The authors develop a prediction model to identify T2DM at high risk of new-onset HF, including those treated with a SGLT2-inhibitor.

This model was derived from the ALTITUDE trial (aliskiren), was validated internally (with data from ALTITUDE) as well as externally in an independent cohort of patients (from the CANVAS trial).

As a novelty of the study, this model was tested in patients without prior HF who were using an SGLT2-inhibitor for a year, which has not been done.

The authors report that a model that included NT-proBNP, TnT and BMI accurately predicted new-onset HF, with consistent results in internal and external validation.

Additionally, the use of calcium channel blocker (CCB), UACR, HbA1c, age and hematocrit emerged as significant predictors in the derivation cohort. This new model had a superior C-statistics of 0.84 than the conventional WATCH-DM score (C-statistic of only 0.71).

The authors are to congratulated for developing a novel score of notable robustness and predictive value; of note, this study is validated internally (same population) and externally (in a different population).

Of the highest importance, while all models are generated in placebo-arms of trials, this model is validated also in the SGLT2i-treated arm of CANVAS; of note, the predictive ability was excellent, even better than in the placebo arm.

The methods are very solid, the results seem logical, the discussion is balanced, the conclusions are supported by the results, and the manuscript is short and reads well.

I only raise Minor Comments:

• Discussion, first sentence: I am fairly positive Refs 23-24 do not fit here

• Does new-onset HF also include ED visit due to HF reagudization (if the patient is not hospitalized), similar to the DELIVER trial?

• Is this score equally predictive in ischemic vs non-ischemic patients?

The authors have satisfactorily addressed my previous concerns