We investigate the sensitivity of the conversions from entrance surface dose (ESD) or kerma-area product (KAP) to effective dose (E) or to energy imparted to the patient (epsilon) to the likely variations in tube potential, field size, patient size and sex which occur in clinical work. As part of a factorial design study for chest and lumbar spine examinations, the tube potentials were varied to be +/-10% of the typical values for the examinations while field sizes and the positions of the field centres were varied to be representative of values drawn from measurements on patient images. Variation over sex and patient size was based on anthropomorphic phantoms representing males and females of ages 15 years (small adult) and 21 years (reference adult). All the conversion coefficients were estimated using a mathematical phantom programmed with the Monte Carlo code EGS4 for all factor combinations and analysed statistically to derive factor effects. In general, the factors studied behaved independently in the sense that interaction of the physical factors generally gave no more than a 5% variation in a conversion coefficient. Taken together, variation of patient size, sex, field size and field position can lead to significant variation of E/KAP by up to a factor of 2, of E/ESD by up to a factor of 3, of epsilon/KAP by a factor of 1.3 and of epsilon/ESD by up to a factor of 2. While KAP is preferred to determine epsilon, the results show no strong preference of KAP over ESD in determining E. The mean absorbed dose D in the patient obtained by dividing epsilon (determined using KAP) by the patient's mass was found to be the most robust measure of E.