, rapidly progressive joint problems requiring total joint replacement
A reduction in the risk of severe joint problems from 0.5% to 0% did not cause respondents an increase in utility (this small variation on this risk was not ordered as expected but was not significantly different from zero). Therefore, the willingness to pay was not reported for this variation.
Negative values of the 95% CI for WTP are possible in cases in which the mean WTP is not statistically significantly different from zero dollars at the 95% level of confidence (i.e., the CIs overlap zero).
From level
To level
Mean ($)
95% CI ($)
Symptom control (patient global assessment)
Poor
Fair
64.65
55.05
74.26
Poor
Good
91.95
79.93
103.97
Poor
Very good
105.67
91.22
120.11
Incremental treatment-related risk of severe, rapidly progressive joint problems requiring total joint replacement
40 people out of 1,000 (4%)
No additional risk (0%)
11.60
5.29
17.91
40 people out of 1,000 (4%)
5 people out of 1,000 (0.5%)
15.64
9.85
21.43
Incremental treatment-related risk of heart attack
5 people out of 1,000 (0.5%)
No additional risk (0%)
17.37
11.39
23.34
5 people out of 1,000 (0.5%)
2 people out of 1,000 (0.2%)
10.11
3.81
16.41
2 people out of 1,000 (0.2%)
No additional risk (0%)
7.26
1.18
13.33
Treatment-related risk of physical dependence
250 people out of 1,000 (25%)
No additional risk (0%)
73.81
63.11
84.52
250 people out of 1,000 (25%)
50 people out of 1,000 (5%)
49.46
41.60
57.31
50 people out of 1,000 (5%)
No additional risk (0%)
24.36
17.99
30.73
Mode and frequency of administration
Oral pills two or more times a day
Oral pills once a day
11.81
4.97
18.65
Injection every 4 weeks
Oral pills two or more times a day
16.28
8.89
23.67
Injection every 8 weeks
Oral pills two or more times a day
14.25
6.25
22.24
Injection every 4 weeks
Oral pills once a day
28.09
20.15
36.02
Injection every 8 weeks
Oral pills once a day
26.05
17.42
34.69
Injection every 4 weeks
Injection every 8 weeks
2.03
−6.03
10.09
CI = confidence interval; MAR = maximum acceptable risk.
∗ The MAR for each risk relative to each improvement in the levels of all other attributes provides the average percentage increase in treatment-related risk that yields a level of disutility equal to the utility generated by improving a treatment from one level to the other level included in table. A reduction in risk of severe joint problems from 0% to 0.5% did not cause respondents an increase in utility (this small variation on this risk was not ordered as expected but was not significantly different from zero). Therefore, the MAR was not reported for this variation.
† It is possible to estimate a specific value for the MAR above the maximum level included in the survey instrument only by making the strong assumption that the disutility of each incremental increase in risk remains constant beyond the maximum level included in the survey. Rather than making this assumption, we only state that the MAR is greater than the maximum level included in the survey, and we did not calculate the actual MAR.
‡ Negative values of the mean or 95% CI for MARs are possible in cases in which the MAR offsets a change in another attribute that results in a decrease in utility (i.e., negative mean) or when the mean MAR is not statistically significantly different from zero at the 95% level of confidence (i.e., the CIs overlap zero).
§ A reduction in the risk of severe joint problems from 0.5% to 0% did not cause respondents an increase in utility (this small variation on this risk was not ordered as expected but was not significantly different from zero). Therefore, the willingness to pay was not reported for this variation.
‖ Negative values of the 95% CI for WTP are possible in cases in which the mean WTP is not statistically significantly different from zero dollars at the 95% level of confidence (i.e., the CIs overlap zero).
The largest WTP values were associated with improvements in symptom control (Table IV). Respondents were willing to pay, on average, approximately $106 per month out-of-pocket for an improvement in symptom control from poor to very good, whereas smaller improvements in symptom control from poor to good or from poor to fair resulted in WTP amounts of $92 and $65 per month, respectively. Respondents were also willing to pay approximately $74 per month to avoid a 25% risk of physical dependence and approximately $49 per month to reduce the risk of physical dependence from 25% to 5%.
Relative importance of additional risks
In BWS analyses of an expanded set of risks, the five most important risks to avoid, in decreasing order of importance, were a 0.6% increased annual risk of having a moderate stroke because of a medicine, a 25% annual risk of becoming physically dependent on a prescription pain medicine, a 0.5% increased annual risk of heart attack because of a medicine, a 4% increased annual risk of severe joint problems because of a medicine, and a 1% risk of having a bleeding stomach ulcer when first starting a medicine (Fig. 4).
Fig. 4Best-Worst Scaling Main-Effects Item Relative Importance. Note: To facilitate interpretation of relative importance estimates, the relative importance of one attribute, annual risk of stroke, was scaled to 10. The relative importance of each of the other attributes was scaled relative to the conditional importance of this attribute. The horizontal bars surrounding each relative importance weight estimate denote the 95% confidence interval (computed by the delta method).
These treatment attributes were followed, in decreasing order of importance, by a 10% risk of feeling foggy and drowsy, a 10% risk of mild-to-moderate nausea and vomiting, and a 10% risk of moderate-to-severe constipation; the least important risks to avoid were a 6% risk of having a tingling or burning sensation in the fingers or toes and a 5% risk of mild-to-moderate swelling in the ankles and feet.
Subgroup analyses
Subgroup analyses explored heterogeneity in preferences for ten prespecified subgroup pairs (Table (C)-1in Supplemental Appendix (C)). Preferences were statistically significantly systematically different only for the subgroup pair defined by the number of correctly answered comprehension questions. Respondents who answered fewer than three comprehension questions incorrectly had systematically different preferences (P < 0.05, placing greater weight on pain control than did respondents who answered three or more comprehension questions incorrectly. Nevertheless, the preference weights for the two subgroups in this pair were qualitatively very similar (see Figure C-1 in Supplemental Appendix (C)).
No statistically significant systematic differences in preferences were observed between subgroups defined by condition (OA, CLBP, or both); opioid experience; age; time living with chronic pain; baseline pain level; score on the internal, chance, and powerful others locus of control questions on the MHLC Scale–Form C; or experience with joint replacement.
Discussion
On average, respondents strongly preferred better symptom control and avoiding treatment-related risk of physical dependence. Avoiding incremental annual treatment-related risks of heart attack or severe joint problems was much less important than improving symptom control or avoiding the risk of physical dependence. Respondents placed greater value on efficacy than on convenience attributes. Specifically, respondents were willing to pay $106 out-of-pocket to achieve very good symptom control, consistent with previous research on the monetary value of OA treatment, which estimated a WTP for viscosupplement injections of up to $71 per treatment among a European population
While mode of administration was not a significant driver of preferences, daily oral administration was preferred to injections. On average, respondents were willing to accept increases in the annual risks of heart attack and severe joint problems that exceed those levels of risk believed to be associated with NSAIDs or NGF inhibitors, respectively, to gain symptomatic benefit. While respondents were generally averse to the risk of physical dependence associated with opioids, they were willing to accept some level of risk to have noticeable improvements in symptoms associated with OA and CLBP.
The BWS results indicated that the three risks included in both the DCE and the BWS exercises—physical dependence, heart attack, and severe joint problems—were among the most important risks to avoid. For the full sample, only the risk of stroke was more bothersome. That these risks were highly important to respondents (test) in the BWS when compared with other risks suggests that the DCE captured the risk attributes of greatest importance test. Furthermore, rank ordering of the risks of physical dependence, heart attack, and severe joint problems was the same in both the DCE and BWS exercise, supporting the robustness of these findings across two different methods of eliciting preferences, although the inclusion of these attributes in both exercises may have elevated their perceived importance in the BWS exercise.
In subgroup analyses, only respondents who answered fewer than three comprehension questions incorrectly had statistically significantly different preferences compared with respondents who answered three or more comprehension questions incorrectly, but preferences between the two subgroups in this pair were qualitatively very similar. We cannot definitively rule out the existence of differences between subgroups, and other methods of exploring preferences, including latent-class analysis, could reveal different preference segments in the sample. In addition, systematic differences in preferences may exist for other subgroups that were not defined a priori and thus were not included in these analyses, and other approaches to exploring preference heterogeneity may identify meaningful differences in preferences among subgroups.
Some limitations of this study must be acknowledged. The results may be subject to information bias, selection bias, and volunteer bias resulting from the panel-based recruitment method and study design. The respondent sample included a range of ages, socioeconomic status, education level, and working status but was drawn from an internet consumer panel that is not completely representative of the US general population, leading to some limitations in the generalizability of the survey results. Respondents were broadly consistent with the broader population of US individuals with OA and/or CLBP in terms of age and gender composition but included disproportionately more who identified as white or Caucasian
. Respondents may be younger and more educated than the population of patients with these conditions. Respondents' comprehension of the treatment attributes and DCE questions was generally good, and aversion to the risk of severe joint problems appears not to have differed among respondents based on their level of comprehension. However, severe joint problems represent an uncommon outcome associated with an emerging therapeutic class and may not have been as familiar test as the other risks evaluated, and this lack of familiarity may have influenced respondents’ preferences. The impact of this potential lack of familiarity on the study results is unknown.
The survey was administered online, although results from online stated-preference studies are, in general, not statistically different from those results elicited through face-to-face interviews
. In addition, all clinical characteristics, including diagnosis of OA and/or CLBP, were self-reported. Diagnosis and treatment history were not confirmed by a physician, and respondents might have had other pain conditions that they did not disclose at screening. Finally, although the study has a number of strengths derived from the use of best practices for design and analysis
Constructing experimental designs for discrete-choice experiments: report of the ISPOR conjoint analysis experimental design good research practices task force.
, respondents in the study were making hypothetical choices, which might not predict actual decisions made in a clinical setting.
Conclusions
Respondents, on average, preferred better symptom control, avoiding risk of treatment-related physical dependence, and taking oral pills daily rather than injections. Respondents were much less concerned with avoiding an incremental annual treatment-related risk of heart attack or severe joint problems. Respondents were willing to accept some level of each type of risk to improve pain symptom control. No preference heterogeneity explained by clinical or demographic characteristics of the respondents was observed. These results suggest that respondents, on average, prefer alternative pharmaceutical treatments for pain to treatments with characteristics reported for opioids and that respondents, on average, would view NGF inhibitors as acceptable alternative treatments if they have efficacy at least comparable to NSAIDs and opioids.
Contributions
DCT and DAW interpreted the data and revised the article critically for important intellectual content. MB and BH designed the study, provided study oversight, conducted the analyses, interpreted the data, and revised the article critically for important intellectual content. KK conducted the analyses, interpreted the data, and revised the article critically for important intellectual content. LA, JA, AB, JCC, LR, and LV contributed to the study design, participated in securing study funding, contributed to the analyses, interpreted the data, and revised the article critically for important intellectual content. All authors approved the article for publication and agree to take responsibility for the content.
Competing interest statement
This study was completed under a research contract between RTI Health Solutions and Pfizer and was funded by Pfizer Inc. and Eli Lilly and Company. Lucy Abraham, Jo Atkinson, Andrew Bushmakin, Joseph C. Cappelleri, and Leo Russo are employees of and hold stock and/or stock options in Pfizer Inc. Lars Viktrup is an employee of Eli Lilly and Company. Marco Boeri, Brett Hauber, and Kathleen Klein are employees of RTI Health Solutions, who were paid consultants to Pfizer in connection with the development of this manuscript. In the past 36 months, Dennis C. Turk has received research grants and contracts from the US Food and Drug Administration and US National Institutes of Health and has received compensation for consulting on clinical trial and patient preferences from AccelRx, Eli Lilly and Company, Flexion, GlaxoSmithKline, and Pfizer. David A. Walsh has received consulting fees from Pfizer and Eli Lilly and Company Ltd. and is an employee of the University of Nottingham.
Role of the funding source
This study was sponsored by Pfizer Inc. and Eli Lilly and Company. Authors affiliated with Pfizer Inc. and Eli Lilly and Company participated in designing the study, interpreting the data, and writing the manuscript and approved the final version for publication.
Acknowledgments
Kimberly Moon of RTI Health Solutions provided overall project management for this study. Kate Lothman of RTI Health Solutions provided medical writing support, which was funded by Pfizer.
Appendix A. Supplementary data
The following is the supplementary data to this article:
Incidence, prevalence, costs, and impact on disability of common conditions requiring rehabilitation in the United States: stroke, spinal cord injury, traumatic brain injury, multiple sclerosis, osteoarthritis, rheumatoid arthritis, limb loss, and back pain.
Clinical Guidelines Committee of the American College of Physicians. Noninvasive treatments for acute, subacute, and chronic low back pain: a clinical practice guideline from the American College of Physicians.
A discrete choice experiment on preferences of patients with low back pain about non-surgical treatments: identification, refinement and selection of attributes and levels.
Subcutaneous tanezumab versus NSAID for the treatment of osteoarthritis: joint safety events in arandomized, double-blind, active-controlled, 80-week, phase-3 study.
in: American College of Rheumatology (ACR)/Association of Rheumatology Professionals (ARP) Annual Meeting. November 8–13, 2019 (Atlanta, GA, USA)
Constructing experimental designs for discrete-choice experiments: report of the ISPOR conjoint analysis experimental design good research practices task force.
31057-8&id=gr1.jpg){kind=link}
31057-8&id=gr2.jpg){kind=link}
31057-8&id=gr3.jpg){kind=link}
31057-8&id=gr4.jpg){kind=link}