| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
EDITORIAL |
Chicago, USA
1 Plymouth Meeting, USA
Time to intervention is a potentially important factor in clinical trials comparing medical procedures or devices (but not in drug trials as there is usually little delay in treatment). This is because average waiting times may differ substantially between 2 treatments being compared. An example is randomized controlled trials (RCTs) comparing percutaneous coronary interventions (PCI) with coronary artery bypass grafting (CABG), which have shown that average waiting time for CABG is usually 2–3-fold longer than for PCI. If there is a difference in waiting times, a longer wait in one group could increase the number of adverse events during the delay. Such events must be included in analyses of event rates for accurate comparison of the risks of different treatment strategies.
Patients waiting for cardiac surgery remain at risk of a fatal or non-fatal event. In a study by Ray and colleagues1 of 1,854 patients undergoing CABG, aortic valve replacement, or combined procedures, median waiting times were: urgent group, 8 days; semi-urgent A group, 37 days; semi-urgent B group, 64 days; and elective group, 113 days. There were 13 (0.7%) deaths (12 cardiac) during the waiting period. Rexius and colleagues2 found an increased mortality risk of 11% per month after acceptance for CABG, and concluded that long waiting lists are associated with considerable mortality. Studies have also documented major cardiac events (death and myocardial infarction) in patients waiting for PCI.3,4 Alterations in lesion characteristics during waiting have resulted in unsuccessful procedures.5 There was a significant decrease in primary success rate for type C lesions (worst atherosclerotic lesion on coronary angiography) after waiting > 12 weeks. A reasonable explanation for unsuccessful angioplasty related to alterations in lesion characteristics during the waiting time was documented in only 4 of 115 procedures.5 Two recent meta-analyses of RCTs comparing PCI/stents to mini-CABG concluded that mini-CABG resulted in less re-intervention than PCI.6,7 There was no significant difference in post-intervention survival. Although these meta-analyses were statistically sound, they did not fully address the methodology and consenting issues in individual studies with respect to waiting time following randomization.
REPORTING OF WAITING TIME IN TRIALS OF PCI VS MINI-CABG
Effective reporting of RCTs has received increasing attention in recent years. Many journals now require reports to conform to the guidelines in the Consolidated Standards of Reporting Trials (CONSORT) statement that includes a checklist of items that should be included in a trial report.8,9 Among these is the number of participants included in each analysis and whether intention-to-treat (ITT) analysis was performed. This analysis is required to capture all events from the time of randomization forward, to enable reviewers to determine whether differences in waiting times have been accounted for. Careful review of the existing RCTs comparing PCI to mini-CABG revealed that only 2 of 6 addressed waiting time as a variable in either study design or summary of results.10–15 Of those reporting waiting time, only one reported a target (30 days), and this was met in only 40% of patients.12 Median waiting time for stenting was 35 vs 44 days for mini-CABG. These were far shorter in the study by Diegeler and colleagues14 where waiting time for stenting was 2 vs 14 days for mini-CABG. It is difficult to ascertain the true impact of this variation, but clearly this weighted variable will always be present in prospective studies. Furthermore, only 2 PCI vs mini-CABG trials stated that ITT analysis was performed, suggesting that waiting time may be an overlooked confounder in some RCTs. On careful review of all 6 studies, we were unable to find any that reported all relevant factors, including time from randomization to intervention, deaths, and other adverse events, and calculated events after randomization and before intervention. It is possible that the other 4 trials performed ITT analysis but did not report it. If so, confounding from differential waiting times may not be a problem in this evidence base, so long as the delays are similar to waiting times in clinical practice. Even if true, it illustrates the need to report all important study design features, as recommended by CONSORT, so that the validity of the findings will not be in doubt.
However, if ITT analysis was not performed in these trials, differential waiting times together with the inherent bias of lack of blinding can result in overestimation of the beneficial effect of either mini-CABG or PCI. There is no adequate method to correct for this in a systematic review, so the summary effect estimate of a meta-analysis that includes data from trials that did not perform ITT analysis is potentially biased also. Therefore, all RCTs of procedures and devices should conduct ITT analysis, and clearly state that such analysis was performed.
THE PROBLEM OF WAITING TIME IN OBSERVATIONAL REGISTRIES
If differential waiting times is a potential problem in RCTs, it is even more likely to be a confounder in studies comparing outcomes of different procedures based on data from large patient registries or databases. The information in such sources usually begins with the patient’s entry into the hospital. Prior waiting time and events occurring during the waiting period are not typically captured by registries or other databases. Patients who die while waiting are not included in mortality statistics. This was acknowledged as a limitation of a recent analysis of CABG vs PCI outcomes of nearly 60,000 patients with procedures recorded in 2 New York State cardiac registries.16 This showed that risk-adjusted long-term survival rates were significantly higher after CABG than PCI; in contrast, recent meta-analyses of RCTs have not yet shown a significant survival advantage of CABG. Because RCT meta-analyses contain far fewer patients, the difference could be due to the lower statistical power of the meta-analyses. Others have pointed out the more restrictive inclusion/exclusion criteria of RCTs, which limits their ability to generalize. However, the apparent difference in survival could also be due to the inability of registries to capture deaths during the waiting period. As the typical waiting period is longer for CABG than PCI, more deaths are likely to occur among patients waiting for CABG. Thus, the adjusted hazard ratio reported in a registry study is likely to overestimate the survival advantage of CABG. This problem could be overcome if registries counted the time when the patient is first referred for therapy as time zero. This would allow adverse events during the waiting period to be captured, and thus control for the effects of differential waiting times.
SUMMARY
As we evolve in the field of contemporary cardiothoracic surgery and witness modern applications of new techniques and technology, we need to be careful of how statistical methods are executed. Publications with hidden mediators that are not adequately addressed can lead to biased conclusions, especially when meta-analyzed. Public health policies need to be sure that their statements are as unbiased as possible for correct inference, leading to optimal patient safety and well-being. Careful analysis of hidden mediators is important in studies comparing the effectiveness of procedures and devices. Such analysis is critical in identifying mediators such as waiting time that should be considered when constructing interventions to be evaluated in the next RCT. In particular, RCTs of devices and procedures should always conduct (and report) ITT analysis, capturing all events from the time of randomization forward to control for differential waiting time. Similarly, observational registries and databases should count time zero as the time when patients are first referred for therapy, rather than when they enter a hospital to receive treatment; this would ensure that events during the waiting period are captured.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| ANN THORAC SURG | ASIAN CARDIOVASC THORAC ANN | EUR J CARDIOTHORAC SURG |
| J THORAC CARDIOVASC SURG | ICVTS | ALL CTSNet JOURNALS |
