Skip to main content
. 2005 Jul 2;331(7507):19. doi: 10.1136/bmj.38488.385995.8F

Table 3.

Multivariate analysis of study initiation, completion, and publication*

Study initiation (n=649)
Study completion (n=565)
Study publication (n=248)
Study characteristics No Odds ratio (95% CI) No Odds ratio (95% CI) No Odds ratio (95% CI)§
Type of study:
Others
509
1.0
432
1.0
NA
Phase I 140 3.09 (1.35 to 7.06) 133 3.30 (1.34 to 7.84)
Funding:
None
54
1.0
39
21
Mixed
42
13.70 (1.71 to 109.76)
37
21
Private
471
2.56 (1.28 to 5.14)
418
167
Public 82 4.05 (1.50 to 10.94) 71 39
Adverse events:
None
NA 448
1.0
194
Adverse effects observed 117 0.52 (0.29 to 0.94) 54
Results:
Inconclusive results
NA NA 44
1.0
Confirmatory results
188
4.59 (2.21 to 9.54)
Invalidating results 16 0.44 (0.10 to 1.86)
Scope:
National
530
1.0
312
182
1.0
Multinational 119 3.07 (1.27 to 7.40) 253 66 2.25 (1.14 to 4.41)
Design:
Experimental
558
490
212
1.0
Non-experimental (descriptive/observational) 91 75 36 2.83 (1.13 to 7.06)
Analysis:
No interim analysis
495
487
203
1.0
Interim analysis 86 78 45 2.21 (0.98 to 4.95)

NA=not applicable (when testing a variable was a nonsense—for example, adverse effect could never explain initiation of a study as it occurs after initiation).

*

Adjusted by using forward stepwise regression on the subset of variables linked at the 0.25 level with each criterion; odds ratio of 1.00 assigned to reference category.

Variables tested: sponsor status, investigator status, revision (yes/no), planned duration, scope of recruitment (national/international), funding, phase (phase I/others, including observational), design (experimental v non-experimental).

Variables tested: sponsor status, investigator status, revision (yes/no), planned duration, interim analysis, funding, phase (phase I/others), design (experimental v non-experimental), adverse effects.

§

Variables tested: interim analysis, design (experimental v non-experimental), with or without direct benefit for patient, scope of recruitment (national/international), results.

Not included in final model because multivariate statistical significance was >0.25 to enter or >0.15 to stay.