Percutaneous coronary intervention versus medical therapy for chronic total coronary occlusions: a systematic review and meta-analysis of randomised trials

The study protocol was published in the PROSPERO database prior to starting (CRD42018115243) [13]. We searched MEDLINE, Embase and the Cochrane Library from inception to 17 June 2020, using the search terms ‘Chronic total coronary occlusion’, ‘Percutaneous coronary intervention’ and ‘Random’ (complete search string available in Electronic Supplementary Material). No language or publication period restrictions were used. Two researchers (AvV and JE) independently screened the articles and identified all studies that met the inclusion criteria, using online Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia). Disagreement was resolved by consensus after consulting the senior author (JPSH). All randomised trials comparing CTO-PCI with no CTO-PCI or OMT were included. Substudies were included if they described mortality, MACE or ventricular function. We excluded all non-randomised, observational studies, as well as trials studying non-CTO lesions or treatment with coronary artery bypass grafting. The reference lists of all included studies were checked to ensure no relevant studies were missed.

The risk of bias was assessed with the Cochrane Risk of Bias Tool [14]. Data were extracted from the manuscripts, using a standardised form (adapted from the Cochrane Collaboration). If data were not stated in the text, we calculated these from frequencies or retrieved them from the figures, using Digitizelt Software (version 2.3.3, Braunschweig, Germany). Unavailable or incomplete data were considered missing. For the purpose of data synthesis, we calculated the incidence of target lesion revascularisation (TLR) and the left-ventricular end-diastolic volume (LVEDV) corrected for body-surface area (LVEDV index) in the EXPLORE study from the original EXPLORE database [7]. The GRADE criteria were used to score the quality of the data of the included studies [15].

All endpoints were defined according to the definitions of the original trials, and as an intention-to-treat analysis at the longest follow-up available. The endpoints of this meta-analysis were mortality at 12 months and at longest follow-up, MACE at 4–6 months, 12 months and at longest follow-up, angina complaints at 4–6 months and 12 months, left ventricular function in terms of LVEF at 4–6 months and LVEDV at 4–6 months, segmental function [in terms of segmental wall thickening (SWT)] and procedural complications. We analysed MACE as reported by the individual studies (MACE_reported) and, to overcome inconsistent definitions, we analysed the different components of MACE separately, i.e. cardiac death, myocardial infarction (MI) and TLR (or, if not available, target vessel revascularisation). TLR included repeat revascularisation of the treated CTO lesion for patients randomised to CTO-PCI and revascularisation of untreated target CTOs for patients randomised to OMT, in accordance with the definitions of the included trials.

Continuous variables are presented as mean and standard deviation (SD). For the purpose of data synthesis, we assumed medians equal to means and interquartile ranges equal to SD multiplied by 1.35 [16]. Discrete data are presented as frequencies. The results are synthesised in a quantitative manner with random effects models using the Mantel-Haenszel method. Pooled discrete data are presented as risk ratios (RR) with 95% confidence intervals (CI). We calculated the mean difference (MD) with associated 95% CI using the inverse variance method to compare continuous variables among studies. We considered a two-sided p-value of less than 0.05 to be statistically significant. Heterogeneity between trials was assessed with the I² statistic. I² < 25% was considered as low heterogeneity, 25–50% as moderate, and >50% as high. Microsoft Excel 2016 and Review Manager, version 5.3, were used to perform statistical analyses.

All-cause mortality at 1 year was comparable between groups (CTO-PCI 1.6% vs OMT 1.0%; RR 1.70, 95% CI 0.50–5.80, p = 0.40; Fig. 2a), as was cardiac mortality (1.2% vs 0.5%; RR 1.77, CI 0.19–16.06, p = 0.61; Fig. 2b). At long-term follow-up (up to 4 years), all-cause mortality was 5.1% in the CTO-PCI group and 4.5% in the OMT group (Fig. S1; RR 1.14, 95% CI 0.38–3.40, p = 0.81). Cardiac mortality at long-term follow-up was 2.7% in the CTO-PCI group versus 2.3% in the OMT group (RR 1.66, 95% CI 0.31–8.79, p = 0.55; Fig. 2c).

Different definitions were used for MACE (Table S5). There were no significant differences in MACE_reported at a weighted follow-up of 5 months (4.8% vs 3.9%; RR 1.21, 95% CI 0.41–3.60, p = 0.73), or at 1 year (5.9% vs 9.1%; RR 0.69, 95% CI 0.36–1.33, p = 0.27) or long-term at a weighted follow-up of 4 years (17.4% vs 19.7%; RR 0.85, 95% CI 0.60–1.22, p = 0.38; Fig. 3b).

We also analysed the different components of MACE separately. There the occurrence of MI was comparable between treatment groups at 1 year (Fig. S2a) and at long-term follow-up (Fig. S2b). However, TLR at 1 year in the intention-to-treat population was significantly more frequent in the OMT group (12.4% vs 3.1%; RR 0.28, 95% CI 0.15–0.52, p < 0.001; Fig. 3c). TLR at long-term follow-up was numerically, but not significantly, different (CTO-PCI 7.9% vs OMT 13.2%; RR 0.55, 95% CI 0.28–1.09, p = 0.09; Fig. S3).

The Seattle Angina Questionnaire was performed in two studies at 1‑year follow-up [9, 10]. On the different subscales of this questionnaire, no significant differences were found (Table S6). ‘Freedom from angina’, defined as Canadian Cardiovascular Society grading of angina pectoris [18] grade 0, was reported in two trials [9, 12]. Significantly more patients were not free of angina at 1 year in the OMT group (27.3% vs 20.8%; RR 0.65, 95% CI 0.50–0.84, p = 0.001; Fig. 4).

There was no difference in LVEF between treatment groups at weighted 5‑month follow-up (MD 2.07%, 95% CI −1.12–5.25, p = 0.20; Fig. S4a), or in LVEF change from baseline (MD 0.28%, 95% CI −0.70–1.27, p = 0.57) or LVEDV change from baseline (MD 0.03 ml/m², 95% CI −2.93–2.99, p = 0.98). Segmental function was assessed in two studies [8, 17]. SWT and SWT recovery in the CTO territory did not differ significantly at 5‑month follow-up. There was a non-significantly greater improvement of SWT in dysfunctional CTO segments (segments with SWT at baseline <45%) in the CTO-PCI group (MD 5.19%, 95% CI −0.47–10.84, p = 0.07; Fig. S4b).

Previous meta-analyses of observational studies demonstrated that successful CTO-PCI was associated with lower mortality and fewer MACE compared to failed CTO-PCI [19]. Nonetheless, this comparison is limited since factors leading to procedural failure, such as severe calcifications and longer CTO lesions, may also reflect an overall impaired health status [20]. Therefore, aggregation of solely randomised data presents a fairer comparison of the two treatment strategies. We found no difference in mortality during up to 4 years of follow-up. Additionally, the incidence of MACE—although defined differently in the studies—did not differ significantly at several time points during follow-up. This was also demonstrated in a previous meta-analysis by Iannaccone et al. [21], but observational studies were also included which could have distorted the results. It is noteworthy that both mortality rate and MACE incidence are low, even after 4 years of follow-up. Thus, OMT for CTO patients seems to be a safe strategy. When analysing the different components of MACE separately, we found that the frequency of TLR was significantly higher after 1 year in the OMT group, but not at long-term follow-up. Most studies stated that TLR was clinically or ischaemia-driven, but this was not clearly specified. Presumably, TLR was often due to residual angina and the difference between randomisation groups could become less apparent during follow-up due to the occurrence of symptomatic reocclusions. In order to interpret the frequency of TLR properly, it should be noted that in the DECISION-CTO trial [10] 20% of patients in the OMT group crossed over to CTO-PCI within 3 days after randomisation. These crossovers were considered protocol deviations rather than TLR; thus the number of TLRs is possibly underreported. Because of the high number of TLRs in the OMT group, the actual number of patients with angina in this group might in fact have been higher if no TLR had been performed, leading to an underestimation of the effect of CTO-PCI. The physiology of this observed angina relief could be explained by a reduction of ischaemic burden after CTO-PCI [22]. To test this, our research group designed the REVISE-CTO trial (ClinicalTrials.Gov NCT03756870), which will study the effect of CTO-PCI on ischaemia reduction, compared with OMT, in patients with predefined ischaemic burden.

A significant improvement of LVEF after CTO-PCI has been described when compared to before the procedure, but no control group was included [5]. We could not find an improvement of LVEF after CTO-PCI when compared with a true control group (OMT). Yet, baseline LVEF in this meta-analysis was not impaired (i.e. 53%); thus a significant improvement of global ventricular function is unlikely to be achieved.

Minimal changes in ventricular function could, however, manifest at segmental level through an improvement of wall thickening in the specific CTO segment. We found a minimal trend towards improvement of SWT in CTO segments that were dysfunctional, but viable, at baseline. It has previously been suggested that preselecting those patients with viable CTO segments might contribute to improved ventricular function and clinical outcomes after CTO revascularisation [23, 24]. The trials studied did not include patients based on imaging criteria, possibly causing an underestimation of the treatment effect of CTO-PCI in a select patient group. Future studies are warranted to further explore appropriate imaging/ischaemia criteria for CTO-PCI.

Various limitations should be acknowledged. First, the included trials differed in patient selection, which resulted in a heterogeneous population. Therefore, we analysed all available outcomes with and without the EXPLORE [7] and IMPACTOR-CTO [11] trial, and no important differences were found. Moreover, the I² was low for almost all comparisons. Including post-STEMI (ST-segment elevation myocardial infarction) patients and those with CTO of the right coronary artery allowed us to yield the maximum cohort size for this meta-analysis. Second, different definitions were used for MACE. By analysing the different components separately we tried to overcome this obstacle. Finally, there was a significant number of crossovers from OMT to CTO-PCI in the DECISION-CTO [10] trial. These crossovers were not considered to be adverse events; therefore the number of TLRs would probably be higher than currently reported. With individual patient data of all included CTO-PCI trials, as-treated and per-protocol analyses could be performed.