Gim et al.: imaging-guided PCI versus CABG in diabetic patients with left main or three-vessel disease

Summary

The study included 3402 diabetic patients with left main or three-vessel disease, out of 6962 patients. Intravascular imaging-guided PCI (n = 392) came from RENOVATE-COMPLEX-PCI, which enrolled from May 2018 to May 2021. CABG came from the Samsung Medical Center registry, 2001 to 2017.

The primary outcome was the composite of all-cause death, non-fatal MI or stroke at 3 years.

Imaging versus CABG: adjusted HR 0.88 (95% CI 0.58 to 1.32; p = 0.525). Propensity-matched: HR 0.85 (0.53 to 1.36). The authors conclude that outcomes are comparable.

Strengths

• The clinical question is relevant and has no randomized trial yet.
• Individual data from a randomized trial of imaging-guided PCI.

Methodological issues

• Absence of difference treated as equivalence. The result is non-significant, in an observational hypothesis-generating study, with no prespecified non-inferiority margin. A CI ranging from a 42% reduction to a 32% increase in hazard does not establish equivalence.
• Heavy censoring in the imaging group. Numbers at risk in Figure 3A are 392, 359, 243 and 142 at years 0, 1, 2 and 3. With 36 events, about 214 patients (55%) were censored before 3 years. The 3-year estimate rests on 142 patients.
• Possibly informative censoring. Many RENOVATE-COMPLEX-PCI patients did not reach 3 years before data lock. The CABG group has a complete risk set. Kaplan-Meier is valid only if censoring is non-informative, and this cannot be checked. If it is informative, the imaging group event rate is likely underestimated. Propensity matching balances covariates at time zero; it does not correct censoring that varies during follow-up.
• Temporal confounding. CABG spans 2001 to 2017; imaging-guided PCI is mostly after 2018. Medical therapy, stents and surgical technique changed between eras.
• Different anatomy. Left main disease in 55.9% of the imaging group and 22.3% of the CABG group. Three-vessel disease in 63.5% and 93.4%. Left main anatomy is usually more amenable to PCI than diffuse three-vessel disease. SYNTAX scores were not collected, so residual anatomical confounding cannot be measured.