In this randomized two-arm parallel group design, Vogt et al. evaluated two modes of combined therapy in children with monosymptomatic nocturnal enuresis (MNE). The authors claim that the “combined therapy proved effective in children with MNE after 6 months, with no statistically significant difference between the two different orders of treatment”. Here we show that the study was scientifically invalid. It thus violates issue 2 of the seven ethical requirements that should be fulfilled in clinical research . First, we show that the aim of the study cannot be achieved with the study design used. Second, we show that the study is completely under-powered to detect any reasonable difference between the treatment regimens applied in the study. Third, despite the small sample, we show that there is almost a significant difference between the treatment arms. Finally, we stress that authors should strictly follow the revised CONSORT statement  for reporting results, to improve the quality of reporting of randomized controlled trials.
(i) The USA Food and Drug Administration (FDA) has well-established principles for developing combined therapies. Specifically, it is required that ‘each component makes a contribution to the claimed effects’ (21 CFR, 300.50). As a result, the combination must be better than the single components . Vogt et al. randomized a child to one of two treatment arms. In treatment arm 1, desmopressin was given in the first 3 months, and the combined therapy of desmopressin and an alarm for the second 3 months. In treatment arm 2, an alarm was used as monotherapy in the first 3 months and the combined therapy in months 4–6. In summary, neither desmopressin nor alarm was used as monotherapy for the entire study period. Thus, this study design is not suited to evaluate the effectiveness of the combined alarm/desmopressin therapy.
(ii) No fewer patients than required and no more patients than required should be included in a clinical trial. Vogt et al. only reported a power analysis with 15 children per group. Thus, they did not describe their sample size calculation. For a binary endpoint, the differences between the treatment groups need to be very large if only 15 patients are recruited per treatment arm, and the power is very low.
To estimate the power of the study, we used the usual 5% test level. Success rates of alarm monotherapy have been reported to be 66–75% for a 3–6-month treatment period . Vogt et al. randomized a total of 43 children, 19 to treatment arm 1 and 24 to treatment arm 2. If we assume an unlikely high success rate of 99% for treatment 2, the power of the study was only 48–76% to detect this effect. In a slightly more conservative scenario with a success rate of 90% for treatment 2, the power decreases to ≈30%. In any case, this study does not have the power to conclude a difference between the two treatment arms.
(iii) The authors aimed to show a significant difference between treatment arms 1 and 2. According to FDA guidelines, an intention-to-treat analysis should be used. However, the authors excluded nine patients who were dry after 3 months, although these should be included because of their success. They further excluded four patients from analysis who discontinued. Interestingly, these four patients belonged to treatment arm 1, in which desmopressin was given in the first 3 months. The reasons for not continuing the trial were not given and we must therefore assume treatment failure in these patients. Table 1 reports the success rates if patients who were dry after 3 months and patients who discontinued are not excluded from analysis. With the intention-to-treat analysis, while the absolute risk difference between both treatment arms was only 9.8% in the unjustified analysis of Vogt et al., it is 21.7% (95% CI –8.1 to 52.9%; P = 0.174). In both analyses, there is an advantage to treatment arm 2, i.e., alarm first, and it is strong with the intention-to-treat analysis. This strong tendency for the alarm treatment arm is in line with current knowledge . In summary, there is a strong tendency that the alarm treatment arm is better than the desmopressin treatment arm. This is most likely not significant because the study is underpowered (see point ii).
|Treatment||Dry after 6 months, n (%)||Total|
|desmopressin/alarm||15 (63)||9 (38)||24|
|alarm/desmopressin||16 (84)||3 (16)||19|
(iv) Clinical trials should be reported according to the revised CONSORT statement to improve the quality of reporting of randomized, controlled trials . We have applied the 22-item check-point list to Vogt et al. (available from Dr Ziegler on request), and must conclude that their reporting is far from optimal. In detail, the authors do not comply to seven items of the check list (5, 7, 9, 10, 11, 19, 21). An additional five items (6, 12, 13, 17, 18) are addressed only in part. To give an example, standard adverse events for oral desmopressin include rhinitis, pharyngitis, headache, and increased cough . However, the authors did not state whether adverse or serious adverse events were ascertained and how often they were observed. In summary, more than half of all items listed in the revised CONSORT statement lack an adequate description. The study is inadequately reported and does not comply with the revised CONSORT statement.
To conclude, the principal investigator and the sponsor of a study are strongly encouraged to seek qualified personnel when conducting a clinical trial. This might lead to studies of substantially higher quality.