Dissection of regional lymph nodes in cutaneous melanoma

Abstract

Natale Cascinelli and colleagues1Cascinelli N Morabito A Santinami M MacKie RM Belli F on behalf of the WHO Melanoma ProgrammeImmediate or delayed dissection of regional nodes in patients with melanoma of the trunk: a randomised trial.Lancet. 1998; 351: 793-796Summary Full Text Full Text PDF PubMed Scopus (563) Google Scholar rely too heavily on significance testing when they suggest that their trial may be interpreted as “finally settling the debate between advocates and opponents of elective lymph-adenectomy” for melanoma of the trunk. This conclusion implies that their trial was merely a test to establish whether elective lymphadenectomy is better than delayed surgery. This is too narrow a vision. Rather, we should regard their study as an estimate of the value of immediate surgery. Their results, unadjusted for covariates, show that we can be 95% confident that elective lymphadenectomy may, at worst, reduce 5-year survival by 2% but, at best, may improve it by 23% (I have calculated the 95% CI for the difference in survival from the published data). Thus, although we cannot be certain that elective lymphadenectomy improves survival, we cannot rule out the possibility that it will result in greater than 20% more patients surviving at 5 years, which hardly seems to settle the debate in favour of delayed surgery. In fact, adjustment for covariates, which the investigators do not provide for the 5-year survival data, should further favour immediate surgery. (The confidence interval for the adjusted hazard ratio which is reported is more difficult to translate into a clinically meaningful form.) The hazards of interpeting negative studies2Freiman JA Chalmers TC Smith Jr, H Keubler RR The importance of beta, the type II error and sample size in the design and interpretation of the randomized clinical trial: survey of 71 negative trials.N Engl J Med. 1978; 299: 690-694Crossref PubMed Scopus (1263) Google Scholar and the value of confidence intervals over significance testing3Rothman KJ A show of confidence.N Engl J Med. 1978; 299: 1362-1363Crossref PubMed Scopus (231) Google Scholar have been well described. We should avoid the erroneous view that significance testing provides a yes or no answer to the question of a treatment's value. A significant p value only indicates that the difference between two groups is not likely to be zero. It does not tell us the magnitude of the difference, which is what we need to know in order to balance benefits and risks. To determine the clinical significance of a trial we need to look at confidence intervals.4Berry G Statistical significance and confidence intervals.Med J Aust. 1986; 144: 618-619PubMed Google Scholar In this case, those confidence intervals are wide enough, I am sure, to lead many readers to question the clinical significance of the results. Reseachers need to consider carefully the sample size required to reduce confidence intervals to clinically meaningful proportions.5Goodman SN Berlin JA The use of predicted confidence intervals when planning experiments and the misuse of power when interpreting results.Ann Intern Med. 1994; 121: 200-206Crossref PubMed Scopus (476) Google Scholar Dissection of regional lymph nodes in cutaneous melanomaAuthor's reply Full-Text PDF