Refitting the predictor variables included in a model in a new cohort usually exaggerates its calibration performance.