Response Letter to the Editor from Zhu et al: “Performance of the Aldosterone-to-Renin Ratio as a Screening Test for Primary Aldosteronism: A Systematic Review and Meta-Analysis”
Year: 2021
Type: letter
Abstract: We thank Zhu et al. for their interest in our systematic review and meta-analysis and for the opportunity to respond to their comments (1, 2). First, we excluded the Aldosterone-renin Ratio for Primary Aldosteronism (AQUARR) study (3) because it did not meet the a priori inclusion criteria. We clearly stated that for inclusion, studies had to implement guideline-recommended confirmatory testing among individuals with both “positive” and “negative” aldosterone-to-renin ratio (ARR) screening (1). In contrast, AQUARR did not use confirmatory testing as the reference standard and instead used the diagnosis of an aldosterone-producing adenoma (APA) via the ‘4 corners criteria’ (3). Therefore, in addition to exclusion based upon a different reference standard and being limited to a narrow subtype of primary aldosteronism (PA; aldosterone-producing adenoma), which was simply not what our meta-analysis was examining, the design of AQUARR was such that individuals with negative ARR tests were not subject to the reference standard verification. This is because it would have been unethical to test the “4 corners” criteria, including adrenal venous sampling and surgical adrenalectomy, for all participants. Applying a reference standard to only individuals with positive screening tests leads to verification bias and overestimation of diagnostic test performance (4). Notably, a large number of individuals (N = 1125) that made up AQUARR (3) came from the PA Prevalence in Hypertensives study (5), which was already included in our analysis. In regard to including studies with normotensive individuals, we acknowledge that the ARR is most commonly used in clinical practice for hypertensive individuals. However, in reality, a modern-day understanding of PA includes normotensives as well (6). Because our interest was the diagnosis of PA, rather than hypertension, we did not exclude studies on this basis. Regarding sensitivity and specificity, these measures are generally believed to be independent of disease prevalence, which is purposefully why we selected these measures in our meta-analysis as opposed to positive and negative predictive values, which do vary by disease prevalence (7). If sensitivity and specificity appear to vary with disease prevalence, this effect is usual the result of alternative mechanisms such as disease spectrum (8). We would also like to highlight that the diagnostic odds ratio, which was suggested as an alternative, is actually derived from sensitivity and specificity and would not avoid Zhu et al.’s suggested problem. We agree, and state as much within our manuscript (1), with the point that ARR performance varies based on testing conditions that were not standardized. This made comparisons between studies challenging. In regard to ARR reproducibility, though we acknowledge this to be an important feature to the test’s generalizable applicability in clinical practice, it does not mean that the ARR is accurate. It is important to maintain a distinction between the precision and accuracy of diagnostic tests. We agree that the assumptions underlying confirmatory testing (i.e., nonsuppressibility of aldosterone) may not necessarily be true for all cases of PA and could potentially result in misclassification. However, without using confirmatory testing as a surrogate reference standard, the only way to confidently know the false-negative rate of ARR testing is to subject everyone with a negative ARR to gold standard verification, which would have to entail clinical outcomes (e.g., response to mineralocorticoid receptor antagonist therapy or surgical adrenalectomy). This is not captured by the 5-corners criteria as it would be unethical to subject individuals with negative ARR testing through the necessary procedures. Therefore, until there is an acceptable gold standard that can be applied completely to both individuals with positive and negative ARR measures, we are forced to settle for the surrogate reference standard. Finally, we agree that resetting the threshold for any diagnostic test can affect (for better or worse) a test’s sensitivity and specificity. This is simply a mathematical function and not specific to PA. Therefore, it is at the discretion of the clinician to select the threshold to be used based on their priorities between sensitivity and specificity because there will be inherent tradeoffs between the 2. This “moving threshold” approach to ARR has not been used in most studies pertaining to ARR interpretation and, as such, our manuscript simply highlights the fallibility of a single ARR threshold for a disease (PA) that exists on a biochemical continuum. Disclosures: The authors declare no conflict of interest. Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.
Language: English
Subfield: Endocrinology, Diabetes and Metabolism
Field: Medicine
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W3173284991