Fig 1.
Illustration of the study design.
Fig 2.
Judgment of experts with respect to the egg burden and treatment efficacy of 2 clinical trial arms.
The labels below the bars denote the page and question (1: question at the top of the page, 2: middle, 3: bottom) in the example questionnaire presented in the S1 File. Numbers above bars represent the number of experts with a valid response (i.e. excluding “don't know” responses). Abbreviations: Q: Question; p: page. Top panel: baseline, middle panel: follow-up, bottom panel: efficacy. a) consensus agreement (absolute majority criterion—more than 50% of experts favor one arm) b) arm pair excluded, experts did not favor any arm c) excluded from the sensitivity analysis (number of trial participants in 1 arm below 30).
Fig 3.
Percentage agreement between experts and different means.
Raw percentage agreement between expert opinion and the calculated means for egg burden at baseline and follow-up and drug efficacy (superiority of a certain trial arm). Both, expert opinion and calculated means were dichotomized into 'A > B' and 'B > A'. Number of trial arm comparisons N: left panel: bl = 22, fu = 22, ef = 23; right panel: bl = 7, fu = 14, ef = 16. AM: arithmetic mean, GM: geometric mean, Hö: Hölder mean, Le: Lehmer mean, Wi: winsorized mean, tr: truncated mean. Numbers behind Hö/Le indicate parameter p, numbers behind Wi/tr denote proportion discarded/replaced. The rank denotes the rounded row mean rank. All: simple majority definition, consensus: absolute majority criterion, more than 50% of experts favor one arm, i.e. only those comparisons marked with footnote a) in Fig 2 are considered. S2 File explains how Fig 2 and Fig 3 are related.
Fig 4.
Relationship between the calculated difference among 2 trial arms estimated by different means and experts' rating scores.
The symbols in the first 3 panels show the association between the rater scores and the differences in egg counts or egg reductions between trial arm pairs calculated by 4 different means (different means are represented by different colors). The lines represent the corresponding loess smoothing lines. The bar plots at the top show the experts’ rating scores in the same way as in Fig 2. Some bar plots were placed at the bottom to avoid over plotting. Note, that rater scores (and bar plots) which favored arm B have been converted to favor arm A, e.g. a rating score of 4 would be converted to a score of 2 (a score of 3 indicates no difference between the trial arms). In 3 comparisons at follow up (numbered 1 to 3 in the top right panel) the estimates were especially strong diverging. The corresponding raw data are presented as strip plot and histogram in the bottom right panel. S2 File explains how Fig 2 and Fig 4 are related.
Fig 5.
Relationship between rater scores, means and cure rates.
Differences between ERRs and CRs in percentage points. Lines and shaded areas represent the loess smoothing line and the corresponding 95% confidence band. Grey crosses and the dotted line represent the experts' score and its corresponding loess smoothing line.