Response to Response to Letter Regarding “An Artificial Neural Network-Based Model to Predict Chronic Kidney Disease in Aged Cats”

Abstract

I thank Biourge et al. for their response to my letter and for highlighting the importance of early chronic kidney disease (CKD) diagnosis in cats. However, I am disappointed that the authors have elected not to disclose the number of cats with a baseline serum creatinine concentration between 1.6 and 1.9 mg/dL included in the testing dataset, on the basis that all cats “were healthy based on history and physical examination and had serum creatinine concentrations below the diagnostic threshold for CKD according to the reference interval of the laboratory used, and thus did not have a diagnosis of CKD at the time of screening.” The authors cite a 2019 International Renal Interest Society (IRIS) article, which states that cats with a serum creatinine concentration reaching the threshold for stage 2 CKD but within the laboratory reference interval must also have other independent evidence of CKD before a diagnosis can be made, to minimize the potential for false positive diagnoses [1]. Since this article was written, it has been shown that a serum creatinine concentration > 1.76 mg/dL has a 92.9% and 90.9% specificity for borderline-low and low glomerular filtration rate (GFR) in cats, respectively [2]. Although the specificity of a > 1.5 mg/dL creatinine cutoff was not provided in this study, an older article reports a specificity of 73% for the detection of a GFR ≤ 70% of normal [3]. Furthermore, current IRIS guidelines state that CKD may be diagnosed based on the sole finding of a serum SDMA persistently > 14 μg/dL [4], which has a 75.0% and 75.7% specificity for a borderline-low and low GFR in cats, respectively [2]. If this level of specificity is considered acceptable, it stands to reason that a serum creatinine concentration > 1.5 mg/dL (although ideally > 1.76 mg/dL) should also be sufficient.

Accordingly, Biourge et al.'s suggestion that diagnosing a cat with CKD based on a sustained serum creatinine concentration between 1.6 and 1.9 mg/dL (especially with a urine specific gravity ≤ 1.035) represents “a common misconception about the IRIS staging system,” and their following statement that these cats “are at a stage where plasma creatinine concentration is still within the laboratory reference interval and would thus be considered as normal in a regular senior screening” appears misguided. I am also not sure that their opinion is widely shared, with various key opinion leaders stating that CKD or renal azotemia in cats be considered with a serum creatinine concentration ≥ 1.6 mg/dL [5-7].

I again urge Biourge et al. to share the aforementioned data from their study. If cats with a baseline serum creatinine concentration between 1.6 and 1.9 mg/dL were considered “healthy,” their model might simply be predicting the likelihood that cats with early IRIS stage 2 CKD progress to a serum creatinine concentration ≥ 2.0 mg/dL within the subsequent 12 months. This is distinctly different from predicting the likelihood of CKD development, as the authors claim. The fact that the positive predictive value of their model (53%) is nearly equivalent to the expected percentage of cats with CKD that develop progressive azotemia within 12 months (47%) [8] is, in my opinion, somewhat suspicious for the former scenario. If this is not the case, I congratulate Biourge et al. for their model and encourage them to make it accessible to the veterinary community for widespread adoption.

Kind regards,