Patients’ Preferences Related to Benefits, Risks, and Formulations of Schizophrenia Treatment

Schizophrenia is a major psychotic disorder, with symptoms including changes in perception, feeling, behavior, judgment, ideation, thought process, and motivation. Symptom manifestation is heterogeneous and variable over time (1). The lifetime prevalence of schizophrenia is approximately 1 in 100 (1), and the incidence of schizophrenia in the United States is 11.1 per 100,000 (2).

Treatment of schizophrenia with antipsychotics requires a balance between alleviation of symptoms and minimization of adverse events (3). Desired benefits include avoiding relapse or hospitalization, amelioration of positive and negative symptoms, and improvement in psychosocial and occupational functioning. Common risks include weight gain and metabolic disturbances (hyperglycemia, diabetes, and hyperlipidemia), extrapyramidal symptoms (EPS), and prolactin elevation. Treatment formulation is also important, especially considering the potential of long-acting injectables (LAIs) to simplify dosing regimens and improve outcomes among patients with poor adherence to medication (4).

Numerous choices of antipsychotic medications with various efficacy and side-effect profiles are available to the clinician. When deciding among treatments, physicians must consider both available evidence and the preferences of patients (5), which can be quantified by using stated-preference methods, such as discrete-choice experiments (DCEs), also known as conjoint analyses (6,7). Previous studies of antipsychotic preference have demonstrated that patients, physicians, and family members collectively place greater importance on productive activity (work or school) compared with positive symptoms or social functioning and place less importance on negative symptoms and side effects (8,9). Medication side effects are of greater concern among patients and their families than among clinicians (10–14).

Our study built on this research by identifying key attributes (benefits and risks) of antipsychotics and quantifying the trade-offs considered by patients when balancing attributes with formulation and adherence. Novel aspects of our work included use of a structured benefit-risk framework to identify a set of the most critical benefits and harms that physicians consider in antipsychotic treatment decisions, a quantitative approach to assess judgments for trade-offs between formulation and benefits, and a method for quantifying the impact of adherence on these trade-offs.

Methods

DCE Study

DCE studies quantify respondent preferences via a series of choice tasks requiring respondents to indicate which of several hypothetical treatment alternatives they prefer. Treatment alternatives are defined by systematically altering various treatment attributes, exposing respondents to hypothetical treatments that are reminiscent of but not representative of any existing treatment. DCEs are now common (7,15–17) and are increasingly applied in online surveys (6,7,16). Regression analysis of the association of respondents’ choices and the attributes of the treatment alternatives allows for an estimation of the relative importance of the attributes considered (6,18).

Study Sample

We targeted a sample of 300 patients, a sample size consistent with both current DCE guidelines (19,20) and current DCE practices in health (18). Respondents who were at least 18 years of age and who had a self-reported physician diagnosis of schizophrenia were identified via a prescreening survey and were recruited through Kantar Health’s online patient panel in May 2012. In total, 811 respondents received e-mail invitations to participate in the online survey. Respondents received “points” equivalent to 5–10 euros ($6–$13 U.S.) that could be redeemed for merchandise or services or donated to charity. The Office of Research Protection and Ethics at RTI International approved this study.

Survey Instrument

This survey was developed in conjunction with a similar instrument targeted to physicians (21,22). Applying an approach based in multicriteria decision analysis and the Benefit-Risk Action Team (BRAT) framework (23,24), we conducted written and telephone interviews with expert academic psychiatrists and assessed product inserts and publications to determine what attributes to include in the survey. Seven attributes were chosen, including improvements in three domains of symptoms (positive symptoms, negative symptoms, and social functioning) and incidence of four adverse events (weight gain, EPS, hyperprolactinemia [irregular periods or difficulty getting or maintaining erections], and hyperglycemia). To assess validity and reliability of our instrument, a draft version was tested in 12 open-ended cognitive interviews (25), after which minor changes were made to the wording to improve respondent comprehension.

Our survey used two types of choice tasks, one to assess treatment preferences and another to assess the impact of formulation and adherence. In the preference task, respondents randomly received one of six blocks of eight randomly ordered choice questions generated from a main-effects, D-optimal experimental design consisting of 48 paired treatments (26,27). Levels for positive and negative symptoms corresponded to the absent, mild, moderate, and severe levels in the Positive and Negative Syndrome Scale (PANSS), with descriptions based on symptom lists from the PANSS scoring convention. The choices consisted of a pair of treatments, each characterized by profiles of the seven attributes (Table 1); patients were asked to choose which treatment was better for “Pat,” a hypothetical patient with schizophrenia. [An example of a preference choice task question is available in an online supplement to this article.] Consistent with previous applications of DCEs among patients with schizophrenia (10), respondents’ preferences were elicited by making a judgment about which choice is better for a third party.

In the adherence task, respondents randomly received one of nine blocks of four randomly ordered choice tasks considering the trade-off between formulation (daily pill, monthly injection, or injection every three months) and risk of experiencing mild and severe positive symptoms (Table 2). Positive symptoms were assessed by using three distributions of mild and severe symptoms (50%/50%, 40%/60%, and 20%/80%). In these tasks, the same treatments were compared for both an adherent patient (“Pat”) and a patient who misses his oral antipsychotic medications (“Jaime”). [An example of a formulation and adherence choice task question is available in the online supplement.]

Statistical Analysis

Responses to the preference choice tasks were analyzed by using a random-parameters logit model, in which a regression examined the association of respondents’ choices and attribute levels in each scenario, allowing estimation of the relative importance of each attribute level (6,25,28). For the adherence choice tasks, conditional logit models were used to estimate the relative importance weights for formulation and chance of improvement in positive symptoms, given information regarding patient adherence history. All analyses were conducted by using NLOGIT 4.0.

Results

Sample Characteristics

Of the 811 patients invited to participate, 684 (84%) responded and 329 (41%) were eligible and provided informed consent. Of eligible respondents who consented, 301 (91%) answered at least one choice question.

Of these 301 respondents, 30 chose the same response (medicine A or B) for all eight preference choice questions. Given the random assignment of attribute profiles into column A or B, this pattern should occur for only 2.4 respondents, suggesting that these 30 respondents did not focus on the survey (29). Data from these respondents were excluded because of validity concerns, leaving 271 respondents in the final sample.

Table 3 summarizes the demographic and clinical characteristics of the final sample: 73% were white, 60% were male, 33% were married, 60% were diagnosed as having schizophrenia between 15 and 25 years of age, and 90% took prescription medication for schizophrenia; the mean±SD age was 38.4±11.9 years. Compared with participants in the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) study (30), our sample reflected greater rates of employment, married status, white race, female gender, and use of prescription medication, and more years of education. Significance tests showed that greater educational attainment (high school or less versus some college or more) was not associated with statistically significant differences in preferences.

Mean Relative Importance Weights for Outcomes

Statistical analysis of the preference choice tasks indicated that survey participants considered improvement in positive symptoms the most important outcome (Figure 1). That is, a treatment associated with an improvement in level of positive symptoms from severe to none and that did nothing else would provide more perceived benefit than a treatment that improved any other attribute over the range of levels studied. This change was assigned an importance value of 10.0.

The second most important feature was eliminating hyperglycemia, which had a mean relative importance of 3.6 (95% confidence interval [CI]=2.6–4.6), indicating that avoiding hyperglycemia was approximately one-third as important as complete improvement in positive symptoms (p<.05). The relative importance of the other attributes, listed in order of decreasing importance, was improvement in negative symptoms from severe to none (3.0, CI=1.6–4.3), 30-pound weight gain (2.6, CI=1.2–4.0), hyperprolactinemia (1.7, CI=.9–2.6), improvement in social functioning from severe to mild (1.5, CI=.4–2.5), and EPS (1.0, CI=.3–1.8). The adverse event with the greatest relative importance was hyperglycemia, which was considered more than 3.5 times as important as the adverse event with least relative importance, avoiding EPS (p<.05).

The model also yielded insights on the changes within an attribute that were most important (data not shown). For weight gain, an increase of 15 to 30 pounds was three times more important than an increase of 0 to 15 pounds. For negative symptoms, the only statistically significant difference in importance was a change from mild to no symptoms. In contrast, for positive symptoms, the changes between levels were of similar importance.

Weights for Formulation and Adherence

The adherence choice tasks inquired separately about two levels of nonadherence: missing one to two doses per week and missing three to four doses per week (roughly missing 20% and 50% of doses, respectively). Because there was no difference between the results for these two levels of nonadherence, the results for both responses were pooled and referred to as nonadherent.

Preference weights showed different patterns for an adherent and a nonadherent patient (Figure 2). For an adherent patient, respondents preferred a daily pill to an equally effective monthly injection (p<.01) and preferred monthly injection to three-month injection (p<.01). For a nonadherent patient, respondents preferred monthly injection to a pill (p=.01). The preference difference between monthly and 3-month injections for a nonadherent patient was not statistically significant.

Finally, the clinical value of an injectable formulation for nonadherent patients can be inferred by comparing the change in preference for switching between an oral and injectable with the change in preference for achieving a given reduction in positive symptoms (Figure 2). Participants indicated that for a nonadherent patient, an increase of up to 26% (CI=3%−49%) in the chance of severe positive symptoms was an acceptable trade-off for switching from an oral formulation to a one-month injectable. For adherent patients, because the oral formulation was preferred over injectables, participants indicated that to be an acceptable trade-off, switching from an oral formulation to a monthly injectable would require a reduction of at least 15% (CI=7%−23%) in the chance of severe positive symptoms.

Discussion

Understanding the importance that patients place on the benefits and risks of antipsychotics and how formulation affects those trade-offs provides insight into past decisions and useful information for future regulatory and treatment decisions. This study built on prior work in this area in three key ways. First, by using a structured benefit-risk approach with input from key opinion leaders and a literature review, we identified a key set of benefits and risks that physicians consider when making decisions about antipsychotic treatment. Second, by incorporating formulation into the choice questions, we obtained quantitative estimates of respondents’ willingness to accept trade-offs between formulation and degree of benefit. Third, by providing information on a hypothetical patient’s adherence, we assessed how perceptions of adherence affected these trade-offs.

Patients regarded complete removal of positive symptoms as more important than any other symptom or adverse event assessed. Notably, an improvement in positive symptoms from severe to moderate levels was as important as or more important than improvement in any other adverse event included (Figure 1). This finding suggests that the main driver in antipsychotic treatment decisions is stopping severe positive symptoms. Once a patient with severe positive symptoms shows some improvement, the trade-off between improved efficacy and adverse events becomes more important. Although the rationale behind these measurements was not examined, one possible explanation is that patients understand the dangers associated with severe positive symptoms more than those for negative symptoms, and they may view adverse events as amenable to control by adjustments in dose, choice of antipsychotic, or both.

Other findings related to the value respondents placed on switching from an oral antipsychotic to a LAI. As might be expected, this value depended on the hypothetical patient’s adherence behavior. For adherent patients, oral formulations were judged superior to injectables. For nonadherent patients, respondents showed a statistically significant judgment in favor of a one-month LAI over an oral form. The importance of switching to a monthly LAI was similar to the importance of a 26% change in chance of reducing the level of severe positive symptoms. In other words, given the choice between a highly effective oral drug and a somewhat less effective LAI, respondents would choose the LAI for nonadherent patients. If confirmed, such results may be useful for regulatory decision making and clinical practice.

A surprising result is related to patient perceptions about improvements in negative symptoms. An improvement in negative symptoms was considered important only for the elimination of mild symptoms, not for improvements of severe or moderate symptoms to moderate or mild levels. Although patients noted distinctions among the survey descriptions of severe, moderate, and mild negative symptoms (Table 1) during pretesting, they did not value changes in these levels of severity while taking the survey. This finding may be a consequence of differences between viewpoints of patients and physicians and patients’ limited understanding of the effect of negative symptoms. The absence of approved treatments for negative symptoms (and social functioning) could diminish understanding of these symptoms and the value placed on them, given that there is not much real-world expectation for improvement.

The findings of this study were similar to those of other published studies in which respondents valued improvements in social functioning and positive symptoms more than improvement in negative symptoms and avoidance of side effects (8,9). A related study in which the authors assessed psychiatrist judgments about medications’ benefits and risks showed that the psychiatrists and patients had several results in common (21,22). Like patients, psychiatrists showed that positive symptoms were their dominant concern. Psychiatrists showed little difference in their opinions about the importance of various formulations for an adherent patient, but as adherence decreased, psychiatrists preferred both one-month and three-month injectables over oral formulations (p<.01). Like patients, psychiatrists would accept up to a 20% to 25% reduction in efficacy in order for a highly nonadherent patient (missed 50% of doses) to switch from a monthly injectable to an oral formulation.

An important consideration is the degree to which the judgments of a panel of patients can be used in the context of individual patient treatment. Mean results from a panel may not be informative about individual preferences. However, panel results can be considered in a manner similar to the consideration of clinical data in treatment guidelines. Both provide evidence about patient populations and enable physicians to better compare the small samples they treat with the larger populations described in the guidelines. Treatment decisions potentially can then be based on both clinical study data and preference data. Mean preference results may be particularly valuable when treating a patient who chooses not to indicate personal preferences or is incapable of doing so.

There were several noteworthy limitations of this study. First, we designed the survey to help respondents interpret the attributes consistently and as intended. However, evaluating choice tasks could be cognitively difficult for patients with schizophrenia, although recent work suggests that is not the case (10). A training section of the survey provided attribute definitions and practice questions. Second, as in all DCE studies, experimental control over the decision stimuli required respondents to evaluate hypothetical choice alternatives. Thus there was potential for hypothetical bias. The study minimized hypothetical bias by using patient-friendly descriptions of treatment attributes that reflected clinically realistic outcomes.

A third limitation was that constraints on cognition limited the number of endpoints that could be considered simultaneously by survey respondents. We used a structured approach for endpoint selection to mitigate this limitation.

Fourth, unlike positive and negative symptoms, social functioning did not include a level for no symptoms or complete cure. This level was excluded because, when pretesting the physician survey on which this patient survey was based, physicians were unable to accept a scenario in which negative symptoms were unaccompanied by limitations on social functioning. We posited that patients would have a similar concern. If the full range of social functioning was included, social functioning might have shown greater importance compared with negative symptoms. Additionally, using alternative definitions of social functioning that reference jobs, independent housing, or time with friends or family may have led participants to attach greater importance to this attribute.

Fifth, this study surveyed a convenience sample of patients from the United States with access to the Internet. This design limits the confidence with which these results can be generalized. As suggested by Table 3, this sample represents a more educated population than is typical for patients with schizophrenia. Future research using a randomized patient sample to verify our findings would be valuable. Readers should also understand that these relationships may change over time as understanding of schizophrenia and its treatment evolves.

Sixth, this study was not designed, nor powered, to address the question of whether judgments would change depending on experience with a particular adverse event, such as EPS, or with adherence challenges. Any potential differences were averaged out. However, a post hoc test on results for the 198 patients who answered a question about self-assessed adherence showed no difference between the groups who did or did not adhere to medication. With a larger sample size, a study could show meaningful differences in judgment on the basis of personal or family medical history and could potentially provide better guidance for individual patient treatment.

Seventh, the survey assessed judgment, which is distinct from personal preference or choice. Schizophrenia patients are very sensitive about revealing personal information, and prior DCE work showed that such patients answer hypothetical questions about others (judgments) more easily than about themselves (preferences) (10). Formulating the questions as judgments also avoided confusion with patients’ experiences and expectations about how treatments could affect their own emotional responses, encouraged objectivity, and reduced potential yea-saying bias compared with a question about what patients would choose for themselves (10,31–34). Use of the results as indicators for personal treatment decisions assumes that the respondents’ judgments are a valid representation of personal preference.

Finally, although DCE surveys could be conducted through in-person interviews, many researchers believe that computerized administration assures respondents of anonymity and reduces yea-saying and interviewer bias. Results from online DCE studies are generally not statistically significantly different from those elicited through face-to-face interviews (35,36), and several DCE studies using online patient panels have been published (6,7,16,37).

Conclusions

Balancing the benefits and risks of treatments is the core of treatment decisions by health authorities, physicians, and patients. This study demonstrated that patients evaluated treatments primarily on the basis of improvement in positive symptoms and that hyperglycemia was judged the most important adverse event. Patients judged oral formulations to be better for adherent patients but judged a monthly injection to be better for nonadherent patients. The results are consistent with our prior DCE study with psychiatrists. Studies of this type can help in understanding the importance people place on the benefits and risks of antipsychotics and how formulation affects those trade-offs, providing insight for both regulatory approval and shared decision making between patients and physicians.