Trial is error? Evaluating the effectiveness of single-troponin diagnostic pathways

Abstract

The most tangible benefit of implementing highsensitivity cardiac troponin (hscTn) assays is the potential to reduce length of stay for patients presenting to the hospital with symptoms compatible with acute myocardial infarction (AMI). There is now abundant evidence to demonstrate the high diagnostic accuracy of rapid diagnostic algorithms that aim to identify patients who have a very low probability of AMI within just a few hours of arriving in the emergency department (ED). By using very low cutoffs set at or close to the limit of detection (LoD) of hscTn assays, a single blood test is sufficient to ‘rule out’ the diagnosis of AMI for many patients. 2 Those patients do not require any further inpatient investigation and may be suitable for immediate discharge, allowing the patients to benefit from earlier reassurance and relieving ED crowding. The ‘singletest’ strategy has substantial potential benefit for busy EDs. Avoiding a second venepuncture saves staff time. Further, more patients will have all the necessary investigation results to hand at the time of their consultation with a clinician. If clinicians do not have that information to hand, they are more likely to see new patients and are less likely to be immediately available once the required investigation results become available, thus increasing the waiting time for patients and reducing efficiency. The implementation of such rapid ruleout strategies has recently been evaluated in two randomised controlled trials (RCTs). The Limit of Detection of Troponin and ECG Discharge (LoDED) trial randomised 632 patients to receive standard care or care guided by a pathway that allowed patients to be discharged if they had an hscTn concentration below the LoD of the assay. The trial found no difference in the proportion of patients safely discharged from the ED, although there was no increase in the incidence of major adverse cardiac events in the intervention group. The HighSensitivity Cardiac Troponin on Presentation to Rule Out Myocardial Infarction (HISTORIC) trial had a stepped wedge design. Seven acute care hospital sites implemented a singletest ruleout strategy using the Abbott hscTnI assay in random order, including a total of 31 492 patients. Again, the trial was ‘negative’ for the primary outcome. In this instance, implementing the singletest strategy successfully reduced length of stay (from 10.1 to 6.8 hours, p<0.001). However, while there was no observed increase in the incidence of AMI or death at 30 days (0.4% vs 0.3%), the 95% CIs were too wide to conclude noninferiority based on the prestated noninferiority margin. By the letter, therefore, we now have two negative trials evaluating singletest ruleout strategies, but for different reasons. The LoDED trial failed to show an increase in the number of patients discharged within 4 hours, while the HISTORIC trial failed to show noninferiority with regard to safety outcomes. However, neither trial showed an increase in the incidence of adverse events after discharge. Further, the findings of the LoDED trial could be explained by noting that the control group received standard care, which may have included using other singletest ruleout strategies (eg, based on the HEART score). Arguably, therefore, while both are ‘negative trials’, the findings of each trial still support implementation. Perhaps the inherently inflexible design of an RCT means that this method is suboptimal to evaluate the implementation of diagnostic pathways of this nature. In this issue, Barnes et al report on the realworld implementation of a singletest ruleout strategy in Perth, Australia. They used an optout consent model to prospectively collect data preimplementation and post implementation. It is particularly impressive that implementation was so rigorously planned in advance, enabling robust evaluation. The preimplementation phase involved a prospective evaluation for almost 6 months, with an additional 6month ‘washout’ period, prior to making the change. The group used the Abbott hscTnI assay (ARCHITECT) with a ‘ruleout’ threshold of 5 ng/L, although only patients who were identified as ‘low risk’ using the History, Electrocardiogram, Age, Risk factors, Troponin (HEART) score (≤3 points) were eligible for immediate discharge. The study showed that implementation of the singletest pathway led to an increase in the proportion of patients being discharged from the ED within 3 hours of arrival (21% vs 38%, p<0.001), with no deaths or AMIs occurring in either group at 30 days. There was no apparent change in the use of functional or anatomical imaging for coronary artery disease after discharge. The findings therefore suggest that implementation of this rapid diagnostic pathway was safe and effective, thus providing support for implementation of similar pathways at other centres. Indeed, if anything, the pathway evaluated may be too conservative. Observational data suggest that hscTnI <5 ng/L has high negative predictive value in patients presenting >2 hours from symptom onset 5 and that use of the HEART score may reduce efficiency without increasing safety. Further, the absence of adverse events at followup suggests that there is potential to reduce outpatient referrals among the patients who are discharged early. It therefore seems that the study reported by Barnes et al teaches us two important things. First, a singletest ruleout strategy can be implemented safely and effectively in realworld settings. Second, this work questions whether such wellplanned realworld evaluations should now be considered the optimum way to assess the effectiveness of implementing diagnostic pathways in practice.