Preference Assessment of Recruitment into a Randomized Trial for Adolescent Idiopathic Scoliosis

Abstract

Background: Randomized controlled trials are powerful tools to evaluate the outcomes of clinical treatments. However, these trials tend to be expensive and time-consuming, and their conclusions can be threatened by several limitations. This study estimated the strength of three common limitations (underenrollment, selective enrollment, and nonadherence to protocol) in a proposed study of adolescent idiopathic scoliosis.

Methods: Patients with scoliosis and their parents were asked to complete a web-based survey about their preferences concerning a hypothetical randomized trial. Adolescents without scoliosis and their parents also participated. Surveys included questions about treatment preference, likelihood of participation, required risk reduction, and propensity to drop out or choose a different treatment while enrolled in the study.

Results: Ninety adolescents and eighty-three parents participated. Observation was preferred to bracing by the majority of subjects. Overall, 33% of the parents and adolescents would both agree to participate in the hypothetical trial. Of the subjects who would not agree to participate, the majority would rather share the decision-making responsibility with the physician than have the treatment chosen in a random fashion. Many of the subjects would consider changing treatments during the course of the trial if they were not satisfied with the outcomes; the majority of parents who preferred bracing would consider crossing over to the bracing arm if their children were randomized to observation.

Conclusions: Recruitment into a randomized trial of bracing compared with observation for the treatment of adolescent idiopathic scoliosis may well be problematic, considering the relatively small percentage of families who said they would consider randomization. Additionally, the threat of nonadherence to protocol may be strong and must be addressed in the protocol of the trial. Most families wanted to make the treatment decision with the physician in lieu of randomization; therefore, the role of the physician in patient recruitment and retention should not be underestimated.

The randomized controlled trial has emerged as the gold standard for all clinical research. The combination of randomized assignment to groups, use of strict inclusion and exclusion criteria, standardized protocols, and a priori power analysis serve to rule out many threats to the internal validity of research results. Thus, it is the most powerful methodology available to researchers investigating the relative efficacy of treatments. This power comes at a cost, however, as most randomized controlled trials require a dedicated research infrastructure, large sample sizes, multi-institutional cooperation, and several years to complete. Additionally, the process of recruitment, randomization, treatment, and follow-up are subject to considerable threats such as underenrollment, selective enrollment, and nonadherence to the research protocol¹.

Rather than waiting for evidence of threats to validity to emerge during the course of a trial and then reacting, researchers should be proactive and investigate these threats during the planning phase of the trial. Halpern termed such an investigation a prospective preference assessment¹. Prospective preference assessment involves the presentation of a hypothetical trial to subjects who are representative of the target population, followed by a series of questions. These questions concern the willingness to participate and the potential barriers to participation such as demographic characteristics, preference for one treatment over another, and propensity to change treatment after the results of randomization are known. This method has been used in an orthopaedic application by Creel et al.², who found that over one-fifth of patients with symptomatic meniscal tears would be willing to participate in a randomized controlled trial comparing arthroscopy with nonoperative therapy. They also found that patients with a strong preference for one treatment over the other were much less likely to participate in the randomized controlled trial than were those who had no preference.

We undertook a prospective preference assessment as an integral part of our planning for a randomized trial of bracing compared with observation in adolescent idiopathic scoliosis. The work in the present study was funded by a clinical trials planning grant from the National Institute of Arthritis and Musculoskeletal and Skin Diseases, the purpose of which is to grant funds to researchers to provide support for the development of Phase-III clinical trials. This program supports the establishment of the research team, data management tools, recruitment strategies, and the finalization of the protocol and other essential elements of the study.

Our ultimate goal is to successfully conduct the first randomized controlled trial comparing the use of bracing and observation on subsequent progression to surgical indication in patients with adolescent idiopathic scoliosis. Adolescent idiopathic scoliosis is a structural lateral and rotatory curvature of the spine arising in otherwise normal children during puberty. Although many nonoperative treatments for adolescent idiopathic scoliosis have been described^3-5, the only modality with any evidence of effectiveness is orthotic (brace) treatment. The expectation is to prevent progression of an existing spinal curvature until the patient reaches skeletal maturity, at which time the risk of curve progression greatly diminishes. Many investigators have examined the effectiveness of bracing in adolescent idiopathic scoliosis over the past half century^5-38. However, critical analysis of the methodology and results of those studies yields inadequate evidence concerning the effect of bracing on curve progression, the rate of surgery, and the burden of suffering associated with adolescent idiopathic scoliosis. Clinicians are left with conflicting and questionable inferences regarding bracing effectiveness—hence the need for a randomized clinical trial.

The present study was designed not only to increase our general knowledge about the willingness of patients and parents to participate in a pediatric orthopaedic randomized controlled trial but also, for a very practical purpose, to gather specific information needed to write the grant and plan the logistics of the proposed randomized controlled trial. The aims of this study were fourfold.

The first aim was to evaluate the threat of underenrollment by estimating the number of adolescents, parents, and adolescent-parent pairs who would consent to be randomized into the proposed randomized controlled trial. Underenrollment threatens the value of outcomes when, despite the use of substantial resources to enroll an adequate sample, the sample size remains too small to provide adequate statistical power. Underenrollment has been attributed to the characteristics of physicians who refer or enroll their patients into trials, the characteristics of potential subjects, the characteristics of the trial itself, and the logistical barriers to trial participation.

Second, we wanted to evaluate the threat of selective enrollment by determining the characteristics of potential participators and nonparticipators. Whereas underenrollment is a threat to internal validity, selective enrollment is a threat to external validity. Selective enrollment occurs when subsets of the target population enroll in greater or lesser proportions than their representation in the population. Narrow recruitment strategies, differential access to the study, different levels of willingness to participate, and different levels of willingness of clinicians to enroll certain subgroups of patients can contribute substantially to selective enrollment. For example, if only those at low risk for curve progression chose to enroll, the generalizability of the findings to the adolescents at higher risk is compromised. Without knowledge of the extent of selective enrollment, researchers run the risk of generalizing trial findings to other patients whose characteristics were not accounted for in the outcomes of the trial.

Third, we wanted to estimate the threat of nonadherence to the protocol by estimating the strength of patient preference for one arm relative to the other, and the propensity for crossover. Nonadherence to protocol can be seen as a threat to both internal and external validity through its affect on the dose-response relationship. If treatment effects are attenuated by nonadherence to the bracing protocol, the statistical power of the study may be decreased and false inferences concerning the relative effectiveness of the treatments can result. Thus, the same outcomes might not be seen in the target population if the sample is more or less adherent. Nonadherence, then, if a function of patient characteristics, can be seen as a special case of selective enrollment with the same implications for generalizability of study findings. Crossover, i.e., when the subject changes from the assigned treatment to the alternative, decreases the interpretability of findings with use of an intention-to-treat analysis because outcomes are attributed to the treatment the subject is randomized to and not the treatment that was actually received. This could be countered by an as-treated analysis, but those findings could also be questioned because of potential unknown or unmeasured factors that could potentially influence either the treatment received or the response to the treatment.

The fourth aim was to determine the effect size to use in power calculations for the proposed randomized controlled trial by measuring the risk reduction desired by the subjects. Required risk reduction was defined as the number of surgeries avoided by the use of a brace before the participant would favor bracing over observation, given their understanding of the side effects and inconvenience of bracing.

Materials and Methods

Sample

This protocol was reviewed and approved by the human research subject boards of the University of Iowa Hospitals and Clinics, Iowa City, Iowa; the University of Virginia, Charlottesville, Virginia; Columbia University, New York, NY; Children's Hospital of San Diego, San Diego, California; and Shriners Hospital for Crippled Children, Erie, Pennsylvania. These sites were chosen to increase the sample size and to try to match the sociodemographic diversity of subjects wanted in the proposed trial. The target population was the adolescents who met the inclusion criteria for the proposed randomized controlled trial and their parents (Fig. 1). We sought subjects from two groups: patients with adolescent idiopathic scoliosis between the ages of ten and fifteen years who had not been managed with braces or surgery (the scoliosis group) and adolescents in the same age group without a scoliosis diagnosis (the surrogate group). The decision to include a surrogate group has precedence in several similar studies that have argued that an untreated sample is necessary to avoid potential bias due to previous treatment recommendations, an internalized sense of the risk of disease progression, or actual treatment decisions and outcomes^39-43. Also included in the study was a parent of each adolescent, since it is the parent or guardian who ultimately has the responsibility for making treatment and research participation decisions. Only one parent was included for convenience.

Fig. 1.

Proposed design of the randomized controlled trial. AIS = adolescent idiopathic scoliosis.

Potential subjects were identified in two ways. The scoliosis group was recruited during their routine clinic visits and the surrogate group was recruited by (1) a note sent home through the local school district, (2) a notice in local newspapers, and (3) a sign posted at the local public library. After the subjects were identified, they were either given or mailed a consent form. Once the consent form was signed and returned, and the names, addresses, and telephone numbers of the subjects were recorded, they were given study identification numbers and specific instructions on how to access the presentation and survey via the web. Once their survey results were retrieved from the database, they were telephoned to ask about their experience and then were mailed a $10 bill as compensation for their time.

Presentations and Survey

The design of this study was based on previous work concerning recruitment issues in randomized clinical trials, specifically the method of prospective preference assessment described by Halpern¹. Typically, these assessments start by describing a hypothetical trial including information about randomization, the research question motivating the trial, previous research, and the potential risks and benefits of both treatment arms^39,44-46. The description is then followed by an interview or survey to measure patient preferences and decision-making strategies as they pertain to trial participation. Such surveys frequently include a treatment trade-off task to determine the amount of risk reduction required by the subjects before they prefer potentially more effective, but also more toxic, treatment. This type of exercise has been seen in the literature concerning a wide range of elective therapeutic options, including participation in clinical trials^1,47-50, and the results can be used to determine the necessary effect size when planning a clinical trial. The treatment trade-off task has recently been used in a pediatric application concerning the preferences of parents and health professionals for the management of febrile neutropenia⁵¹.

Three web-based presentations were tailored to the roles and developmental levels of the participants and thus included a parent version, a version for thirteen to fifteen-year-old adolescents, and a version for ten to twelve-year-old children. Each described the proposed randomized controlled trial. The content presented to the parents and the older adolescents included an overview of the natural history of adolescent idiopathic scoliosis, surgical treatment, and the results of previous studies on bracing effectiveness. The presentation included an evaluation of previous research into bracing effectiveness, and it stated that some clinicians believe bracing is beneficial, but, conversely, some clinicians believe the effectiveness estimates are based on limited and flawed experimental evidence. The presentations also included information about the perceptions of patients and parents who had experienced bracing treatment, which we had gleaned from our experience with families during routine clinic visits, focus groups concerning decision-making patterns in adolescent idiopathic scoliosis, and support groups for adolescents with adolescent idiopathic scoliosis⁵². This patient-derived information was incorporated into sections dealing with the positive and negative aspects of each treatment arm. The treatment arms were summarized side by side to facilitate comprehension and comparison as used in previous studies in a randomized controlled trial context^39,41 (Table I).

TABLE I.

Summary of the Observation and Bracing Treatment Arms from the Presentation to Parents

Observation	Bracing
Plan of Care
Observation is not commonly used in the United States, but it is used in many centers in Europe. Some doctors believe that there are no good treatments for adolescent idiopathic scoliosis, and so the patient should just be observed.	Bracing is the standard of care for scoliosis in the United States. Some doctors believe that wearing a brace decreases the risk of curve progression to 45° before skeletal maturity.
Observation involves having x-rays and a clinical examination every 6 months until one of two things happens: (1) the patient reaches skeletal maturity or (2) the curve progresses to 45°, at which time surgery would be discussed.	Bracing involves wearing a plastic brace up to 23 hours per day and having x-rays every 6 months until one of the following two things happens: (1) the patient reaches skeletal maturity or (2) the curve progresses to 45°, at which time surgery would be discussed.
Possible Side Effects
There are no known side effects associated with observation and regular checkups. However, if braces work, by not wearing one, the chance of curve progression and surgery may be increased.	Possible: skin irritation, back discomfort, and social and family problems associated with not following doctor's orders.
	Rare: decreased kidney function, decreased respiratory function, abnormal breast development, and discrimination.
Possible Benefits
Observation allows for frequent assessment of curve status and will allow us to know if your curve has progressed.	Bracing may decrease the risk of curve progression to 45°, the size at which most doctors would consider suggesting surgery to correct the curve.
Patients do not wear a brace or have any of the problems associated with wearing a brace.	The curve is stabilized, but the brace provides no permanent correction of the curve.
Possible Consequences
A certain number of curves do progress and require surgery. If surgery is needed, the curve will be significantly corrected.	A certain number of curves do progress despite brace treatment. Therefore, some patients potentially spend years getting unnecessary treatment. If surgery is needed, the curve will be significantly corrected.

The survey was based on previous literature concerning treatment decisions and trial participation^39,53-57. The items and scales used to operationalize these concepts are outlined below:

• The decision to participate (agree or refuse).

• The reason for agreeing (the physician suggested it; benefits to future treatment; lack of treatment preference; my child, or my parent, would want to participate; or other).

• The reason for refusing (wanting to make own decision; my child, or my parent, should make decision; physician should make decision; or a preference for shared decision-making)⁵⁸.

• The treatment preference (bracing or observation) and the strength of the preference (slight, somewhat, or strong).

• The propensity to cross over (i.e., switch to the bracing arm) if randomized to watchful waiting and the curve progressed (definitely, maybe, or definitely not).

• Demographic factors (age, sex, education, or race).

• Required risk reduction derived from treatment trade-off task^43,47,48. This task asks the subject to consider possible surgical rates after observation and bracing, while considering the side effects and inconveniences of bracing. The baseline surgical rate for both treatments was set at 60%. The subjects were asked to choose their preferred treatment when surgical rates were equal, and then again as the surgical rate after bracing was decreased from 60% to 50%, 40%, 30%, 20%, 10%, and 0%. The surgical rate at which the subjects prefer bracing over observation is an estimate of their required risk reduction, given their perceptions of the bracing experience.

Statistical Analysis

Since the primary goal of this study was to estimate the number of subjects agreeing to participate, the necessary sample size was calculated on the basis of a 90% confidence interval around the hypothesized agreement rate. In order to achieve a confidence interval with a spread of 10% on either side of the point estimate of 30%, fifty-seven subjects would be required. For a spread of 7.5%, 102 subjects would be required. The sample and the relationships between variables and the decision to participate were characterized through the use of descriptive statistics, Fisher exact tests, and Student t tests. Alpha was set at 0.05. No a priori analyses were done to suggest sample size or power for detecting these relationships.

Results

Demographic Characteristics of the Adolescents

Table II summarizes the demographic characteristics of the sample. The adolescent group included ninety subjects. The average age was twelve years (range, ten to fifteen years); forty-two completed the version of the survey for the thirteen to fifteen-year-old group, and forty-eight completed the version for the ten to twelve-year-old group. Forty-seven subjects (52%) were girls, and twenty-four (27%) had scoliosis. The scoliosis group was, on the average, sixteen months older than the surrogate group (p < 0.0004), and it contained more girls (75% compared with 44%; p < 0.02), although these factors were not found to significantly influence decisions concerning participation.

TABLE II.

Summary of Demographic Data, Decisions, and/or Treatment Preferences of Adolescents by Group (Scoliosis or Surrogate)

	Adolescent Comparisons
Item	Scoliosis (N = 24)	Surrogate (N = 66)	P Value
Average age (and standard deviation) (yr)	13.39 (1.46)	11.89 (1.64)	0.0004
Female	18 (75%)	29 (44%)	0.02
Agree to participate	7 (29%)	36 (55%)	0.06
Reason for agreeing			0.68
Physician suggestion	1	2 (6%)
Find best treatment	6	22 (61%)
No preference		6 (17%)
Parent and/or child would like it		3 (8%)
Other		3 (8%)
Refuse to participate	17	30
Reason for refusing*			0.66
Make own decision	7	8 (29%)
Parent and/or child should decide	1	1 (4%)
Physician should decide	3	4 (14%)
Share decision	6	15 (54%)
Preference for treatment	24	66	0.54
Yes	19 (79%)	37 (56%)
No	0	9 (14%)
Don't know	5 (21%)	20 (30%)
Prefer observation	15	26 (70%)	0.15
Strength of preference†
Slightly	1	2	0.22
Somewhat	5	8
Strongly	10	4

^*Two responses specifying the reason for refusing were missing in the surrogate group.

^†The ten to twelve-year-old group did not answer this question. Of the fifty-six subjects with a preference, thirty were in the older group and therefore answered the question. Twenty-six subjects were not presented with the strength question.

Decision-Making of the Adolescents

Of the ninety adolescents, forty-three (48%; 95% confidence interval, 36% to 57%) indicated a willingness to participate in the proposed randomized controlled trial of bracing (Table II). Seven (29%; 95% confidence interval, 14% to 44%) of the twenty-four subjects in the scoliosis group were willing to participate compared with thirty-six (55%; 95% confidence interval, 43% to 67%) of the sixty-six subjects in the surrogate group. The majority of subjects in both groups stated they would participate in order to help find the best treatment for patients in the future. Of those who would refuse to participate, six of seventeen subjects in the scoliosis group and fifteen (54%) of twenty-eight subjects in the surrogate group (responses from two subjects were missing) stated that they would refuse because they did not want their treatment decided by randomization but instead wished to share in the treatment decision-making process. Less than 25% of those in each group felt that their parent or the physician should unilaterally make the treatment decision.

Preference for Treatment Among the Adolescents

The majority of both the scoliosis group (fifteen of nineteen adolescents) and the surrogate group (thirty-seven of sixty-six subjects; 56%) had a preference for a treatment arm (bracing or observation). The observation arm was overwhelmingly preferred by both groups (93% and 70%, respectively). Overall, having a preference was related to a decision to participate and was stronger in the scoliosis group. Of the thirty-seven subjects with a preference in the surrogate group, twenty-one (57%) would participate in the trial compared with five of the nineteen scoliosis subjects (p < 0.05).

Demographic Characteristics of the Parents

Characteristics of the parent sample are summarized in Table III. Eighty-three parents participated. Twenty-three (28%) were the parent of a child with scoliosis. The average age was forty-seven years (range, thirty-four to sixty-five years), seventy-four (90% of eighty-two who answered the question) were white, four (5%) were Hispanic, and four (5%) were of other racial backgrounds (the response from one parent was missing). The sample was predominately female (sixty parents; 72%). The majority of the sample had attended at least one year of college. These demographic characteristics were not significantly different between the parent groups nor were they significantly associated with the decision to participate.

TABLE III.

Summary of Demographic Data, Decisions, and/or Treatment Preferences of Parents by Group (Scoliosis or Surrogate)

	Parent Comparisons
Item	Scoliosis (N = 23)	Surrogate (N = 60)	P Value
Average age (and standard deviation) (yr)	46.61 (5.40)	47.18 (5.72)	0.67
Female	20 (87%)	40 (67%)	0.13
Agree to participate	9 (39%)	37 (62%)	0.08
Reason for agreeing	9	37	0.06
Physician suggestion	3	1 (3%)
Find best treatment	5	25 (68%)
No preference		5 (14%)
Parent/child would like it		3 (8%)
Other	1	3 (8%)
Reason for refusing	13	23	0.005
Make own decision	3	0
Parent and/or child should decide	1	2 (9%)
Physician should decide	4	2 (9%)
Share decision	5	19 (83%)
Preference for treatment	23	60
Yes	10 (43%)	29 (48%)	0.64
No	3 (13%)	12 (20%)
Don't know	10 (43%)	19 (32%)
Prefer observation	7	15 (52%)	0.46
Strength of preference
Slightly	1	2 (7%)	0.99
Somewhat	5	16 (55%)
Strongly	4	11 (38%)

Decision-Making of the Parents

Of eighty-three parents, forty-six (55%; 95% confidence interval, 46% to 66%) would consent to having their child in the proposed randomized controlled trial. The potential participation rate was not significantly different (p < 0.09) between the scoliosis parents (39%; 95% confidence interval, 22% to 59%) and the surrogate parents (62%; 95% confidence interval, 49% to 73%). Of the nine parents in the scoliosis group who would agree to participate, five did so in order to help to find the best treatment for the future, three agreed because the physician suggested it, and one parent agreed for an unlisted reason. The pattern of responses was somewhat different in the surrogate group (p < 0.06): twenty-five (68%) of thirty-seven parents agreed to help to find the best treatment and one agreed because the physician suggested it. However, unlike the parents of the scoliosis subjects, some parents (five; 14%) of the surrogate subjects agreed to participate because they had no preference for treatment, three (8%) agreed because they felt their child would like to participate, and three (8%) agreed for some unlisted reason.

Of the thirty-six parents who would refuse to let their child participate in the proposed randomized controlled trial, most (twenty-four; 67%) did so because they wanted to share responsibility for making the treatment decision. The reasons for refusal were different between the groups (p < 0.005). Four of the thirteen parents of the scoliosis patients would refuse to participate because they desired the physician to make the treatment decision, whereas only 9% of the twenty-three parents of the surrogate subjects desired this. Three parents of scoliosis patients would refuse so that they could make the treatment decision themselves, and two parents of surrogate subjects would refuse so that their child could make the decision. Of the twenty-four parents in both groups who wanted to make a shared decision, 44% felt they and the physician should make the decision, 36% felt they and their child should make the decision, and 20% felt their child and the physician should make the decision.

Preference for Treatment Among the Parents

The parent sample was split in terms of preference for treatment; thirty-nine (47%) of eighty-three parents had a preference for treatment (either bracing or observation), fifteen (18%) had no preference, and twenty-nine (33%) were not sure. There was no difference in these rates between the scoliosis and surrogate parent groups (p = 0.64). Thirty-eight percent of the parents with a preference for treatment arm had a strong preference compared with the slim majority who slightly or somewhat preferred one treatment relative to the other. Of the ten scoliosis parents with a preference, seven preferred the observation arm compared with fifteen (52%) of the twenty-nine parents of surrogate subjects (p < 0.46). There was a nonsignificant relationship between preference and agreeing to participate in the randomized controlled trial (p < 0.22). Eleven of the fifteen parents with no preference agreed to participate, 56% of the thirty-nine with a preference agreed, and 44% of the twenty-nine who did not know whether they had a preference agreed to participate. The bracing arm was preferred by 45% (ten) of the twenty-two who agreed to participate and by seven of the seventeen who would refuse to participate (p < 0.99).

Treatment Trade-off for Required Risk Reduction Among the Parents

The treatment trade-off exercise produced the required risk reduction of the parents, indicating the decrease in the rate of surgery required before they would endorse bracing over observation. The exercise set the baseline surgery rate at 60% for both treatments. Therefore, a required risk reduction of 0% indicated the parent preferred bracing even when the surgery rates were 60% after each treatment, and a required risk reduction of 100% indicated the surgery rate would have to decrease to zero before bracing was preferred. The median required risk reduction was 50%, or a reduction from a 60% surgery rate after observation to a rate of 30% (range, 0% to 100%) after bracing. This rate did not differ between the parent groups (p < 0.58) or between those who would agree or those who would refuse to participate (p < 0.85).

Propensity for Crossover Among the Parents

The parents were asked what they would do if their child's curve progressed while he or she was randomized to the observation arm. Four (5%) of eighty-three parents indicated they would definitely consider crossing over to the bracing arm, fifty (60%) would consider crossing over, and twenty-nine (35%) would definitely continue with observation. These rates did not differ between parent groups (p < 0.50) or between those who agreed and those who refused to participate (p < 0.28). As expected, parents who preferred bracing (but whose child was theoretically randomized to observation) would more frequently consider crossing over than would those who preferred observation (76% of seventeen compared with 31% of twenty-two; p < 0.01).

Concurrence Between Parents and Adolescents on Participation

The decisions to participate made by the parents and the adolescents are summarized in Table IV. Overall, fifty-five (66%) of eighty-three parent-adolescent pairs concurred in their decision to participate; twenty-eight pairs would refuse and twenty-seven pairs (33%; 95% confidence interval, 23% to 43%) would agree to participate. Of fifty-nine pairs in the surrogate group, forty (68%) concurred; 27% would refuse to participate, and 41% (95% confidence interval, 29% to 53%) would agree to participate. However, the rate of concurrence and willingness to participate was lower in the scoliosis families. Of the twenty-four parent-adolescent pairs in the scoliosis group, 63% concurred; 50% would refuse to participate, and 13% (95% confidence interval, 4% to 31%) would agree to participate. Of the twenty-seven pairs from both groups in which both the adolescent and parent had a treatment preference, twenty (74%) were in accord: 52% jointly preferred observation and 22% jointly preferred bracing.

TABLE IV.

Joint Agreement to Participate by Total Sample and by Group (Scoliosis or Surrogate)

	Adolescents
Parents	Agree	Refuse
Total sample
Agree	27 (33%)	19
Refuse	9	28 (33%)
Scoliosis group
Agree	3 (13%)	6
Refuse	3	12 (50%)
Surrogate group
Agree	24 (41%)	13
Refuse	6	16 (27%)

Discussion

This study was designed to lend insight into the preferences of subjects considering the decision of whether to participate in a randomized trial and, specifically, to help us to plan the randomized controlled trial. It was particularly important for us to estimate the potential for underenrollment. Using the three estimates of agreement to participate (33% overall, 41% for the surrogate group, and 13% for the scoliosis group), we set the estimated randomization rate for the proposed randomized controlled trial at 25%. We also set the effect size (relative risk reduction) at 50%. Both of these estimates were used to help us to determine that approximately twenty participating centers would be needed to recruit 449 subjects in the first two years of the project. Periodic evaluations are to be conducted during the trial to determine the predictive validity of the participation rate found in the present study.

We were also concerned about selective enrollment. Several barriers, such as distance and financial considerations, low levels of physician participation, narrow inclusion criteria, and the complexity of the informed consent process, have been extensively studied in the literature^59-62. Sociodemographic and attitudinal variables of the patients, such as those in the present study, have also been shown to present barriers to trial participation^63-65. With the numbers available to us, we could find no significant relationship between demographics (age, sex, race, and education) and the decision to participate. Therefore, we have no evidence to expect a problem in recruiting from families of different sociodemographic characteristics. It should be noted, however, that the sample in this study consisted predominantly of white participants, whereas the expected racial breakdown for the proposed randomized controlled trial includes white participants at a rate of only 71%. Therefore, if there is an undetected relationship between race and decision to participate in a trial such as the one proposed, we might see different participation decisions in the actual trial with a different racial makeup.

There is evidence from this assessment that preferences for decisional control and a specific treatment will influence participation decisions and potentially result in selective enrollment. The responses of the adolescent group were fairly predictable. Over one-half had a preference for treatment, overwhelmingly preferring the observation arm; the majority of those with a treatment preference would refuse to participate. It is reasonable to expect adolescents to overwhelmingly prefer observation to bracing. Wearing a brace is not fashionable and can cause skin irritation and other inconveniences. Even when adolescents believe in the relative benefit of bracing in terms of long-term health and appearance, they often value current comfort over some future benefit. Therefore, those preferring observation would opt not to participate in order to avoid the risk of randomization to the bracing arm; instead, these adolescents might try to influence the parent and physician to opt for observation outside the trial.

This study examined two issues that have rarely been explored in orthopaedics. The first issue involves the unique challenges that arise when subjects of the proposed trial are minors, and both the parent and child preferences need to be considered, separately and jointly. Relatively little empirical information is available concerning the determinants of parental permission for participation in a randomized controlled trial⁶⁶. However, attitudes, beliefs, and values, such as the desire for decisional control, have been cited as barriers to participation. For example, Wiley et al., in a study of parents of children in cancer trials, found that parents who refused participation desired decisional control, commenting that they trusted their physician's judgment over that of computerized randomization⁶⁶. On the other hand, many parents feared choosing a cancer treatment that might later prove to be ineffective and preferred that randomization remove this dilemma. Thus, the anticipation of decision regret could explain the behavior of parents who consented to participation, as well as the desire of many parents to either share decision-making or to have the physician make the decision for them. Anticipatory regret could also explain the propensity for crossover found in the present study.

Parents, as proxy decision-makers, are not free to make the same decisions for their dependents as they might for themselves. Decisions made for children are expected to be made in their best interests, implying that parents may have less latitude to be as idiosyncratic or unconventional as they might be when making decisions for themselves⁶⁷. This may explain why parents' decisions about participation were not in accord with what might be expected. Strength of preference, or having a preference at all, was not strongly associated with a willingness to participate, although previous studies have shown that having a preference is a barrier to accepting randomization². The majority of parents preferred the observation arm, although we expected that parents would prefer the active treatment, especially since bracing is known as the standard of care.

The second issue rarely explored is what patients and parents expect from bracing in terms of effectiveness. Since the primary outcome of the proposed randomized controlled trial is the percentage of subjects reaching the surgical threshold (defined as a Cobb angle of ≥50°) in each treatment arm, this study assessed the reduction in surgical risk desired by the participants for them to find bracing superior to observation despite the inconvenience and potential side effects. The median required risk reduction was from 60% to 30%, indicating that parents felt the brace should reduce the risk of surgery by half. Few studies have described surgical outcomes for contemporary cohorts of patients managed with bracing and those managed with observation. The latest follow-up of Swedish patients³⁰ found a 23% surgery rate in the observation group compared with 0% in the braced group (a 100% risk reduction). A recent systematic review of surgical rates after bracing found a range from 0% to 44%⁶⁸, but the authors concluded that the current evidence is not of sufficient quality to inform clinicians or patients what the actual or potential risk reduction could be. Thus, the literature does not indicate whether bracing is able to meet the expectations of parents in the present study. It is interesting that the protocol committee assembled to plan the randomized controlled trial also agreed that the risk reduction should be ≥50% before they would declare bracing to be superior to observation.

It was assumed that subjects would believe in the physician's equipoise concerning the two treatment arms, and they would therefore assume that the physician was in favor of participation. The role of the physician has been shown to influence decisions concerning trial participation. Confidence in the doctor who recommended the trial was a principal reason for parents to allow their children to be randomized to a hypothetical drug trial⁶⁹. In general, patients who perceive their physicians to have affiliative communication styles are more likely to be satisfied with their care, and they are more likely to comply with physician recommendations, including participation in clinical trials⁷⁰. In a study of prospective participants in a cancer trial, those who believed the clinician was truly uncertain trusted the study and could accept randomization as an acceptable method for making treatment decisions⁷¹. There was no physician-subject contact in the present study. Perhaps in a true clinical situation, where the physician's equipoise could be more personally and directly communicated, participants would perceive that, by accepting randomization, they are indeed making a shared decision or making the decision desired by the physician. This perception may then result in higher participation rates, since the most frequent reasons for not wanting to participate were the desire to share decision-making or to have the physician make the decision.

The major limitations of this effort were its hypothetical nature, the relatively small sample size, and, potentially, the use of surrogates as subjects. What subjects say they would do in a hypothetical situation may not predict what they would do when faced with an actual enrollment decision, although stated willingness to participate has been correlated with actual participation decisions⁴⁴. Potentially, many of the factors studied could either be confounders or interact in ways that substantially affected the decision to participate, but the size of the sample did not allow for reliable tests of these effects.

One could argue that the preferences and participation decisions from the surrogate group should not be used to predict the responses of patients approached during actual trial recruitment. Many previous simulations of recruitment for a randomized controlled trial have used surrogates in the place of patients with the actual condition under study^40-43,45,72. There are two main reasons for the use of surrogates in place of actual patients. The first is to avoid interfering with actual treatment decisions and to avoid the bias that may result from already making the decision and experiencing the treatment. The second reason for the use of surrogates is logistical. It can be difficult to identify a representative sample of prospective participants, especially when the target population is not readily accessible or not composed of regular users of the health-care system¹. We sought scoliosis subjects at five different institutions. Most patients seen in these clinics had either undergone bracing or surgery or were offered treatment during that clinic visit. This left the small group of patients who were still between ten and fifteen years old but, because of the curve size or skeletal maturity, had not been offered treatment.

We anticipate that many of the patients who meet eligibility criteria when they are initially seen will have little firsthand knowledge of scoliosis and will not have an established relationship with the pediatric orthopaedist; thus, they will be very similar to the surrogates in this study. We expect that patients who are presented with the option of participating in the proposed randomized controlled trial on their first orthopaedic visit will respond similarly to the surrogate group. Conversely, the patients in the scoliosis group in this study had not yet been recommended for bracing or surgery, but they were presumably familiar with the disease, perhaps already had been exposed to bracing as the standard of care, and, biologically, may have been at lower risk for progression. If this is true, the patient group may have been less representative of the target population for the randomized controlled trial than was the surrogate group. We expect that subjects who become eligible for the study after one or more follow-up visits would make participation decisions similar to those of the patient group in this assessment. If surrogates had made up the entire sample, an overestimation of participation rates may have occurred, as the participation rate in the scoliosis group was one-third of that of the surrogate group. The patients and parents in the scoliosis group were found to have a relatively more passive orientation toward health care, as they were more likely to want the physician to make the treatment decision and were less likely to participate. These findings point to sick-role theory as suggested by Stiggelbout and Kiebert⁵³ in their comparison of patients and companions with respect to decision-making preferences. They posited that “the mere fact of being a patient leads to a shift in preference away from participation,” implying that, in general, people want to participate in health-care decisions but are less enthusiastic when they are sick.

One of the major tasks in planning a randomized controlled trial is to determine the target sample size. This is generally done through consideration of biological plausibility, clinical plausibility, clinical judgment, conventional wisdom, and feasibility⁴³. Our initial effect size of a 50% reduction in surgical rates was determined by our protocol development committee. The exercise presented to subjects in this study allowed us to estimate this figure from the perspective of potential consumers. Some studies have found that actual patients are willing to accept a smaller benefit from treatments than are doctors and nurses. We would then anticipate a difference in the required risk reduction between the scoliosis and surrogate groups, but instead we found both groups had a median required risk reduction of 50%. The agreement between the findings in the present study and the judgment of the protocol committee lend credibility to the power and sample size analysis to be included in the subsequent grant proposal.

Combining the estimates of randomization rate (25%) and required risk reduction (50%), the required sample size is approximately 450 subjects. The results of the present study indicated that recruiting subjects into a randomized trial of bracing compared with observation could be problematic. Joint agreement to participate was low, leading us to estimate that centers in the proposed randomized controlled trial will have to screen 100 families to find twenty-five who will agree to participate. Treatment preference was associated with a propensity for crossover for both parents and adolescents, indicating that if the subject were randomized to his or her nonpreferred arm, or experienced progression during observation treatment, the rate of nonadherence to the protocol could be quite high.

It is difficult to infer from these findings the extent to which selective enrollment might unduly affect the generalizability of the trial results. The sample size did not allow us to compare rates between different risk profiles within the scoliosis group, so we cannot anticipate whether risk, or its perception, would result in differential randomization. On the other hand, it is unlikely that altruism, treatment preference, or the desire to make a shared decision would have any direct effect on the relationship between bracing and curve progression. It is plausible that these factors could result in differing levels of adherence to the protocol, which could then have a substantial influence on the randomized controlled trial outcomes. For example, researchers have found a positive correlation between altruism as the motivation for trial participation and subsequent adherence to trial medication⁷³. It is hard to predict how the effect of altruism on compliance and the propensity for crossover will play out in the actual trial. These uncertainties only reinforce our decision to use an embedded temperature monitor (as used in a recent study from Texas Scottish Rite Hospital⁷⁴) to objectively determine the “dose” of brace wear actually received, regardless of the randomization assignment. The as-treated analysis will use hours of brace wear as the independent variable.

The findings in the present study also reinforce the important role that physicians have in promoting recruitment into a randomized trial. The surrogates were more likely to choose randomization rather than to share the decision with the physician, whereas parents in the scoliosis group, who presumably had an existing relationship with an orthopaedist, tended to either want to share the treatment decision or to have the physician make the decision. Several parents indicated they would agree to participate because (hypothetically) a physician had suggested the study. Therefore, recruitment efforts should allow for physician-subject interaction so that the physician's equipoise and commitment to the study are more directly and effectively communicated to the adolescents and their parents. This in and of itself could sufficiently address the issue of underenrollment.

This preference assessment provided invaluable data for the planning of the proposed randomized controlled trial. We encourage all researchers planning a large-scale randomized controlled trial to conduct a similar investigation.