Sources of between-study heterogeneity in the proportions of ER+, PR+, and HER2+ cases from meta-regression analyses.

aAdjusted for all other variables in the table except tumor grade (this variable was not included in the model because of potential for over-adjustment). The variable study design was not included in the final models because it was not associated with the frequency of ER+, PR+, or HER2+ in the crude or adjusted analyses.bMissing values were included as separate categories.cDefined according to the last year in which patient recruitment took place.dMean or median age of study cases at the time of breast cancer diagnosis; for studies that provided only age categories the mean was estimated from the mid-point and frequency of each category.eThese numbers are higher than the total number of North African studies included in the review (Table 3) because one study [26] presented separate ER+ and PR+ estimates for ages [55] presented separate ER+, PR+, and HER2+ estimates for inflammatory (IBC) and non-IBC locally advanced breast cancer (LABC).fp-values for interaction between region and timing of receptor testing for ER+: ppp = 0.10 in the crude analysis, p = 0.17 in the adjusted analysis; HER2+: p = 0.37 in the crude analysis, p = 0.52 in the adjusted analysis.Sources of between-study heterogeneity in the proportions of ER+, PR+, and HER2+ cases from meta-regression analyses.