 |
|
|
APPENDIX 2 - DXA quality control
- At time of installation, (as part of commissioning procedure) or
after any major maintenance procedure or major software upgrade
- Machine calibration and testing by supplier. Accuracy
and precision evaluation
- In vitro: short-term precision
Scan manufacturer's recommended QC phantom 10-20 times. From these
measurements calculate the mean BMD of the Phantom and the coefficient
of variation (CV) of the results. (See Appendix
3).
CV = Standard deviation x 100%/Mean
- In vivo: short-term precision
On each machine, obtain voluntary followup scans within a short
timeframe (maximum 1 month), or duplicate scans on the same day,
on at least 28 subjects. The volunteers should be in an age range
appropriate to the generally scanned population and should give
informed consent. From these values obtain the in-vivo short-term
precision for BMD (See Appendix 4).
- Ongoing quality control procedures
Calibration and quality control according to manufacturer's specifications.
The QC phantom shall be scanned at least twice weekly (and preferably
daily) using the same scanning parameters. This phantom is not the daily
calibration phantom, but is an anthropomorphic (or quasi-anthropomorphic)
phantom recommended by (or at least acceptable to) the manufacturer.
The QC phantom data shall be recorded and checked for medium-term precisional
error and systematic bias using the appropriate statistical analysis.
This analysis may include one or more of the following tests: Multi-rule
Shewart Chart analysis (Appendix 5), CUSUM (Appendix
6), and Running Mean. Further information to assist in the choice
of statistical methods (particularly regarding reporting of results)
is given in Appendix 3. The ongoing QC program
must include a regular maintenance schedule, with periodic testing of
accuracy using a suitable QC phantom, as recommended by the manufacturer.
This phantom may be the same as that used for the regular (at least
twice weekly) QC. All interventions by the maintenance/repair provider
shall be recorded so that variations in QC parameters may be correlated
with breakdowns and repairs if necessary - partly to facilitate retrospective
analysis and correction of BMD data, if this should, with hindsight,
be deemed necessary.
- Long term Quality Control
Examination of trends in QC parameters (particularly BMD) over periods
of years is a necessary precaution for establishing the integrity of
long term studies of individual patients, or for the reporting of results
from clinical trials. The statistical methods used to examine long term
trends are the same as those use for medium term studies (see above).
However, it should be noted that statistical analyses become more powerful
and illuminating, with the acquisition of larger data groups - so that
analyses are worth periodically repeating as the QC history of a machine
accumulates.
Each machine has specific operating characteristics which leads to some
QC parameters being more important than others, depending on the circumstances.
For example, experience with Hologic pencil beam systems has shown that
variations in the parameters "k" and "d0" are frequently
more sensitivity to the onset of malfunctions than are BMD, BMC or Area.
Such information should be incorporated into the operations manual for
a machine.
- Use of internationally accepted anthropomorphic
standards in ongoing multi-centre QC studies
It is likely that some form of multi-centre participatory QC will eventually
be either strongly recommended or mandatory. The use of such standards
allows the BMD of a particular centre to be expressed as a machine-
and location-independent "standardised" BMD (sBMD), which
is much more amenable to allowing comparisons between centres in multi-centre
clinical trials.
However, sBMD is not currently recommended to compare the measurements
of the same patients between machines or between centres.
|
|
