This trial design is the gold standard for establishing the efficacy and short-term safety of an experimental agent, especially when the ultimate goal is approval from regulatory agencies. This trial design allows to test treatment arms with clinical equipoise and to determine the assay sensitivity (i.e., the ability to demonstrate effectiveness in the experimental arm versus the active comparator/placebo). Statistical analysis is simple and straightforward and usually involves assessment of effect size. However, the use of placebo or experimental agents with unknown efficacy when other effective treatments are available raises important ethical concerns, and the generalization of results to routine clinical care is questionable as a select group of patients are enrolled in clinical trials. This study design has been employed in several trials in children with JIA [9–12]. More recently, this design has been used for new drug treatment of systemic JIA such as tocilizumab and canakinumab. The consideration of placebo-controlled trials in systemic JIA may be justified because the disease has few therapeutic options, the placebo effect and the required sample sizes are expected to be low, and the placebo treatment duration is short.

This study design has been used in clinical trials of children with JIA [13–15] and implies the substitution of placebo by an active comparator (e.g., standard therapy), usually in the format of a non-inferiority, equivalence, or superiority study. This design has the advantage of allowing comparison with current-standard therapy and is more patient friendly because all subjects receive an active treatment. However, especially when the comparator drug is very effective, sensitivity can be difficult to evaluate. Double-dummy designs are often necessary in situations in which the active agent has a different dosing pattern than the comparator. The most important disadvantage is that very large sample sizes are needed, with the potential for exposing a large number of subjects to either an inefficacious or dangerous treatment.

This was originally proposed for use in pediatric rheumatology studies [16]. Eligible children are treated in an open-label fashion with the experimental therapy to be tested in the trial for a few weeks or months. After this first phase, only responders are randomized in a double-blind fashion either to continue the experimental therapy or to switch to placebo. In this segment of the study, called the double-blind withdrawal phase, patients who experience a disease worsening or recurrence are withdrawn from the double-blind withdrawal phase and usually re-treated with the experimental therapy in an open-label fashion. This approach has several advantages: a placebo-controlled segment is included in the double-blind withdrawal phase but only responders may be exposed to placebo. All subjects receive treatment with the experimental agent, and information from all subjects is used in the analysis of efficacy and safety. This study design is considered efficient because it minimizes the number of patients exposed to a potentially ineffective or unsafe experimental treatment, and the proportion of response in the open-label phase of the study more closely approximates routine clinical care than an RCT. On the other hand, the randomized withdrawal design may be affected by a bias toward responders (e.g., only responders are randomized), and the reduced amount of patient-years on placebo produces a potential for limiting validity of safety comparison with the experimental agent. The requirement for estimation of the response rate during the initial open segment as well as an estimate of the time to flare and the requirement of performing a complex evaluation of response during the study visit are also limitations of this approach. Other possible caveats include the use of nontraditional outcomes (e.g., time to flare or proportion who flare in the double-blind withdrawal phase) and the impracticality for use in treatments with very long duration of biologic effect, such as B-cell depleting antibodies. The withdrawal design has proven to be a very effective design and has been used in the majority of recent trials of biologic agents in children with JIA [17–22]. Data gathered by use of this design led to the approval of biologic agents for children with JIA by both the FDA and EMA. Another key issue in this trial design is the sample size calculation. For parallel RCT trials, it is based on expected ACR Pediatric 30 response (see next section) rates among those in the test group versus those in the placebo or active comparator agent. Conversely, in randomized withdrawal designs, sample size estimations for the pivotal phase of the trial are based on the expected rate of flare among those in the experimental group compared to placebo (Table 6.1).

The standardization of the clinical response assessment is fundamental for the efficiency of a clinical trial and for allowing the comparison of different therapeutic approaches. In the field of pediatric rheumatology, important steps forward in this direction have been made. Until the mid-1990s, the assessment of clinical response in JIA trials was not standardized, and multiple different outcome measures were used [23]; the amount of change that configured a clinically relevant improvement or worsening was not established. An important step forward was made in 1997, when the core set of outcome measures and definition of improvement in JIA were developed [24]. The core set includes the six variables (Table 6.2). The definition of improvement allows classifying patients as responders in a clinical trial if they demonstrate an improvement of at least 30 % from baseline in at least three of any six core set variables, with no more than one of the remaining variables worsening by more than 30 %. These criteria were endorsed by the American College of Rheumatology (ACR) and are currently known as the JIA ACR 30. Soon after their publication, these criteria became the gold standard for the assessment of response to therapy in JIA and are accepted by both the FDA and the EMA for all phase III trials in JIA seeking drug registration [25]. Recently, the ACR Pediatric 30 was adapted for use in clinical trials in systemic JIA, by adding, besides the six core set variables, the demonstration of the absence of spiking fever (38 °C) during the week preceding the evaluation [26, 27].

The ACR Pediatric 30 criteria were introduced when the therapeutic options for the management of JIA patients were limited, with low-dose methotrexate being the most effective treatment [10]. In a number of cases, even a moderate improvement was considered a success. In the following years, the use of higher doses of methotrexate [13] and the introduction of biologics markedly increased the possibility to achieve a better disease control. Therefore, besides the 30 % improvement in outcome variables, more stringent levels of improvement were defined with the same method: an improvement of at least 50 %, 70 %, 90 %, or 100 % in at least three of any six JIA core set variables, with no more than one of the remaining variables worsening by 30 %, identifies, respectively, the ACR Pediatric 50, 70, 90, and 100 response criteria. Most of the clinical trials conducted in the last decade have also reported these more stringent definitions of improvement [12, 17–20, 26, 27].

The definition of flare currently used in JIA is essentially the inverse of the ACR Pediatric 30 criteria for response [28]. This definition requires that there be at least a 30 % worsening in three of the six JIA core set parameters, with no more than one improving by more than 30 %. Because this definition employs percentages, patients with low numbers of joints with active joints or with limitation of motion may have an artificially high percent increase simply because the denominator for the calculation is so small. Thus, the definition has been modified in this situation. If either the number of joints with active arthritis or the number of joints with limitation of motion is used in the flare definition, there must be at least a two-joint increase in the number of active joints or those with limited range of motion. Also, if the physical or parent global scales are used in the flare definition, worsening ≥2 cm (on a 0–10 cm scale) is required. Additionally, for systemic JIA patients only, reappearance of spiking fever (suggested definition >38 °C, lasting for at least 2 days in the week preceding the evaluation) not due to infections would signify that a flare has occurred. These flare criteria have been validated and shown to be effective in limiting potential damage in subjects randomized to placebo or among those who lose response to the experimental agent.

The ACR Pediatric response criteria are a tool specifically devoted to the measurement of a change in disease activity. They are well suited for evaluating the efficacy of a new therapy in comparison to another therapy or a placebo in the setting of a clinical trial, where all enrolled patients are supposed to present a similar level of disease activity at baseline. However, the nature of their calculation does not enable the measurement of patients’ actual disease activity at the beginning or at the end of a clinical trial. Furthermore, response criteria do not allow discernment of whether one group of patients has more active disease than another group, in particular when the initial conditions are different. The comparison of one patient’s response with that of another patient may be misleading, because when the baseline conditions are different, an equal level of percentage improvement may correspond to a wide difference in the absolute change and in the absolute state at the final time point. Thus, in the interpretation of the results of a clinical trial, it must be considered that the evaluation of the sole percentage improvement may lead to some important information about the effectiveness of the medications under scrutiny to be missed.