Beckett RD, Loeser KC, Bowman KR, Towne TG. Intention to treat and transparency of related practices in randomised controlled trials of anti-infectives. BMC Med Res Methodol. 2016;16(1):106 doi.org/10.1186/s12874-016-0215-2. The purpose of a study is to determine the proportion of people in a group who can be expected to receive a particular treatment. Of course, those who do not complete the treatment cannot be expected to benefit. Thus, the proportion of responders among those who complete treatment provides an exaggerated estimate of the effect of treatment – this does not accurately reflect the positive effect that can be expected in clinical practice in those who are prescribed this particular treatment. The multivariate linear regression model showed two notable correlations. A more conservative limit of ITT subCI was associated with a lower proportion of the ITT population contained in the PP population for the treatment arm and a higher proportion of the ITT population in the PP population for the control group. These variables determine the proportion of the excluded population that would then affect the width ci as described below. The linear regression model was only an exploratory analysis for the following reasons.

First, the predictor methods used in the model were often not described in detail in the journal articles. For example, only 39% of studies described how they handled missing data. Second, many other factors may have contributed to the analysis being more conservative, such as. B the trend of absenteeism and non-compliance [11]. Data may be randomly absent or missing in terms of response to treatment [10, 11]. Non-adherence may also be related to response to treatment or study group if there were differences in side effects [10]. These factors cannot be captured from empirical evidence. Finally, the exclusion criteria for the ITT and PP analyses were heterogeneous across studies. As an exploratory analysis, univariate linear regression was used to estimate associations between study-level characteristics and the difference between the lower bound of the CI of the itT and PP analyses. Possible predictors were methods of processing missing data, risk of bias, and inclusion and exclusion criteria for TTI and PP populations as binary variables. Variables with univariate P < 0.2 were entered into a multivariate linear regression model.

Why these questions should be asked when talking about intention-to-treat (ITT) vs. Pro Protocol (PP)? Well, let`s start with some general definitions and explanations: Before our study, only two studies compared ITT and PP analyses. These two studies included 11 and 20 studies [12, 13], respectively, while our study included 154 studies. Ebbutt and Frith found larger CIs in pp analysis and otherwise no consistent pattern of differences in both directions between the two analyses [12]. In contrast, perhaps due to the larger number of studies in our systematic review, we found that the itT analysis had broader CIs and tended to be more conservative, a finding consistent with Brittain and Lin`s study [13]. Newcombe RG. Estimation of the interval for the difference between independent proportions: Comparison of eleven methods. Stat Med. 1998;17(8):873–90. doi.org/10.1002/(SICI)1097-0258(19980430)17:83.0.CO;2-I.

The co-primary endpoints were point estimate and lower AI. We converted all risk differences to standard ARR, which is calculated as the success rate in the treatment arm minus the success rate in the control arm, so a negative ARR means that the results favor the control arm and a positive ARR means that the results favor the treatment arm. Based on this alignment, the lower limit of CI can be interpreted as representing the worst plausible treatment effect for the treatment group. A conclusion of non-inferiority was based on a comparison of this lower limit of CI with the margin of non-inferiority (Fig. 1). Deviations that could be affected by actual treatment should not be used as an exclusion criterion: e.B. “Early study termination” may not be a good choice of exclusion criterion from PP analysis if this discontinuation is due to a lack of efficacy (and therefore related to the treatment received). The excluded population is defined as patients in the ITT population who were excluded from the PP population.

We chose the Agresti-Caffo method because it tends to have a more conservative CI width than the other two methods [25]. We also used the method described by Newcombe [26] to calculate ci as a sensitivity analysis. Porta N, Bonet C, Cobo E. Discord between reported intention-to-treat and pro-protocol analyses. J. Clin Epidemiol. 2007;60(7):663–9 doi.org/10.1016/j.jclinepi.2006.09.013. We included studies published in English that were identified as non-inferiority RCTs in humans and compared two or more systemic antibiotic regimens to treat a bacterial infection. Studies were included when the treatment and control arms were specific antibiotic regimens. Each arm of the study should have a different antibiotic diet. Montori VM, Guyatt GH principle of intention to treat.

CMAJ. 2001;165(10):1339–41. ITT analysis is considered more conservative (less likely to find a difference between groups) than PP analysis in superiority RCTs, as the estimated treatment effect can be diluted using ITT analysis by including participants who did not receive the intervention [5]. However, in non-inferiority studies, this dilution and tendency to make results similar in both treatment arms may lead to inappropriate claims of non-inferiority [6,7,8,9]. Following this line of thinking, PP analysis would be more conservative (less likely to explain non-inferiority) than ITT analysis and would be preferable as the primary analysis of non-inferiority studies [6]. Centre for Drug Evaluation and Research (CDER). Guidelines for bacterial pneumonia acquired in industry: development of drugs for treatment. 2014. www.fda.gov/media/75149/download.

Retrieved 8 June 2020. Orthoevidence. Why are researchers concerned about the “intention to treat” analysis? OE Original. 2019;2(9):1. Available from: myorthoevidence.com/Blog/Show/42 Recent studies have challenged the idea that PP analysis is more conservative in non-inferiority studies. Simulation studies identified scenarios where PP analysis was more conservative and other scenarios where ppD was not [10, 11]. However, there is little empirical evidence so far. One study found no significant differences between ITT and PP analyses in asthma studies [12]. Another study on antibiotic non-inferiority studies found a trend that ITT analysis might be more conservative than PP analysis, but could not draw firm conclusions [13].

To be included in this secondary analysis, studies must have reported both ITT and PP analyses and the results as a percentage of absolute risk differences. The most conservative approach between PP and ITT analyses was defined as one with the lowest limit of the smallest (most negative) CI, as the smaller limit is less likely to exclude a margin of non-inferiority. In a clinical trial (we are only talking about superiority studies here, because the situation is different in non-inferiority studies), we want to recognize a benefit of treatment A (e.B. verum) compared to treatment B (e.B. placebo). The aim is to refute that “treatment A is no better than treatment B (the so-called `null hypothesis`). This is equivalent to proving that “treatment A is actually better than treatment B” (this is how statistical tests work). The use of ITT analysis ensures that comparability between groups as obtained by randomization is maintained, maintains sample size and eliminates bias. .