Clinical background
Treadmill exercise testing gives more than a binary verdict on the presence or absence of ischaemia. It quantifies exercise capacity, the ST response and symptoms in a single test, and the score summarises these three dimensions into a prognostic figure. The decision the score serves is whether a patient with suspected or known coronary artery disease can be managed conservatively or whether referral for coronary angiography is warranted. Without a structured score, clinicians tend to over-read isolated ST changes and to overlook the fact that exercise capacity is often the single strongest prognostic variable.
Calculating the Duke treadmill score
The score is calculated as:
where the exercise-induced angina index is 0 (none), 1 (non-limiting) or 2 (exercise-limiting). Exercise time refers to minutes on the treadmill using the Bruce protocol. Maximal ST deviation is measured as the greatest horizontal or downsloping ST depression during or after the test, in millimetres. The score theoretically ranges from approximately −25 (highest risk) to +15 (lowest risk).
The derivation cohort consisted of 613 consecutive outpatients with suspected coronary artery disease referred for exercise testing at Duke University Medical Center between 1983 and 1985. Follow-up was 98 per cent complete at four years. The outcome modelled was all-cause mortality at four years [1].
Interpretation in practice
The score divides patients into three risk bands. The calculator uses the thresholds established in the derivation study, which have since become standard in guidelines:
| Score | Risk category | Four-year survival | Annual mortality | Clinical action |
|---|---|---|---|---|
| ≥ +5 | Low risk | 99% | approx. 0.25% | Conservative management. Coronary angiography is not indicated unless the symptoms worsen. |
| −10 to +4 | Intermediate risk | 95 % | approx. 1.25 % | Individual assessment. An imaging ischaemia test (myocardial perfusion scintigraphy or stress echocardiography) can clarify the need for coronary angiography. |
| < −10 | High risk | 79% | approx. 5% | Referral for coronary angiography. |
The intermediate risk group made up approximately 30 per cent of outpatients in the derivation cohort [1]. Their four-year survival was 95 per cent, corresponding to an average annual mortality of about 1.25 per cent [1], and it is the group in which the score offers the weakest decision support. Here additional risk markers, above all heart rate recovery and chronotropic incompetence, add information the score does not capture [2].
Validation and performance
In the derivation cohort the score achieved an AUC of 0.849 for distinguishing patients who died from those who survived four years, and it discriminated better than clinical assessment alone [1].
The most comprehensive external validation was carried out by Lauer and colleagues, who used data from 33,268 patients in a derivation cohort and 5,821 patients in a separate validation cohort, all with a normal resting ECG and suspected coronary artery disease. In this population the Duke treadmill score had a c-index of 0.73 for all-cause mortality, compared with 0.83 for an extended nomogram model that included age, sex, smoking, hypertension, diabetes, typical angina, heart rate recovery and ventricular ectopy in the recovery phase [2]. The extended model reclassified a substantial proportion of patients at intermediate or high risk by the Duke treadmill score as low risk. The score therefore retained its value as a quick and simple tool, but its discrimination was measurably poorer than could be achieved with readily available additional variables.
For women, Alexander and colleagues showed, in a cohort of 976 women and 2,249 men who had undergone both exercise testing and coronary angiography, that the score stratified prognosis as well in women as in men. It was, if anything, better at excluding significant coronary artery disease in women, with a lower proportion falsely classified as low risk [3]. Gulati and colleagues, by contrast, found in a prospective cohort of 5,636 asymptomatic women that the score was an independent predictor of mortality (HR 2.2 for a DTS <5 compared with ≥5), but that exercise capacity alone, measured in METs, was equally strong prognostically. ST depression and exercise-induced angina added no further prognostic information in this asymptomatic population [4].
An age-adjusted modification, in which age was added to the score, improved the AUC from 0.76 to 0.80 in a cohort of 1,759 male veterans [5]. This underlines that the score in its original form does not take age into account, even though age is one of the strongest prognostic factors in coronary artery disease.
Limitations
The score applies only to patients who complete a full exercise test using the Bruce protocol. Patients who stop the test early because of musculoskeletal pain or poor fitness rather than cardiac symptoms record a short exercise time that penalises the score disproportionately. If the exercise test is performed with a protocol other than Bruce, or if exercise time cannot be measured in minutes in a standardised way, the score cannot be calculated.
Patients with an abnormal resting ECG, for example with left bundle branch block or left ventricular hypertrophy, were not included in the derivation cohort, and ST deviation is difficult to interpret in these patients. The Lauer cohort explicitly excluded patients with an abnormal resting ECG [2]. With a pacemaker, ST analysis during exercise is essentially unusable.
The score is prognostic, not diagnostic. A high value does not exclude coronary artery disease, and a low value does not confirm the absence of ischaemia. It indicates the probability of future death, not the presence of stenosis. For a diagnostic question the score should not be used on its own.
Beta blockers and other rate-limiting drugs can reduce exercise capacity and shift the score downwards without any change in the underlying prognosis. This is a systematic source of error when interpreting the score in patients on such medication.
Asymptomatic individuals are not the target population. In asymptomatic women, the ST response and the angina index add no prognostic information beyond exercise capacity [4], and the score should not be used to screen healthy people.
References
- Mark DB, Shaw L, Harrell FE Jr et al. Prognostic value of a treadmill exercise score in outpatients with suspected coronary artery disease. N Engl J Med. 1991. PMID: 1875969
- Lauer MS, Pothier CE, Magid DJ et al. An externally validated model for predicting long-term survival after exercise treadmill testing in patients with suspected coronary artery disease and a normal electrocardiogram. Ann Intern Med. 2007. PMID: 18087052
- Alexander KP, Shaw LJ, Shaw LK et al. Value of exercise treadmill testing in women. J Am Coll Cardiol. 1998. PMID: 9822093
- Gulati M, Arnsdorf MF, Shaw LJ et al. Prognostic value of the Duke treadmill score in asymptomatic women. Am J Cardiol. 2005. PMID: 16054460
- Rafie AH, Dewey FE, Myers J et al. Age-adjusted modification of the Duke Treadmill Score nomogram. Am Heart J. 2008. PMID: 18513516
