Patient Preferences for Psoriasis Treatments:  Process Characteristics Can Outweigh Outcome Attributes FREE

Psoriasis is one of the most frequently occurring dermatologic diseases, affecting 2% to 3% of the population in Europe and the United States.¹ As a chronic debilitating inflammatory disease of the skin and joints, psoriasis can cause considerable physical impairment. Moreover, the disease can strongly affect the emotional and social well-being and socioeconomic status of patients.^2- 5 Individuals with psoriasis often face significant social stigmatization and work discrimination.⁵ The well-being of patients is influenced not only by the disease but also by its management. Many patients, especially those with severe psoriasis, are dissatisfied with the management of their disease and frustrated by the perceived ineffectiveness of treatments.^2- 3,5

In recent years, a wide range of treatment options has been established for psoriasis, including a variety of local therapies, phototherapy, traditional systemic antipsoriatic medications, and biologicals.^6- 11 The clinical goal is to find the most efficient treatment, associated with the fewest possible adverse effects (AEs), at a reasonable cost. What can result is a process of treatment trial and error and inefficiencies in identifying an effective and patient-acceptable therapy.¹² Specifically, the processes and AEs associated with recommended treatments may not fit well with patients' social and work context and preferences.¹³ Poor fit of a recommended treatment could contribute to the high level of dissatisfaction with^5,14 and low adherence to psoriasis treatments.^15- 17 Nonadherence to prescribed treatment has been reported¹⁸ in up to 40% of individuals with psoriasis.

To minimize the risk of nonadherence and poor clinical outcome, evidence-based treatment guidelines¹⁰ call for increased patient participation in treatment decision making to build a sense of active partnership and empower patients for greater self-management of the disease. One important step to achieve this goal is to assess the preferences that individuals have for the attributes of their therapy. Preferences is an economic concept referring to the value that an individual attaches to different options when presented with a choice. More commonly acknowledged in other fields, the role and potential impact of patients' preferences is being increasingly recognized in the health care sector.¹⁹ Previous studies^20- 21 in psoriasis have used time trade-off or standard gamble preference elicitation methods in which participants are asked their willingness to trade off future life years or accept a risk of immediate death in exchange for returning to full health. Studies also have assessed health state utilities in psoriasis and psoriatic arthritis using willingness-to-pay analyses, visual analog scales,^20,22- 23 the EuroQol-5 multiattribute health utility instrument,²⁴ and, recently, conjoint analysis (CA).^25- 27

In CA, also known as discrete choice experiments, participants are asked to choose between treatment options.²⁸ The treatment options are decomposed into and described by their attributes —allowing both process and outcome attributes to be included —which may more accurately reflect factors influential in treatment decision making. The objective of our study was to use CA to explore the relative importance of process and outcome attributes associated with available treatment options for psoriasis. Moreover, we wanted to assess whether individual preferences are influenced by sociodemographic and socioeconomic characteristics because factors such as age, employment, and marital status are known to influence adherence to antipsoriatic treatments.¹⁶

Individuals with psoriasis aged 18 years or older who attended outpatient clinics at the Dermatology Department of the University Medical Centre Mannheim between December 1, 2009, and September 30, 2010, were asked to participate in the study. To ensure that the full range of antipsoriatic treatments (ie, local therapy, UV therapy, classic systemic therapy, or biologicals) would be valid options, only patients with moderate to severe psoriasis according to the criteria of the Committee for Medicinal Products for Human Use²⁹ were recruited. These included (1) those with a Psoriasis Area and Severity Index (PASI) of 10 or more³⁰; (2) patients with psoriatic involvement of the hands, feet, or head; (3) patients with psoriatic arthritis according to Classification of Psoriatic Arthritis criteria³¹; and (4) those receiving systemic antipsoriatic therapy. Patients visiting our department for the first time, patients coming for follow-up visits, and those with and without ongoing antipsoriatic therapy were included. The study was approved by the Ethics Committee of the Medical Faculty Mannheim and performed according to the principles of the Declaration of Helsinki.

After providing written informed consent, participants were assigned an identification code used to access a computerized survey to be completed before their clinical consultation. Technical assistance was provided if requested. Within the survey, participants' preferences were explored using CA. Generation and evaluation of the CA exercises involved (1) identification of key attributes associated with the range of psoriasis treatments; (2) assignment of various levels to the identified treatment attributes; (3) creation of hypothetical treatment scenarios by combining these levels in a random fashion; (4) pairing of the treatment scenarios using an orthogonal design to maximize discrimination of the relative importance of attributes while minimizing the number of choice tasks needed in the set; (5) presentation of choice sets to respondents; and (6) measurement of preferences for the treatment attributes based on the choices made for the presented choice sets.²⁸ The range of currently available treatments was grouped into 6 outcome attributes (probability of benefit, magnitude of benefit, duration of benefit, probability of AEs, AE severity, and AE reversibility) and 5 process attributes (treatment location, frequency, duration, delivery method, and cost for the individual). Four realistic attribute levels, based on currently available treatments, were created for each attribute (Table 1).

The attributes were separated into 2 groups to avoid information “overload ” during the CA exercises.³² The first group included (1) magnitude of benefit, (2) duration of benefit, (3) probability of AEs, (4) AE severity, (5) AE reversibility, and (6) cost. The second group included (1) probability of benefit; (2) treatment location, frequency, duration, and delivery method; and (3) cost. Cost was included in both groups to verify internal consistency and to allow later comparison between groups. The hypothetical treatment scenarios were designed using commercially available conjoint analysis software (http://www.sawtoothsoftware.com). Of the 4096 (4   ×  4   ×  4   ×  4   ×  4   ×  4) treatment scenarios possible for each group of attributes, 12 pairs of scenarios were randomly selected by the software for each respondent. The random selection of scenario pairs was used to minimize possible bias associated with the order in which the scenario pairs were presented (Table 2). Respondents were asked to choose their most preferred treatment option (ie, the utility maximizing option) from each pair of treatment scenarios. Utility is the net benefit an individual would derive from a treatment option, assumed to be based on the option's attribute levels and the sum of their associated “part-worth ” utilities.²⁸

Sociodemographic (ie, age, sex, and marital status) and socioeconomic (ie, household income per month and working status) characteristics and the Dermatology Life Quality Index (DLQI)³³ were also assessed in the computerized survey. The PASI for each participant was scored by 2 of the study team investigators (M.-L.S. and A.S.).

Part-worth utilities for each attribute level were computed using logit regression. The part-worth utilities were then scaled to sum to zero within each attribute. Attribute importance was assessed by calculating the range between the highest and the lowest part-worth utility for each attribute. To allow comparison between attributes, the relative importance score (RIS) for each attribute was calculated as a percentage by dividing each attribute's range by the sum of all attribute ranges and multiplying by 100.²⁸ A RIS for each attribute was calculated for each participant and then averaged across all participants. To compare results between attribute groups 1 and 2, attribute importance was translated to one list by matching the relative importance of cost as assessed in both groups.

Subgroup analysis was performed on the basis of sociodemographic factors (age, 18-35, 36-49, 50-64, or ≥65 years), sex (male or female), marital status (married or living with a partner, single, or widowed), socioeconomic factors (net monthly household income < €1000 [US $1437], €1000- €2000, or > €2000) and working status (employed or unemployed [including homemakers and retirees]), and PASI (0-5, 5.1-10, or >10) and DLQI (0-5, 6-10, 11-20, or >20). For dichotomous variables, the RIS differences between subgroups were tested for significance using analysis of variance. If there were more than 2 groups, a post hoc test was applied. Significance was assessed at P   ≤  .05.

To further investigate the influence of participants' sociodemographic and socioeconomic characteristics on the RIS, multivariate linear regression analysis was performed. In these models, the RIS for each attribute was used as the dependent variable, and age, sex, marital status, income, and PASI and DLQI were included as independent variables. Working status was not included in the regression analyses because of possible multicollinearity (working status was highly correlated with income). A model was created for each treatment attribute using the function RIS  =   β₀  +   β₁ sex  +   β₂ age  +   β₃ marital status  +   β₄ income  +   β₅ PASI  +   β₆ DLQI, with β₀ as a constant for neglected factors and β_1-6 as the coefficients to be estimated.³⁴ β indicates how the value of the RIS changes when one of the independent variables is changed and the other independent variables are held constant. Standardized regression coefficients were calculated and reported to enable comparison of the effect of each independent variable on the RIS, although different measurement units were used for the independent variables. Significance was again assessed at P   ≤  .05.

During the study period, 197 outpatients with psoriasis were asked to participate. Sixteen potential participants (8.1%) declined enrollment and 18 potential participants (9.1%) were excluded because the psoriasis was classified as mild. All 163 participants completed the computerized survey. Mean (SD) age was 49.3 (14.1) years (range, 18-80 years). Further sociodemographic and socioeconomic characteristics are reported in Table 3. Despite the fact that only participants with moderate to severe psoriasis were included, the mean (SD) PASI was relatively low (5.6 [5.5]; range, 0-32.8) because most participants were receiving antipsoriatic treatment at the time of data collection. The mean (SD) DLQI was 7.6 (6.9; range, 0-29, Table 3).

Averaged across all study participants, the attribute regarded as most important was treatment location (RIS, 26.76), followed by probability of benefit (RIS, 23.77) and method of delivery (RIS, 23.49) (Figure 1).

Among outcome attributes, the probability of benefit (RIS, 23.77) and magnitude of benefit (RIS, 18.49) were considered most important, followed by the probability of AEs (RIS, 16.96) and duration of benefit (RIS, 16.37), whereas AE severity and reversibility appeared less relevant (Figure 1).

Among process attributes, participants attached importance to treatment location (RIS, 26.76), method of delivery (RIS, 23.49), treatment frequency (RIS, 19.36), and treatment duration (RIS, 19.31). Remarkably, cost to be covered by the individual was found to be least relevant (RIS, 17.60). Most process attributes had a higher RIS compared with the outcome attributes.

The effect of sociodemographic, socioeconomic, and disease-related factors on treatment preferences was estimated in descriptive analyses (Figures 2, 3, and 4) and multivariate logistic regression models (Table 4). When participants were stratified according to sex (Figure 2A, Table 4) and marital status (Figure 2C, Table 4), no significant differences were found. However, when different age groups were compared, it was striking that, on average, older participants ( ≥65 years) found the probability of benefit to be less important than younger participants did (Figure 2B). This finding was confirmed in our regression analysis ( β  =   –0.24; P  =  .005, Table 4).

On average, participants with an intermediate monthly household income between €1000 and €2000 found the duration of benefit to be more important compared with those with lower income ( < €1000 per month; β  =  0.26; P  =  .02, Table 4) and viewed the severity of AEs as less important (P  =  .02 in descriptive analyses, Figure 3A; β  =   −0.25; P  =  .03 in regression models, Table 4). Furthermore, participants with intermediate incomes ( €1000- €2000 per month) attached less importance to the treatment delivery method compared with participants with lower income ( β  =   −0.32; P  =  .005). Notably, the importance of treatment cost covered by the individual did not differ significantly between income groups (data not shown). When attribute importance was compared between employed and unemployed participants (Figure 3B), employed participants were significantly more concerned about the reversibility of AEs (P  =  .006; data not shown). Preferences for other attributes were largely independent of the working status.

Grouping according to PASI (0-5, 5.1-10, and >10) produced no significant differences in the descriptive analyses (Figure 4A). However, surprisingly, regression analyses indicated that participants with a higher PASI considered the probability of benefit less important ( β  =   −0.18; P  =  .02) but the frequency of treatment more important ( β  =  0.17; P  =  .04; Table 4) compared with participants with a lower PASI. When subgroups according to disease-related quality-of-life impairment were created (Figure 4B), those with a DLQI between 11 and 20 (demonstrating relatively poor quality of life) regarded the probability of improvement as significantly more important than did patients with a DLQI between 0 and 5 (P  =  .02 in descriptive analyses). However, this result did not persist in our multivariate regression analysis (Table 4).

Perhaps the most interesting finding of our study is that participants attach great importance to process attributes in treatment selection. Across all study participants, the attribute regarded as most important —even more important than probability and magnitude of benefit —was treatment location. In Germany, as in some other European countries, local therapy with anthralin (dithranol), often in combination with UV light, is still a major option for treatment of moderate and severe psoriasis.^10,35- 36 This treatment is most often performed in an inpatient or day hospital setting with patients hospitalized for 2 to 3 weeks because application is tricky and anthralin stains skin and sometimes even clothing if not completely washed off after treatment. However, after broad introduction of biologicals, this regimen is often replaced by outpatient therapy with biologicals.^10- 11,37 Although our findings may be partly attributable to bias arising from sampling individuals attending our outpatient clinic, they underscore that patients prefer outpatient treatment. The high impact of treatment location may also limit patient acceptance of UV therapy, excluding home phototherapy, which requires frequent visits at outpatient dermatology offices or clinics. The observation that treatment location, as well as most other process attributes, have a higher RIS compared with AE-related attributes suggests that patients may be willing to accept AEs in trade for a treatment compatible with their personal and professional lifestyle. Treatments poorly compatible with personal and professional demands may include, in addition to inpatient and UV therapy, time-consuming local therapies and systemic therapies requiring frequent laboratory workup.

Moreover, we found that participants were willing to trade an increased risk of AEs for increased probability and magnitude of therapeutic benefit. This finding is in contradiction to the previously reported work of Seston and colleagues,²⁷ who showed that patients with psoriasis prioritized low risk of AEs (skin cancer and liver damage) and that they were willing to trade time to achieve moderate (50%) skin improvement to minimize the risk of these AEs. In the CA exercise performed by the Seston group, AEs were precisely named, which may have influenced the measured preferences. However, preferences to avoid AEs were traded against “time to achieve moderate improvement ” rather than against probability and magnitude of improvement, making it difficult to compare their results with our own.

Other studies^{2,14,23,38- 39} performed on patient preferences for psoriasis treatments have compared selected treatment options. However, the available evidence on preferences for or satisfaction with the existing systemic therapies and phototherapies is surprisingly limited, and, in a recent systematic review,³⁹ it was not possible to rank therapies according to patient preferences. For example, studies^2,14,23,38 comparing preferences for classic systemic antipsoriatic therapies vs phototherapies have reported variable outcomes, with some studies showing preference for oral therapy and some for UV therapy. Patients who obtained biologicals appear to prefer these therapies over other treatments.³⁸ Correspondingly, in a prospective study,⁴⁰ patients with psoriatic arthritis achieved a rapid and sustained gain in health utility when receiving tumor necrosis factor antagonists.

When we performed subgroup analyses, we found that preferences are, in part, dependent on sociodemographic and socioeconomic characteristics. Older patients ( ≥65 years) appear to care less about the probability of improvement compared with younger individuals. In most other studies^20- 21 using time trade-off, willingness to pay, and visual analog scales to assess health state utilities in psoriasis, utilities were largely independent of age. However, psoriasis-related impairment may be perceived differently among different age groups. For example, it has been reported^5,12,40- 41 that older patients experience psoriasis more physically than mentally compared with younger patients. Correspondingly, younger patients with psoriasis were shown²² to have greater willingness to pay for social comfort and emotional health than were older patients. Our descriptive analyses also showed the tendencies that women care more about the probability of improvement than men and that singles place greater value on this attribute compared with individuals living with a partner (Figure 2A and C). Although differences did not achieve statistical significance in our sample, it seems likely that women and singles attach special importance to appealing appearance, not the least because of expectations from society and from potential partners.

The relationship between socioeconomic status and treatment preferences appears to be more complex. According to our results, participants with intermediate household incomes ( €1000- €2000 per month), mostly belonging to the working class, attach great importance to long-lasting improvement. To achieve improvement, they are more willing to risk AEs than are participants with lower household incomes ( < €1000 per month) who are likely in Germany to be supported by social benefits. Psoriasis is known to have a negative effect on individuals' professional life and employment status.^5,42 Therefore, individuals who are employed in moderate-income occupations may be more likely to demonstrate preferences for treatment compatible with work constraints, ie, a treatment with long-lasting benefit that minimizes time lost from an hourly wage job. In descriptive analyses, we also observed the tendency that participants with a monthly income of more than €1000 care more about the probability of benefit than do those with an income below that level (Figure 3A). The same applies to participants who are working compared with those who are homemakers, unemployed, or retired (Figure 3B). Although differences were not significant, our findings may indicate that individuals with an occupation attach greater importance to appearance, possibly because of expectations at work and from society. Remarkably, and in accordance with previous willingness-to-pay analyses in psoriasis,^21- 22 income was not a predictor for the effect of individual cost on preferences. Cost to be covered by the individual was the process attribute regarded as least important, ie, participants were prepared to trade copayments for more convenient treatment modalities independent of their income.

When stratifying participants according to the PASI and DLQI, patients with greater impairment of the quality of daily life (DLQI >10) attached greater importance to the probability of improvement compared with those with a lower DLQI, as expected. By contrast, the PASI was negatively correlated with participants' preference for the probability of improvement. This finding may reflect the low correlation between objective clinical severity and psychological disability in psoriasis.⁴³ However, the PASI measured in this study does not reflect the severity of psoriasis in these participants because many had achieved significant control with use of systemic medications. Finally, patients who place great value on the probability of improvement are likely to adhere to prescribed treatments, resulting in favorable outcomes shown in a low PASI.

A great advantage of the CA exercise performed in our study is that it realistically reflects the decision-making process undertaken by patients and physicians in daily clinical practice. The attributes and attribute levels represented in the choice experiments realistically characterize all treatment options available for moderate and severe psoriasis. To our knowledge, our study is the first to explore and compare preferences for process and outcome attributes for all available psoriasis treatments. Clearly, our findings should be verified in larger and more diverse patient samples before generalizations can be made. Our study included only a subpopulation of patients with moderate to severe psoriasis treated in a German university hospital, and patient preferences may be different among individuals with mild psoriasis. We cannot imply that our findings are transferable to patients in other cultures and health systems. Furthermore, the effects of other disease-related factors, such as disease duration, previous treatment experience, and comorbidities, on patient preferences for psoriasis treatment need to be explored.

In conclusion, although patients with psoriasis attach significant importance to the probability and magnitude of benefit, it appears that process attributes, such as location and method of delivery of treatment, may be even more important. A careful history of patient preferences must be incorporated into therapeutic decision making, with particular attention to age and work context. Synthesizing evidence-based medicine with shared decision making may have the potential to enhance treatment satisfaction, facilitate adherence, and optimize efficiency and clinical outcomes.