Editor—In their survey of all randomised controlled trials published in 1997 in four major medical journals, Hollis and Campbell found that only 48% of the reports explicitly mentioned intention to treat analysis.1 In a considerable proportion it was insufficiently described and sometimes inadequately applied. Their results are confirmed by our assessment of all randomised controlled trials published between 1993 and 1995 in the same four journals.2 In addition to our assessment of ethical issues, we calculated the proportion of randomised controlled trials reporting intention to treat analysis in accordance with different descriptive and methodological characteristics.
In our review of 608 randomised controlled trials, we found that 290 of the trials (47.7%) explicitly mentioned that they applied the principle of intention to treat analysis. The reporting of this issue increased slightly between 1993 and 1995 (although the increase was not significant). Trials with a greater number of participants and those funded by the pharmaceutical industry were more likely to report the application of the intention to treat principle (table). In the multivariable logistic regression analysis, when we controlled for the general characteristics previously described, we found that trials with survival of patients as the principal outcome were more frequently reported to follow the intention to treat principle. In addition, those randomised controlled trials that gave no information about sample size were less likely to report the use of this principle (table). Randomised controlled trials not reporting the number of withdrawals or losses to follow up and those not reporting information about compliance with treatment were also less likely to report the intention to treat principle, although these results were not significant.
Our data support the relation between a higher methodological quality of the trials and the reporting of the intention to treat analysis. Our results reinforce the conclusions of Hollis and Campbell that the application of this principle still needs to improve because it seems that there has been no improvement between 1993 and 1997.1 A joint effort of editors and researchers is needed to meet the CONSORT guidelines3 and the authors' recommendations favouring intention to treat analysis.1 A better quality of reporting will help readers to assess the design, conduct, and analysis of randomised controlled trials more critically.
Table.
Reporting of intention to treat analysis in published clinical trials (1993-5)
| Total No | No (%) reporting intention to treat | Crude odds ratio for not reporting the use of intention to treat (95% CI) | Logistic regression multivariable model adjusted odds ratio (95% CI) | Odds ratio (95% CI) adjusted for descriptive characteristics | |
|---|---|---|---|---|---|
| All sample | 608 | 290 (47.7) | |||
| Descriptive characteristics of the trials | |||||
| Journal: | |||||
| N Engl J Med | 219 | 122 (55.7) | 1 | 1 | |
| JAMA | 81 | 37 (45.7) | 1.50 (0.90 to 2.50) | 1.64 (0.92 to 2.92) | |
| BMJ | 105 | 44 (41.9) | 1.74 (1.09 to 2.79) | 1.80 (0.96 to 3.39) | |
| Lancet | 203 | 87 (42.9) | 1.68 (1.14 to 2.46) | 1.54 (0.95 to 2.50) | |
| Year of publication: | |||||
| 1995 | 211 | 108 (51.2) | 1 | 1 | |
| 1994 | 195 | 92 (47.2) | 1.17 (0.79 to 1.73) | 1.20 (0.78 to 1.86) | |
| 1993 | 202 | 90 (44.6) | 1.30 (0.89 to 1.92) | 1.14 (0.73 to 1.77) | |
| Country of authors: | |||||
| Europe (except United Kingdom) | 164 | 86 (52.4) | 1 | 1 | |
| United Kingdom | 127 | 61 (48.0) | 1.19 (0.75 to 1.90) | 0.90 (0.52 to 1.55) | |
| United States | 240 | 120 (50.0) | 1.10 (0.74 to 1.64) | 1.24 (0.75 to 2.07) | |
| Other | 77 | 23 (29.9) | 2.59 (1.45 to 4.60) | 2.59 (1.38 to 4.85) | |
| Main specialty of authors: | |||||
| Medical specialties | 432 | 213 (49.3) | 1 | 1 | |
| Surgery or medical-surgical | 106 | 45 (42.5) | 1.32 (0.86 to 2.02) | 1.23 (0.76 to 1.99) | |
| Intensive or emergency care | 37 | 18 (48.6) | 1.03 (0.52 to 2.01) | 1.08 (0.51 to 2.29) | |
| Public health | 16 | 7 (43.8) | 1.25 (0.46 to 3.42) | 1.60 (0.54 to 4.74) | |
| Other | 17 | 7 (41.2) | 1.39 (0.52 to 3.72) | 1.29 (0.44 to 3.82) | |
| Number of participating subjects: | |||||
| >500 | 171 | 109 (63.7) | 1 | 1 | |
| 51 to 500 | 322 | 167 (51.9) | 1.63 (1.11 to 2.39) | 1.68 (1.12 to 2.53) | |
| ⩽50 | 115 | 14 (12.2) | 12.66 (6.68 to 24.10) | 12.43 (6.24 to 24.36) | |
| Source of funding: | |||||
| Pharmaceutical industry | 206 | 129 (62.6) | 1 | 1 | |
| Public agency | 165 | 73 (44.2) | 2.11 (1.39 to 3.20) | 2.11 (1.34 to 3.34) | |
| Other | 126 | 52 (41.3) | 2.38 (1.52 to 3.75) | 2.01 (1.22 to 3.30) | |
| Not reported | 111 | 36 (32.4) | 3.49 (2.14 to 5.68) | 2.35 (1.34 to 4.10) | |
| Methodological characteristics of the trials | |||||
| Outcome: | |||||
| Survival | 142 | 104 (73.2) | 1 | 1 | |
| Other | 466 | 186 (39.9) | 4.12 (2.72 to 6.24) | 2.86 (1.77 to 4.60) | |
| Sample size estimation: | |||||
| Shown | 281 | 167 (59.4) | 1 | 1 | |
| Not shown | 327 | 123 (37.6) | 2.43 (1.75 to 3.37) | 2.28 (1.55 to 3.37) | |
| Compliance with treatment: | |||||
| Stated | 532 | 261 (49.1) | 1 | 1 | |
| Not stated | 76 | 29 (38.2) | 1.56 (0.95 to 2.56) | 1.71 (0.98 to 2.99) | |
| Reporting follow up or withdrawals: | |||||
| Reporting the number of patients withdrawn or lost to follow up | 194 | 100 (51.5) | 1 | 1 | |
| Not giving information about number of patients lost to follow up | 414 | 190 (45.9) | 1.25 (0.89 to 1.76) | 1.45 (0.98 to 2.14) | |
A higher odds ratio means a higher probability of not reporting the use of the intention to treat principle.
References
- 1.Hollis S, Campbell F. What is meant by intention to treat analysis? Survey of published randomised controlled trials. BMJ. 1999;319:670–674. doi: 10.1136/bmj.319.7211.670. . (11 November.) [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Ruiz-Canela M, Martínez-González MA, Gómez Gracia E, Fernández-Crehuet Navajas J. Informed consent and approval by institutional review board in published clinical trials. N Engl J Med. 1999;340:1114–1115. doi: 10.1056/NEJM199904083401412. . (Erratum N Engl J Med 1999;341:460.) [DOI] [PubMed] [Google Scholar]
- 3.Begg C, Cho M, Eastwood S, Horton R, Moher D, Olkin I, et al. Improving the quality of reporting of randomized controlled trials. The CONSORT statement. JAMA. 1996;276:637–639. doi: 10.1001/jama.276.8.637. [DOI] [PubMed] [Google Scholar]