Biomarker tests for invasive Candida infection in critically ill patients in the UK (A-Stop): a multicentre diagnostic test accuracy study

Background

The current practice of empirical antifungal treatment for suspected invasive Candida infection in the intensive care unit (ICU) leads to overtreatment. Biomarkers might better target treatment; however, their diagnostic accuracy in this patient group is uncertain. This study aimed to evaluate whether a biomarker or a combination of biomarkers has sufficiently high negative predictive value (NPV) to enable early cessation of empirical antifungal treatment in patients in ICUs.

Methods

In this prospective, multicentre, diagnostic test accuracy study, we recruited patients who had been started on empirical antifungal treatment for suspected candidiasis and had blood cultures drawn within 24 h of starting treatment from 43 UK ICUs. We tested a research blood sample, collected within 24 h of the standard-of-care care blood culture (ie, within 48 h of starting empirical antifungal treatment), with three DNA detection tests (Fungiplex, T2Candida, and CandID) and one (1,3)-β-D-glucan test (Fungitell). We assessed the accuracy of these tests against an international consensus definition of proven invasive Candida infection and a constructed reference standard for proven or probable infection. The primary outcome was NPV for exclusion of proven Candida infection; we also reported sensitivity, specificity, and likelihood ratios.

Findings

Among 4005 patients screened, 1251 were enrolled. A total of 1225 participants, recruited between April, 2018 and Feb, 2023, were evaluable, among whom 54 (4·4%) had proven invasive Candida infection. The median duration of antifungal treatment was 9 days (IQR 5–15), and survival at 28 days was 76·8%. The index tests had similar NPVs: 96·6% (95·3–97·6) for Fungiplex, 97·1% (94·7–98·6) for T2Candida, 96·7% (95·4–97·8) for CandID, and 97·2% (95·5–98·4) for Fungitell; however, sensitivity varied widely among the tests, from 27·5% to 68·6%. The optimal test combination was Fungitell and CandID, which had an NPV of 97·5% (95·7–98·7) and a sensitivity of 75·0%. Combining these biomarkers with standard-of-care blood culture testing further increased the NPV to 98·6% (97·0–99·4), with a sensitivity of 85·1%.

Interpretation

No single biomarker had sufficiently high sensitivity to justify stopping empirical antifungal treatment. Combining Fungitell and CandID with standard-of-care blood culture testing could provide a sufficiently high NPV to allow early cessation of empirical antifungal treatment in this patient group, thereby reducing overtreatment, although this gain in sensitivity is accompanied by a reduction in specificity. Randomised trials are needed to confirm the clinical and cost-effectiveness of this diagnostic strategy before it is adopted in practice.

Funding

UK National Institute for Health and Care Research Health Technology Assessment Programme and the Research Charitable Trust Fund of the Belfast Health & Social Care Trust.