If the studies are sufficiently homogenous in their design, outcome assessment and follow-up, we will conduct meta-analyses using a random effects model (DerSimonian and Laird [
23]) using the current version of R (R Foundation for statistical computing, Vienna, Austria) [
24]. We will use the relatively more conservative random effects model as we expect studies to have included a variety of patient participants (e.g. screening and symptomatic populations) and use radiologists of varying expertise. We will combine the percentage of patients with PICRC in each individual study to estimate a pooled prevalence, along with a 95% confidence interval (CI). If studies report variable lengths of follow-up preventing meta-analysis of prevalence at a 36-month time point, we will attempt to estimate a survival curve for PICRC from the individual study estimates using random effects survival meta-analysis [
25]. If available, we will compare the prevalence of proximal and distal PICRC using an odds ratio (OR) with 95%CI. We will assess heterogeneity between study-specific estimates using the inconsistency index (
I
2 statistic [
26]). If heterogeneity is considerable (
I
2 > 75%) and the
p value <0.1, we will not perform quantitative data synthesis [
18]. We will investigate between-study sources of heterogeneity using subgroup analyses by stratifying original estimates according to study characteristics, CTC technique and radiologist factors as described in the data extraction section above. We plan to investigate for small study effects (and possible publication bias) both qualitatively, by inspecting funnel plots [
27], and quantitatively, using the test proposed by Harbord [
28], although we will defer a decision regarding the suitability of these methods until the number of included studies and between-study heterogeneity is known. If there is evidence of possible publication bias and heterogeneity is sufficiently low (
I
2 < 25%) [
26], we will estimate PICRC rates using the trim and fill method [
29] to gain an approximation of the “missing” study rates and overall summary estimate and will present this as an estimate of the potential impact of missing studies.