Is an RCT the Best Way to Investigate the Effectiveness of Nonoperative Management of Pediatric Appendicitis?
In the January 2015 issue of Annals of Surgery, Svensson et al1 report the results of their pilot randomized controlled trial (RCT) investigating nonoperative management of acute nonperforated appendicitis in children. The authors should be commended for systematically investigating this less invasive medical therapy in children; however, we wish to raise 2 concerns with their study: (1) the potential effects of preference bias on the trial, and (2) the lack of a comparable primary outcome between the 2 treatments.
Although considered by many the “gold standard” for evaluating medical interventions, the RCT may not be the optimal method for evaluating nonoperative management of appendicitis in children due to the potential negative effects of preference bias. Specifically, preference bias can impede recruitment in an RCT investigating interventions with different risk-benefit ratios due to patient or surgeon preferences for one intervention over the other.2,3 This may be especially relevant in a trial involving the emergency and pediatric settings.2,3 In addition, preference bias can alter the outcomes related to a treatment with patients assigned to their preferred treatment having greater treatment effects and those assigned to their nonpreferred treatment having worse outcomes.2,3
The potential effects of patient and family preferences on study recruitment and outcomes may be particularly relevant in a study of nonoperative management of pediatric appendicitis.4,5 First, patients and families with strong preferences may decline to participate in a randomized controlled trial, thereby hindering enrollment and limiting external validity.6 Second, patient-family preferences may affect the success of nonoperative management as it is predicated on the patient and family accepting an ongoing risk for recurrent appendicitis. With this type of participative intervention, there is a greater chance for motivation to influence outcome as the success of nonoperative management depends on the patient's and family's perceptions and willingness to make it succeed.6,7 Both recruitment and the success of nonoperative management may have been impacted by preference bias in the trial by Svensson et al. In their trial, only 40% of approached eligible patients enrolled and 25% of patients randomized to nonoperative management underwent appendectomy within 1 year at surgeon/parental discretion despite not developing recurrent appendicitis.1
Randomization in a clinical trial cannot safeguard against preference bias on the part of the patients and their families. When randomized trials are not feasible due to strong preferences or when only a small proportion of patients will accept randomization, then high-quality parallel group nonrandomized studies in which patients choose their treatment are a valuable alternative.3,8 This patient choice design may minimize preference bias by aligning the perceptions and desires of the family and patient with the treatment, and increase trial participation as patients are more likely to enroll if they can choose their preferred treatment.5,7 Our group has been investigating nonoperative management of uncomplicated appendicitis in children using a patient choice design. In our trial, the patient choice design allowed for a higher enrollment rate with 85% of approached eligible patients agreeing to participate in the trial.9 Despite not being randomized, the nonoperative management and surgery groups in our trial had similar demographics, preoperative characteristics, and preoperative diagnostic evaluations.9 Additionally, the patient choice design may have also minimized the negative effects of surgeon/parent preference bias on outcomes as there have been no appendectomies carried out without a clinical presentation consistent with recurrent appendicitis.9
A second concern with the RCT by Svensson et al is their primary outcome. In trials of appendicitis comparing nonoperative management to surgery, there is no clear primary outcome that should be used.