The study cohort consisted of 271
082 patients, of whom 23
991 (8.9%) underwent non-invasive stress testing within 180 days before surgery (tables 1 and 2). Some patients underwent more than one test; hence, a total of 25
877 stress tests were performed. The median time between testing and surgery was 48 days (interquartile range 16 to 101 days). A total of 7795 (30%) of these tests were ordered by cardiologists, 7234 (28%) by internists, 6281 (24%) by family physicians, 3594 (14%) by surgeons, 477 (1.8%) by anaesthesiologists, and 496 (1.9%) by other specialists. Of the patients who underwent preoperative testing, 914 (3.8%) underwent coronary angiography, 149 (0.6%) underwent percutaneous coronary intervention, and 134 (0.6%) underwent coronary artery bypass graft surgery between the dates of stress testing and surgery.
Table 1 Preoperative characteristics of entire cohort. Values are expressed as number (percentage) unless indicated otherwise
Table 2 Perioperative characteristics of entire cohort. Values are expressed as number (percentage) unless otherwise indicated
Patients who underwent preoperative testing and those who did not differed with regard to all measured characteristics (tables 1 and 2). Patients who underwent testing were typically men who had surgery at a high volume or moderate volume non-teaching hospital and had an increased burden of comorbid disease. They were also more likely to be evaluated by a specialist before surgery, undergo preoperative cardiac procedures, and require intraoperative care such as epidural anaesthesia or intraoperative invasive monitoring.
Of the patients who underwent stress testing, 23
060 (96%) were successfully matched to a similar patient who did not. The covariate balance between the two arms was improved considerably by propensity score matching (tables 3 and 4): the mean standardised difference between the two groups decreased from 14.9% (range 0.3 to 90.9) to 0.48% (0.02 to 1.5). Of the matched patients who underwent testing, 914 (4.0%) underwent coronary angiography, 136 (0.6%) underwent percutaneous coronary intervention, and 119 (0.5%) underwent coronary artery bypass graft surgery between the dates of stress testing and surgery.
Table 3 Preoperative characteristics of the propensity matched pairs. Values are expressed as number (percentage) unless otherwise indicated
Table 4 Perioperative characteristics of the propensity matched pairs. Values are expressed as number (percentage) unless otherwise indicated
Within the matched cohort, one year survival was higher among patients who had undergone preoperative testing than in those who had not (hazard ratio (HR) 0.92, 95% CI 0.86 to 0.99, P=0.03; fig 1). This corresponded to a number needed to treat38
to prevent mortality at one year of 221 (95% CI 111 to 16
067). In hospital mortality and hospital stay were also reduced among patients who underwent stress testing (relative risk (RR) 0.85, 95% CI 0.73 to 0.98; P=0.03 and 8.72 days v
8.96 days, difference −0.24 days, 95% CI −0.07 to −0.43; P<0.001, respectively; table 5). Patients who had undergone stress testing were more likely to be admitted to a monitored bed after surgery than were those who had not undergone testing (RR 1.09, 95% CI 1.06 to 1.12; P<0.001; table 5). Conversely, rates of postoperative mechanical ventilation were similar in the two groups (RR 1.02, 95% CI 0.98 to 1.08; P=0.25).
Fig 1 Survival curves for postoperative all cause mortality in patients who did (n=23060) or did not (n=23060) undergo preoperative stress testing (matched by propensity score) over one year after surgery
Table 5 Processes of care and outcomes in the propensity matched pairs. Values are expressed as number (percentage) unless otherwise indicated
In sensitivity analyses, the association of stress testing with improved one year survival was unaffected when the number of previous acute care hospital admissions within two years before the index surgery was added to the original propensity score (HR 0.92, 95% CI 0.86 to 0.99; P=0.02). Additionally, we found no statistically significant association between stress testing and surgical site infections (RR 1.00, 95% CI 0.94 to 1.07; P=0.89; table 5).
The subgroup of individuals older than 66 years consisted of 15
475 patients who underwent stress testing and 15
475 who did not. In this subgroup, 5626 (36%) of those who underwent preoperative stress testing and 3998 (26%) of those who did not were receiving β blockers (RR 1.41, 95% CI 1.36 to 1.45; P<0.001). Additionally, 1895 (12%) patients who had undergone testing were new β blocker users, compared with 827 (5.3%) who had not undergone testing (RR 2.23, 95% CI 2.12 to 2.48; P<0.001). More patients in the stress testing group were receiving statins (5000 (32%)) than were those who had not been tested (3679 (24%); RR 1.36, 95% CI 1.31 to 1.41; P<0.001). A total of 974 (6.3%) patients who had undergone testing were new statin users, compared with 539 (3.5%) of patients who had not (RR 1.81, 95% CI 1.63 to 2.00; P<0.001).
The association of stress testing with mortality was unchanged when the analyses were repeated in subgroups defined by procedure type (P=0.28 for interaction; fig 2). Conversely, the effects of testing on mortality varied with RCRI class (P=0.005) and, to a degree, ischaemic heart disease (P=0.08). Preoperative stress testing was associated with harm in low risk patients (RCRI 0 points: HR 1.35, 95% CI 1.05 to 1.74); however, it was associated with improved survival in intermediate risk patients (RCRI 1-2 points: HR 0.92, 95% CI 0.85 to 0.99) and high risk patients (RCRI 3-6 points: HR 0.80, 95% CI 0.67 to 0.97; fig 2). These differences corresponded to a number needed to treat to prevent mortality at one year of 156 for intermediate risk patients (95% CI 79 to 6127) and 38 for high risk patients (95% CI 21 to 315). Conversely, the number needed to harm in low risk patients was 179 (95% CI 97 to 1090). The relative rates of intervention use (that is, preoperative medications and cardiac procedures) were higher in low risk patients who underwent stress testing than in intermediate or high risk patients who were tested (web table A). The characteristics of the subgroups defined by RCRI class are presented in web table B.
Fig 2 Association of preoperative stress testing with one year survival in the subgroup analyses. The dashed vertical line represents the overall treatment effect (hazard ratio 0.92) and the solid vertical line represents a null effect (hazard ratio 1) (more ...)