Variables

Study variable.

The study variable is treatment timing, operationalized as the time to coronary revascularization treatment and computed in calendar days. Based on treatment timing and the type of revascularization procedure received, patients were assigned to one of two study groups: delayed CABG or timely PCI. The time to treatment starts on the date when the need for revascularization is clinically established and the patient is ready, willing, and able to undergo revascularization. The time to treatment ends on the date the index revascularization procedure was performed. To establish intervals defining timely and delayed treatment, we used the Canadian Cardiovascular Society (CCS) recommended times [9] to define delayed CABG and timely PCI for semi-urgent and elective CABG and PCI patients, the First Minsters’ Meeting benchmarks [10] for urgent CABG, and CSBC benchmarks for urgent PCI patients (Table 1). In this analysis, appropriate treatment timing is defined by each of the three patient urgency categories. Table 1 defines the “appropriate time’ and ‘with delay’ time intervals for each urgency category”.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Study group assignments by procedure type and urgency and treatment delay in days.

https://doi.org/10.1371/journal.pone.0303222.t001

Statistical methods

We estimated the frequency and percentage of patients by characteristics and by treatment group. Groups were compared using a chi-square test for categorical variables and p-values for between group differences reported. We modelled cumulative mortality over three years using with a flexible parametric approach. This approach uses restricted cubic spline functions, also known as Royston-Parmar models, to model the baseline mortality over three years [13]. The main advantage of this model is that it provides the means to smoothly estimate the survival function, in contrast to the Cox model, where the baseline hazard function can be a noisy step function [13]. We then estimated the unadjusted cumulative mortality proportion over three years for each study group. Finally, we estimated risk ratios comparing the treatment groups. in mortality at three years. A risk ratio of less than one means the delayed CABG group had a lower risk of mortality at 3 years compared to the timely PCI group. A risk ratio of greater than one means the delayed CABG group had a higher risk of mortality at 3 years compared to the timely PCI group. We selected three years’ follow-up as the follow-up time, considering the variation in follow-up time used in randomized controlled trials of CABG versus PCI.

We estimated propensity scores [14] for the probability of belonging to each study group using logistic regression and, using those scores, calculated inverse probability of treatment weights [15], following the ASCERT approach [8]. Inverse probability of treatment weighting creates a synthetic cohort that utilizes all patient information, compared to other propensity score methods where this cannot be assured. Variables were selected starting with those used in ASCERT [8] and informed by a scoping review of the factors of mortality after CABG [16]. Each patient was weighted by the inverse of the probability of being assigned to their treatment group to adjust for differences between the two treatment groups. We compared the performance of the propensity score model by comparing the distribution of covariates and propensity scores before and after inverse probability weighting. Adjusted mortality estimates were obtained using an inverse probability weighted flexible parametric approach. Statistical analyses were performed using Stata 17 (College Station, TX). Flexible parametric models were constructed using stpm2, a Stata software package [17].

Patient and public involvement

We consulted the Pacific Open-Heart Association (POHA) to inform development of our research question. POHA provides peer support to patients undergoing or who have undergone heart surgery in the Vancouver, BC area. They confirmed that BC patients often wait for CABG and that anxiety results when it isn’t known when a planned CABG will occur. In this study, we address the question that patients told us matters most: would it be better for CABG candidates to undergo PCI instead of facing an indeterminate length of time waiting for CABG?

Data access

The authors first gained access to the data for research purposes on June 7, 2019, and had access to the data through April 25, 2024. The authors had no access to information that could identify individual participants during or after data collection.

Setting and participants

We identified 39,176 British Columbia patients who met the selection criteria for our study (Fig 1).

We did not select patients for the analytical cohort if their revascularization record could not be linked to hospital records or their PCI record was for ad-hoc PCI, but the procedure could not be linked to a diagnostic catheterization (n = 591), or that their hospital records did not contain diagnosis codes indicative of stable ischemic heart disease (n = 696). 37,889 patients were eligible for analysis. We set aside patients if their procedure urgency could not be determined (n = 56), if patients with multiple PCI records were unresolved after applying the repeat revascularization algorithm or if there were errors in the administrative data set where date of death preceded date of revascularization (n = 72), if the patient received delayed PCI (n = 2,550), or if the patient received timely CABG (n = 9,711). 25,520 patients were available to be analyzed.

Descriptive data

Table 2 shows the baseline characteristics of patients in the study cohort.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Baseline characteristics of the patients.

https://doi.org/10.1371/journal.pone.0303222.t002

Before adjustment with inverse probability weighting, the patients undergoing delayed CABG were, compared to patients undergoing timely PCI, had higher proportions of male sex, a BMI >30, triple vessel disease, left main disease, and an ejection fraction ≤50%. The delayed CABG group had significantly higher proportions of atrial fibrillation or atrial flutter, congestive heart failure, diabetes, hypertension, and renal disease, compared to timely PCI. The timely PCI group had higher proportions of double-vessel disease and Canadian Cardiovascular Society (CCS) Angina Class 4. Most patients, regardless of study group, were treated in metropolitan hospitals. Clearance time is shorter amongst patients treated with timely PCI compared to delayed CABG. Proportions of neighborhood income decile are balanced throughout the study cohort. Of the patients who underwent PCI, 48.1% received Bare-Metal Stents (BMS), 4.5% received a combination of BMS and Drug-Eluting Stents (DES) and 42.8% received only DES. Of the patients who underwent CABG, 8.5% received only a saphenous vein graft, 71.6% received a single arterial graft, 16.3% received a double arterial graft, and 3.4% received a triple arterial graft. The mean waiting time for the timely PCI group was 11.9 days and the mean waiting time for the delayed CABG group was 71.5 days (S1 Appendix).

As expected, patients in the timely PCI group had a lower probability of being selected for delayed CABG than did those in the CABG group. However, all patients had a positive probability of being assigned to either CABG or PCI, consistent with results of comparative effectiveness studies of CABG and PCI published elsewhere [8].

Outcome data and main results

Unadjusted failure curves are shown in Fig 2; unadjusted cumulative mortality and risk ratios are shown in Table 3.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Cumulative mortality in the CABG and PCI populations, from an unadjusted analysis.

https://doi.org/10.1371/journal.pone.0303222.g002

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Rates of mortality (percent), risk ratios, and 95% confidence intervals in the delayed CABG and timely PCI populations, from an unadjusted analysis.

https://doi.org/10.1371/journal.pone.0303222.t003

Failure curves adjusted with inverse probability of treatment weighting are shown in Fig 3; adjusted cumulative mortality and risk ratios are shown in Table 4.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Cumulative mortality in the CABG and PCI populations, from an analysis adjusted with the use of inverse probability of treatment weighting.

https://doi.org/10.1371/journal.pone.0303222.g003

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Rates of mortality (percent), risk ratios, and 95% confidence intervals in the delayed CABG and timely PCI populations, from an adjusted analysis.

https://doi.org/10.1371/journal.pone.0303222.t004

Key results

This study used data from multiple population-based registries and databases to evaluate the effectiveness of delayed CABG as compared with timely PCI. In this study, we found that amongst British Columbia patients 60 aged years or older, who underwent non-emergency first-time revascularization for angiographically-proven, stable left main or multi-vessel ischemic heart disease in British Columbia, between January 1, 2001, and December 31, 2016, there was a significant difference in both unadjusted mortality and mortality adjusted using inverse probability of treatment weights at 30 days, one year, two years, and three years. When the study cohort was stratified into three time periods, results were consistent with the results observed in the primary analysis (S1 Appendix).

Our findings should be considered in the context of results from other studies. There have been seven randomized, controlled trials comparing CABG with balloon angioplasty, eleven comparing CABG with PCI and stenting in patients with multi-vessel disease and six comparing CABG with PCI and stenting in specifically in patients with left-main disease. A survival advantage for CABG was noted in the Stent or Surgery (SoS) trial at two years [18] and sustained at six years [19]. In the ARTS-I trial, where BMS were used, a survival advantage at five years was not found [20]. However, the ARTS-II trial, which compared the ARTS-I CABG cohort to a new PCI cohort where patients received DES, did find a statistically significant difference in mortality. The FREEDOM trial, where diabetics with multivessel disease were treated with CABG compared to PCI with DES also found a statistically significant difference in mortality at five years [21], and again at seven years [22], consistent with the subgroup analysis reported by the BARI Investigators [23], but inconsistent with the SYNTAX diabetes subgroup study that showed no statistical difference [24], though a trend to better outcomes with CABG was observed. In the SYNTAX study, no difference in all-cause mortality was observed at three years, however, a statistically significant benefit was found for CABG patients with triple-vessel disease [25]. The benefits for patients with triple-vessel disease were sustained in the ten-year SYNTAX extension study [26].

Of note is the finding of a difference in mortality at 30 days of index revascularization or last staged PCI. Randomized controlled trials and observational studies have shown higher mortality in patients undergoing CABG during the first 30 days compared to patients undergoing PCI, attributed to increased risk of death in the immediate post-operative period. We selected only patients with stable ischemic heart disease, who would be expected to have lower surgical risk than patients with more unstable disease. This may account for lower early mortality observed in this study. BC also has a long-standing cardiac surgery quality oversight program, delivered by CSBC in collaboration with the cardiac surgery community. Annual reporting on hospital and surgeon mortality at 30 days and 1 year during the study period may have contributed to improvements in the structures and processes associated with care quality that could contribute to lower mortality.

Observational studies also inform our understanding of these results. The ASCERT study [8] examined the comparative effectiveness of CABG versus PCI in Medicare patients using patient selection criteria and analytical methods common to ours. While no significant difference in mortality was found after one year, lower mortality was observed with CABG compared to PCI at four years, similar to our results. Similar findings were recently reported by Mehaffey and colleagues using contemporary CABG and PCI techniques in Medicare patients [27]. Our findings are consistent with those of a recent systematic review [28] included 23 studies comparing CABG with PCI. CABG was associated with better survival during follow-up in 17 studies and no significant difference between treatments in six, with no study favouring PCI. A recent meta-analysis reported similar results in patients who received CABG compared to PCI with DES [29].

Prior research also shows that the incidence of adverse events on CABG waitlists is low in the non-emergency population. Specifically, the risk of death has been shows to be approximately 1% [30] in one study, with a death rate 0.5 per 1000 patient-weeks in semi-urgent patients and 0.6 per 1000 patient-weeks in non-urgent patients [31]. The emergency surgery rate in the latter study was 1.2 per 1000 patient-weeks in the non-urgent group, while the rate was 2.1 per 1000 patient weeks in the semi-urgent group. Given the low frequency of these adverse events while waiting in the non-emergency population, we would not expect our conclusions to change because of events occurring prior to index revascularization, the time where we begin to measure outcomes in this analysis. We are unaware of any studies that examine the effect of waiting for PCI. This is expected given that PCI is generally available both in Canada and in the United States within relatively short time periods. Our study uses timing intervals recommended by Canadian Cardiovascular Society physicians [9]. Thus, in the case of PCI, a very low event rate would be expected in stable patients given the consensus guidance to treat the lowest acuity stable patients within 42 days.

We conducted a sensitivity analysis to investigate the potential for unmeasured confounding. Using the using the methodology proposed by VanderWeele and Ding [32], we calculated the E-value for the analysis. The E-value is the minimum strength of association, on the risk ratio scale, that an unmeasured confounder would need to have, with both the treatment and outcome, to fully explain away a specific association, conditional on the measured covariates. A large E-value suggests considerable unmeasured confounding would be needed to explain away an effect estimate, while a small one suggests little unmeasured confounding would be needed. We determined that the observed risk ratio of 0.32 could only be explained by an unmeasured confounder with a risk ratio of at least 5.7-fold each (95% CI closest to the null risk ratio, 4.4), beyond the measured covariates. Therefore, it seems unlikely that a single unknown factor could have an effect sufficiently large enough to account, on its own, for the observed difference in mortality between the treatment groups.

While we hypothesized that at some point, patients with stable disease would benefit from changing therapies to PCI, this was not observed. It is known that in BC during the study period that significant CSBC quality oversight structures were in place for CABG whereas for PCI, these were only introduced in the latter third of the study period. Quality review and quality improvement mechanisms may have contributed to lower mortality in the CABG group and possibly higher mortality in the PCI group.

These results are noteworthy in that they demonstrate that amongst this patient population, the benefits of CABG do not appear to be attenuated by delay compared to PCI. For physicians who must advise patients on treatment strategies in the context of scarce resources, these results suggest that PCI as an alternative revascularization strategy may not be indicated if a reduced risk of mortality is desired. Patients can know that waiting for CABG may have benefits over PCI. Policymakers should interpret these results in the context of past CABG research, which shows benefit to earlier timing of treatment [5].

Limitations

This study has limitations. First, this study utilizes observational data as the basis for its analysis and conclusions. While took steps to model our design after a randomized controlled trial using the guidance of Cochran [33] and Rosenbaum [34], it’s possible these efforts were insufficient. It’s possible that group comparability was not the same in this study as in a randomized trial given the many factors, both conscious and unconscious, that may affect a revascularization method selection. Second, it’s possible unmeasured confounders affected our results. While we used inverse probability of treatment weights to balance differences in patient and health system factors and conducted a sensitivity analysis, it is still possible that unmeasured confounders affected these results. Third, we studied patients who underwent treatment between 2001 and 2016, during which stent technology evolved significantly, and the use of antiplatelet therapy evolved significantly. While we accounted for this by adjusting for calendar year of procedure, this may not have been sufficient to address the effect of time on outcomes. Fourth, we were limited by data available from CSBC in how we could establish extent of disease. While revascularization appropriate use criteria [35] suggest the use of SYNTAX scores to differentiate eligibility for CABG or PCI, this data is not routinely collected in BC. While SYNTAX scores are thought to have limited utility due to inter-rater variability [6], the absence of this data limited our ability to stratify our patient groups to match those proposed in appropriate use criteria. Fifth, the dataset lacked specific reasons for delaying CABG. Previous studies have linked variations in time to procedure to the weekly number of patients on the CABG waiting list and the weekly number of urgent cases [36]; therefore, we accounted for these factors by creating a clearance time variable that was included in the propensity score model. Sixth, clinical presentation data was not consistently available during the study period, so we used diagnosis codes to identify patients with stable disease and did not select any patient with an ‘emergency’ priority for our study cohort. These efforts may not have completely excluded patients with more serious acuity who were not eligible for CABG and instead treated with PCI. Finally, this study was performed in a publicly funded health system in Canada where access to care is limited by funded and operationalized capacity. Thus, its results may not generalize to other health systems where capacity is not a factor that constrains access to care.

Interpretation

Our results suggest that there is evidence that the treatment benefit of CABG surgery is not attenuated because of a delay in treatment when compared to PCI provided within appropriate time.

Generalizability

As a population-based study, our results can be generalized to similar populations as those selected for this study. Our results can also be generalized to cardiac services systems with structures and processes that align with those that found in BC. Caution should be taken in applying these results to other populations and systems.