Incidental detection in the DANCAVAS trial was limited by design and did not bias the cost effectiveness conclusion

This commentary refers to ‘Cost effectiveness of population screening versus no screening for cardiovascular disease: the Danish Cardiovascular Screening trial (DANCAVAS)’, by R. Søgaard et al., https://doi.org/10.1093/eurheartj/ehac488 and the discussion piece ‘More costs than benefits from screening for cardiovascular disease with computed tomography in the DANCAVAS study’, by N. H. Andersen et al., https://doi.org/10.1093/eurheartj/ehac634.

Andersen and coauthors raise a concern that the use of CT for general population screening could lead to a huge number of incidental detections of non-cardiovascular abnormalities, which will solely harm citizens and increase health care costs.¹ We understand their concern in the context of routine practice and clinical guidelines for CT-based diagnostic work up, and we agree with the implicit view that the healthcare system need to somehow regulate the huge number of referrals flowing from a CT focused on something else. But that was not the case in the DANCAVAS trial.

After careful consideration and ethical discussion, the DANCAVAS trial design defined a scanning mode largely blinded to neighbouring organs to the abdominal and coronary arteries. Technicians (radiographs) interpreted the scans, and they were trained to measure the coronary artery calcium score and the aortic diameters, and to evaluate the heart rhythm.² They were not trained to evaluate or describe anything else. There were, however, exceptions where a large tumour could not be overlooked, and these scans were referred to the radiologists for ethical reasons. This approach was carefully considered in the design, approved by the ethical committee, and explicitly addressed in the participant consent material stating that ‘we do not look for cancer or any other abnormalities unrelated to the listed cardiovascular conditions’.

After five years of follow up in the DANCAVAS trial, we concluded that screening is cost-effective.³ For the cost side, the conclusion is based on a costing protocol including all general practice services, cardiovascular disease-related prescription medication, and all hospital service related to cardiovascular disease. If we extend the hospital cost to include also service unrelated to cardiovascular disease, the cost difference between the randomization groups changes sign such that health care for the screened group becomes on average cheaper than health care for the control group. Due to the strength of the design (randomized, population scale) and the quality of Danish register-based cost data (almost 100% completeness and sensitivity) we can say with great certainty that unobserved incidental detections do not bias our results in the direction suggested by Andersen and coauthors. We are currently investigating the more detailed downstream consequences to point at exactly where cost savings appear to occur.

For the benefit side, an important difference between mortality risk reduction and the quality-adjusted life year outcome, which is recommended for cost effectiveness evaluation, is that harm but also good to quality of life is considered.⁴ This means that the cost effectiveness evaluation does not just adapt the simple mortality risk reduction outcome⁵ but contributes with a more comprehensive outcome actually reflecting citizens’ and patients’ preferences.

References

Author notes