Rules for proper conduct of a random control clinical trial (from Conclusions):
(1) The investigator shall have no preconceived opinion about the superiority of one or other arm, either at the start of the trial or during its conduct.
(2) The investigator shall remain in ignorance of any trend in results: interim analyses are bad ethically and statistically.
(3) The avoidance of bias is central to the proper conduct of random control trials. The investigator shall be unable to influence the allocation of patients. thus, certain methods of randomization are unacceptable.
(4) The investigator shall have no knowledge concerning the arm of the trial to which the patient has been allocated when judging the occurrence or non-occurrence of an event.
(5) No treatment may be given to any patient without his or her properly informed consent. No treatment may even be suggested which is not in the patient's interests.
(6) The control group shall receive the best standard treatment whereas the test group shall have the best treatment varied only by inclusion of the regimen being studied.
(7) Data shall at all times be treated with due deference. They shall never be "tortured until they confess." Bias can enter into the process of analysis as easily into the design and conduct of a trial.
(8) The principal requirements in presentation are accuracy, clarity and brevity.
Evaluation Domains:
(1) design and conduct: 15 items
(2) data analysis: 10 items
(3) presentation: 8 items
The responses are yes or no, with positive responses scored from 1 to 5 and negative responses scored 0.
subscore for design and conduct =
= SUM(points for 15 items)
subscore for analysis =
= SUM(points for 10 items)
subscore for presentation =
= SUM(points for 8 items)
total score =
= (subscore for design and conduct) + (subscore for analysis) + (subscore for presentation)
Interpretation:
• minimum subscore and total score: 0
• maximum design and conduct subscore: 50
• maximum analysis subscore: 30
• maximum presentation subscore: 20
• maximum total score: 100
• The higher the score, the closer the study is to the ideal.
Limitations:
• I am not sure why patient consent is only graded 1 point.
• There may be variation in scoring depending on the scorer's assessment of compliance. For example, one reviewer may mark references as inaccurate if there are several errors present, while another might mark as inaccurate references with a single page discrepancy.
• A possible modification would be to give multiple responses along a spectrum, with graded points accordingly.
