Abstract
ISPOR
A methodological
investigation to define a clinical
relevant cut-off point in the ordinal scale of
the EORTC QLQ-C30 questionnaire.
All the advisory board will
be included as co-authors
Objectives
The objective
of this analysis was to identify a clinical relevant cut-off point in
the EORTC QLQ-C30 ordinal pain score by comparing patient and clinician
reporting for the same symptom.
Methods
A retrospective
pooling of closed European Organisation for Research and Treatment of
Cancer Randomized Controlled Trials, where the symptom pain was scored
at baseline by the patient (EORTC QLQ-C30) and
the clinician [Common Toxicity Criteria (CTC)], allowed secondary analysis
to quantify and test the optimal cut-off point. The CTC was dichotomized
as 0,1,2 vs. 3,4; defined as a clinical relevant cut-off point for clinical
practice. Percent agreement with various dichotomizations of the QLQ-C30
pain scale was calculated, and McNemar’s test applied. Verification
of the accuracy and generalizibility of the findings was evaluated with
a validation set and by applying the same cut-off point on another symptom,
i.e. fatigue.
Results
Data at entry
of study were available for pain [number of trials (t)=8, number of
patients (n)=1214] and fatigue [t=5, n=1237]. Model and validation set
were obtained by splitting the dataset in half. Agreement between patient
and clinician dichotomized scores using different cut-off points for
the QLQ-scale produced the following percentage agreement and p values
for McNemar tests; median (<2.19 vs >2.19) (64%, p<.01), quartile
(<=3.0 vs >3.0) (81%, p=0.55), decile (<4.0 vs 4.0) (85%, p<.01).
These results indicate that the quartile split reflects best the dichotomized
CTC score. This was confirmed in the validation set (quartile cut-off
point: 82%, p=0.86). However, when the quartile cut-off was applied
to the QLQ-C30 fatigue scale, a significant difference (p<.01) between
patient and clinician results was found.
Conclusion
Our results indicate that a quartile split of the QLQ-C30 pain score is optimal. However, a single cut-point may not generalize to other QLQ-C30 symptoms; symptom-specific cut-points may be required.