Halpern and Preston used a quality score to evaluate the quality of clinical trials. This was a modified version of the quality score developed by Detsky et al. Halpern, Preston and Detsky are from the University of Toronto.
Parameters:
(1) randomization
(2) description of randomization
(3) concealment of treatment group from investigators
(4) description of outcome measures
(5) objectivity of outcome measures
(6) concealment of treatment from assessors (blinding)
(7) description of inclusion and exclusion criteria
(8) number of patients excluded
(9) description of treatment for treatment group
(10) description of treatment for control group
(11) statistical test described and p value given
(12) statistical analysis appropriate
(13) analysis performed if negative test
(14) sample size calculation
Parameter |
Finding |
Points |
patients assigned randomly |
no |
0 |
|
yes |
1 |
randomization adequately described |
no |
0 |
|
partially |
1 |
|
yes |
2 |
treatment group concealed from investigator |
no |
0 |
|
yes |
1 |
adequate description of outcome measures |
no |
0 |
|
yes |
1 |
outcome measures objective |
no |
0 |
|
partially |
1 |
|
yes |
2 |
assessors blinded to treatment group |
no |
0 |
|
yes |
1 |
inclusion and exclusion criteria well defined |
no |
0 |
|
partially |
1 |
|
yes |
2 |
number of patients excluded given and reasons logical |
no |
0 |
|
partially |
1 |
|
yes |
2 |
full description of therapy in treatment group |
no |
0 |
|
partially |
1 |
|
yes |
2 |
full description of therapy in control group |
no |
0 |
|
partially |
1 |
|
yes |
2 |
statistical test stated |
no |
0 |
|
yes |
1 |
p value reported |
no |
0 |
|
yes |
1 |
statistical analysis appropriate |
no |
0 |
|
partially |
1 |
|
yes |
2 |
calculation of confidence intervals or post hoc power calculations |
positive trial |
0 |
|
no for negative trial |
0 |
|
yes for negative trial |
1 |
sample size calculation before the study |
no |
0 |
|
yes |
1 |
where:
• In the original report by Detsky et al, the statistical test and p value are 2 separate items. In the table in Halpern and Preston they are listed together. However, if you add the values for the items together, you get maxima of 20 and 21 (and not the 21 and 22 listed). The only way to get the maxima reported is to split the 2 items out.
• The choices for the description of randomization were adequate, partial and inadequate.
• Logical reasons for exclusion included being ineligible, refused consent, etc.)
total quality score =
= SUM(points for all 4 items)
Interpretation:
• minimum total score: 0
• maximum total score for a positive trial: 21 (if p value a separate item)
• maximum total score for a negative trial: 22 (if p value a separate item)
• The higher the score the better the quality score.