In this critical review of the current practice of patient dose verificatio
n, we first demonstrate that a high accuracy (about 1-2%, 1 SD) can be obta
ined. Accurate in vivo dosimetry is possible if diodes and thermoluminescen
ce dosimeters (TLDs), the main detector types in use for in vivo dosimetry,
are carefully calibrated and the factors influencing their sensitivity are
taken into account. Various methods and philosophies for applying patient
dose verification are then evaluated: the measurement of each field for eac
h fraction of each patient, a limited number of checks for all patients, or
measurements of specific patient groups, for example, during total body ir
radiation (TBI) or conformal radiotherapy. The experience of a number of ce
nters is then presented, providing information on the various types of erro
rs detected by in vivo dosimetry, including their frequency and magnitude.
From the results of recent studies it can be concluded that in centers havi
ng modern equipment with verification systems as well as comprehensive qual
ity assurance (QA) programs, a systematic error larger than 5% in dose deli
very is still present for 0.5-1% of the patient treatments. In other studie
s, a frequency of 3-10% of errors was observed for specific patient groups
or when no verification system was present at the accelerator. These result
s were balanced against the additional manpower and other resources require
d for such a QA program. It could be concluded that patient dose verificati
on should be an essential part of a QA program in a radiotherapy department
, and plays a complementary role to treatment-sheet double checking. As the
radiotherapy community makes the transition from the conventional two-dime
nsional (2D) to three-dimensional (3D) conformal and intensity modulated do
se delivery, it is recommended that new treatment techniques be checked sys
tematically for a few patients, and to perform in vivo dosimetry a few time
s for each patient for situations where errors in dose delivery should be m
inimized. (C) 1999 Elsevier Science Inc.