A 1,000 – 1,500 word limit exists
i) What is the research question?
In the treatment of patients with Kawasaki disease (KD), is the addition of corticosteroid therapy to standard therapy recommended over standard therapy alone?
KD, an acute autoimmune disease that most commonly affects children and infants, results in the inflammation of the medium-sized blood vessels throughout the body. In rare and extreme cases, KD can affect the heart by causing coronary artery aneurysms and other abnormalities. Conventional primary treatment of KD involves the administration of intravenous immune globulin (2 g/kg) and aspirin (80-100 mg/kg) each day until fever subsides. Newburger et al. (2007) have set out to confirm whether primary treatment with intravenous methylprednisolone therapy (at 30 mg/kg) could, as previous retrospective studies have suggested, further reduce the risk of coronary-artery abnormalities.
ii) Is it concerned with the impact of an intervention, causality, or determining the magnitude of a health problem? (2+1= 3 marks)
The study is concerned with the impact of an intervention on a health problem; specifically, it is concerned with the impact of primary corticosteroid treatment on the coronary-artery outcomes and incidence of adverse events in KD patients.
i) What is the study type?
The study type is randomized controlled trial (RCT). RCTs are commonly used in clinical research trails to evaluate the effectiveness of new treatments on health-related biomedical outcomes.
ii) Is the study type appropriate to the research question? (1+1=2 marks)
Since the study in question is evaluating the effects of a medical intervention on health-related outcomes, this study type can be considered appropriate.
i) What is the reference population and source population?
The reference population is children with acute KD. The source population, from which the study participants are drawn, is comprised of 199 patients from eight centers within the Pediatric Heart Network in North America.
ii) Is there any selection bias?
Thanks to randomization, there is no selection bias in this study. There is no significant difference between the participants making up the control group versus the participants making up the treatment group. Furthermore, there is no significant difference between the sample of participants involved in the study and the general population to which the study results are expected to apply.
iii) If yes, does this threaten the external validity of the study? (1+1+1=3 marks)
There is no selection bias, so the external validity is not threatened.
i) Was the assignment of patients to treatments randomised?
Yes.
ii) Do you think the randomisation process was adequate? (1 + 2=3 marks)
Yes, I believe the randomization process was adequate because it minimized at least two sources of bias – allocation bias and confounding. In a randomized trial, the study participants are randomly allocated to different study groups; in this case, they are randomly allocated to either the treatment group or the placebo group. Since all of the participants involved in this study have an equal chance of ending up in one or the other group, the study has effectively eliminated allocation bias. Confounders, which are lurking variables that are not accounted for but which nevertheless have an effect on the outcome of a study, are also controlled for by randomization. Since randomization distributes the known and unknown confounding variables evenly among the treatment and placebo groups, the impact of confounding bias is minimized in this study.
Was blinding done in this study? If yes, who were blinded? (2 marks)
Yes. This study implemented a double-blind technique. This means that not only do the study participants not know whether they have been assigned to the treatment group or the control group, the researchers themselves do not know whether each individual participant belongs to the treatment group or the control group. This technique reduces two further sources of bias – ascertainment bias and observer bias. Participant ascertainment bias is a bias that results when the participants’ knowledge of whether they are receiving medication or a placebo subconsciously influences their reporting of their health outcomes. Observer bias, on the other hand, results when experimenters’ knowledge of whether participants are receiving medication or a placebo subconsciously influences their reporting of patients’ health outcomes.
What are the outcome measures and how are they measured? (2+2= 4 marks)
The primary outcome measures were coronary-artery outcomes and number of adverse events at 1 week and 5 weeks after randomization. Coronary-artery outcomes were the diameters, expressed as z-scores, of seven coronary artery segments, including in the left main coronary artery (LMCA), left anterior descending coronary arteries (LAD), and right coronary arteries (RCA). In addition to the adjusted z-scores, the absolute dimensions, as well as the changes in dimensions, of the seven coronary artery segments were measured. In addition, coronary artery outcomes were classified as the absence or presence of coronary artery aneurysms.
Adverse events were classified according to their severity and expectedness, as well as their likelihood of being attributable to IV methylprednisolone or placebo. Measured adverse events included shock and respiratory failure, profound sensorineural hearing loss, negative blood cultures, possible nonocclusive thrombus in the RCA, anaphylaxis to IV immune globulin, hypotension, and an episode of hypokalemia.
Secondary outcome measures also included the presence or absence of retreatment, the time to discharge from the hospital, and the total number of days of fever. Other laboratory measures, including haemoglobin and C-reactive protein levels, were also considered.
i) Are statistical tests considered?
Yes; statistical tests were used to compare the distributions of data between the treatment group and the placebo group.
ii) Were the tests appropriate for the data?
Yes. Where data was found to be normally distributed, a t-test of significance was used for continuous variables; if the data was not normally distributed, a Wilcoxon’s rank-sum test was used. In addition, a log-rank test was used to compare the survival distributions, over the time interval that lasted from patients’ admission to the hospital to discharge, of the treatment and placebo groups.
iii) Were patients analysed in the groups to which they were randomised?
Yes. The results were based on the groups as they were originally assigned.
iv) Do the conclusions drawn follow logically from the results of the analyses? (4 marks)
Yes. The conclusions regarding the significance of results follow directly from the statistical values determined in the analysis.
Are the results clinically or socially significant?
No. The researchers determined that the inclusion of intravenous methylprednisolone therapy for the primary treatment of Kawasaki disease does not significantly reduce the incidence of adverse events or significantly improve coronary artery outcomes.
Was the sample size adequate to detect a clinically or socially significant result? (2+2=4 marks)
Yes. The sample size of the study was 199 patients, of which 101 were randomly assigned to the treatment group, and 98 were randomly assigned to the control group. However, the researchers calculated that they needed a sample size of at least 194 patients in order to achieve a significance level of 0.05. Since the sample size of the study slightly exceeds this minimum size, the sample size can be considered adequate.
What conclusions did the authors reach about the study questions? Do you accept the results of this study? (2+ 3 = 5 marks)
In response to the research question, the authors found that they were unable to recommend adding corticosteroid therapy to the standard therapy regimen for the treatment of Kawasaki disease. Statistical analysis did not result in significant differences in coronary artery outcomes, measured in terms of z-score distributions, between treatment groups or placebo groups, at either 1 week after randomization or 5 weeks. Of the seven coronary artery segments for which z-score distributions were computed, only one segment, the posterior descending artery, was found to have a statistically significant difference in mean diameter between the two groups. Furthermore, the percentage of patients with coronary artery abnormalities did not achieve a statistically meaningful difference between treatment and control groups. In terms of the total number of days spent in the hospital or with fever, there was no statistically significant difference between the groups. Neither was there a statistically significant difference in the number of episodes of retreatment, or the percentage of patients requiring retreatment, between the two groups. The number of patients experiencing adverse events also did not differ significantly between the groups. I believe the data to be thoroughly assessed and the research methodology to be free of bias, and so I accept the results of this study.