Keeping balance: Author’s reply
Bibliographic record
Abstract
We thank Dr Müssigbrodt1 for his interest in our study and the complex question of arrhythmogenic right centricular cardiomyopathy (ARVC) diagnosis.2 To continue the discussion in this reply, we elaborate on the elements that he found surprising and rectify inaccuracies in his statements. His first concern regards our choice for an expert panel, instead of ‘reproducible genetic, phenotypic, and clinical criteria’. However, unfortunately, such a ‘gold-standard’ test does not exist for diagnosing ARVC as none of the clinically available tests has sufficient sensitivity or specificity. Relying on genotype would exclude nearly half of cases that are ‘gene-elusive’. Diagnosis must thus result from a composite reference standard: the 2010 Task Force Criteria (TFC).3 There are two recommended evidence-based options to validate composite reference standards; (i) an expert panel or (ii) latent classification learning algorithms,4 of which the latter is suboptimal for relative small cohort sizes in ARVC-research. The appealing strength of an expert panel is that the experts can use the totality of available evidence on an individualized basis, allowing a more comprehensive assessment with multiple shades of grey instead of a limited black-and-white approach from pre-defined criteria with fixed cut-off points. Such expert panels have previously been proven to be valuable in the field of cardiology, for example, to validate the diagnostic value of B-type natriuretic peptide for heart failure.5 Secondly, Dr Müssigbrodt expressed concerns about our control group, a cohort of patients referred for evaluation of ARVC, instead of healthy individuals, and about the disappointing results of several criteria. We consider our control group a major strength of our analysis. In the real world, patients referred for ARVC evaluation have specific traits that resemble ARVC, making them much harder to discriminate. Most of the previously published diagnostic results of the 2010 TFC tests are based on comparison with healthy controls, leading to an overestimation of the true diagnostic accuracy. This explains why several tests, such as signal-averaged electrocardiogram (SAECG), showed lower values in our study. This did however not come as a surprise to us as the recently published expert consensus review of the 2010 TFC advocates for the elimination of SAECG based on ‘non-specific findings and limited diagnostic accuracy’.6 The last point raised by Dr Müssigbrodt, that our statement ‘ECG and arrhythmia criteria alone can rule out ARVC with high sensitivity’ is ‘overconfident’, might be based on a misunderstanding. We fully agree with Dr Müssigbrodt’s statement that ECG and arrhythmia criteria are insufficient to diagnose (i.e. ‘rule-in’) ARVC. Our study rather highlights the importance of the finding of a 100% sensitivity for ‘ruling-out’ ARVC, for example in relatives subject to frequent re-evaluations. As these are two different concepts not to be confused, we want to emphasize the importance of other tests such as cardiac magnetic resonance imaging (CMR) and genetics for diagnosing ARVC. As the first study validating the 2010 TFC as a whole in a real-world population, our results highlight the potential for improvements in ARVC diagnostic criteria. We, like Dr Müssigbrodt, advocate that new diagnostic criteria should maintain a balance between over- and underdiagnoses, and be based on evidence-based approaches. Conflict of interest: none declared.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.005 | 0.042 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.002 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.008 | 0.004 |
| Scholarly communication | 0.009 | 0.006 |
| Open science | 0.003 | 0.004 |
| Research integrity | 0.148 | 0.088 |
| Insufficient payload (model declined to judge) | 0.012 | 0.009 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".