OCVC-BIF OCT substudy: adjunctive side branch DCB after kissing balloon

Summary

OCVC-BIF randomized 299 patients with bifurcation lesions, after main vessel stenting and kissing balloon, to adjunctive side branch DCB (n=149) or no DCB (n=150). The OCT substudy included 221 patients with analysable follow-up OCT. 198 had paired lumen volume.

The main endpoint was the change in side branch lumen volume at 9 months:

Within groups, the change was +0.135 mm³ (95% CI −0.984 to 1.254; p = 0.811) with DCB and −2.149 mm³ (p = 0.028) without DCB. Baseline side branch lumen volume was about 33 mm³.

Strengths

• The discordance between the parametric and nonparametric analyses appears in the abstract and in the central illustration.
• High-resolution serial OCT, with strut healing analysis in the main vessel.
• Prespecified nonparametric sensitivity analysis.

Methodological issues

• Skewed distribution. In the no-DCB group the mean was −2.15 mm³ and the median −0.65 mm³, with a tail reaching −50 mm³. The mean does not describe the typical patient. The Wilcoxon test gives p = 0.093. The cumulative distribution curves nearly overlap except at the extremes: a few patients carry the result. Variances differ (ratio about 2.8), but with nearly equal n (100 vs 98) this matters little.
• Two OCT endpoints without a hierarchy. Change in ostial lumen area and in lumen volume were declared without alpha allocation. Ostial area was null (p = 0.39), and the ostium is where side branch restenosis occurs. With alpha split between the two endpoints (0.025), p = 0.042 does not pass.
• Multiple comparisons. Tables 4 and 5 and the subgroup figure carry about 45 unadjusted comparisons. The probability of at least one false positive at alpha 0.05 is about 90%. Bonferroni over 45 tests gives alpha 0.0011.
• No sample size of its own. The sample size comes from the parent trial, computed for angiographic restenosis (20% vs 7%). With about 100 per arm and a pooled SD of 7.8 mm³, the minimum detectable difference at 80% power is about 3.1 mm³. Power for the observed effect (2.28) is close to 54%. In this setting a significant result tends to overestimate the true effect (type M error).
• Wide interval relative to the segment. The CI of 0.08 to 4.48 mm³ over a baseline volume of about 33 mm³ runs from 0.24% of the segment to about 7%.
• The difference comes from the control group. The DCB arm did not change (+0.135 mm³; p = 0.811). The discussion refers to late lumen enlargement and positive remodelling. Inhibiting neointimal hyperplasia predicts less loss, and less loss is what the data show.
• Loss to follow-up. Of 299 randomized, 198 had paired volume: 34% without the main endpoint. Complete-case analysis, with no multiple imputation and no tipping-point analysis. Sex, diabetes, prior MI and Medina class were imbalanced, with no standardized differences and no adjusted analysis. ANCOVA on baseline volume would be the conventional choice for a change score.
• Safety without a margin. Safety is presented as noninferiority with no margin, no M1/M2 justification and no power calculation. The minimum detectable difference in uncovered strut ratio is about 0.62 percentage points. In relative terms, the study cannot exclude a relevant increase.