Reporting of the clinical relevant dose to organs at risk (OR) and other normal tissues is crucial in trials and protocols where the aim is to assess late complications and to increase the therapeutic ratio for external beam radiotherapy. The dose distribution in normal tissues and ORs are, however, most often heterogeneous, at least when more than two opposing beams are applied. To decide the most clinical relevant dose with respect to late occurring complications is therefore not a straight forward problem. In this work we discuss what parameters characterise the dose-volume-histogram (DVH) best by calculating normal tissue complication probabilities (NTCPs) by using the Lyman model and various sets of statistical parameters drawn out from the DVHs. These NTCPs are compared to NTCPs calculated from the full DVHs, when the sets of parameters are evaluated. Our calculations indicate that the NTCP based on the Lyman model is best correlated to the Dmax value for serially organised tissues such as the spinal cord. For organs, described largely as tissues organised in parallel, the Dmedian or Dmean of the DVH may be applied. Our calculations reveal that Dmean is the parameter of choice when Dmedian is quite small, but when the two parameters approach each other. Dmedian will be a better choice, using a unity volume fraction. For ORs characterised by a mixed serial and parallel functional structure, as the heart, neither Dmax, Dmedian nor Dmean may predict the actual NTCP.