Further insights into the application of the TFR prognostic score (TPS) in CML patients attempting TFR

We thank the authors for their comments on our recently published treatment‐free remission (TFR) prognostic score (TPS) and for proposing a dynamic interpretation of its clinical use.

The TPS integrates age at diagnosis, sex, time to MR ⁴ and duration of sustained deep molecular response (DMR; ≥ MR⁴) prior to treatment discontinuation (duration of MR ⁴ ) to stratify patients by probability of TFR (pTFR). ¹ Among these variables, duration of MR ⁴ emerged as the strongest predictor, consistent with findings of the EURO‐SKI trial ² , ³ and supported by the recent J‐SKI study results. ⁴

Although a longer duration of MR ⁴ is strongly associated with improved TFR outcomes and is the only potentially modifiable factor in the model, this relationship does not establish that prolonging MR⁴ in an individual patient will increase the likelihood of TFR. Accordingly, the TPS was developed as a prognostic tool at the time of TKI discontinuation, rather than as a dynamic prediction model.

While recalculating the TPS at later time points after deferring discontinuation may appear reasonable, such an approach would require validation in longitudinal or time‐updated modelling frameworks, which was beyond the scope of our study.

Furthermore, continued TKI therapy during DMR does not guarantee improved outcomes, as some patients may lose DMR over time and become ineligible for TFR.

However, the model can nonetheless inform decision at the time of assessment to defer TKI discontinuation in patients with a low predicted pTFR; this approach necessarily entails continued therapy and is therefore likely to result in a longer DMR duration at a later attempt.

Decisions regarding TKI discontinuation remain preference‐sensitive and are shaped by patient–physician discussions, including comorbidities, treatment tolerance, risk attitude and willingness to adhere to intensive monitoring. The TPS, once externally validated, should be considered a decision support tool to inform shared decision‐making, rather than a prescriptive instrument.

Importantly, the validity of the TPS and its associated calculator is limited to clinical settings consistent with the derivation cohort, namely, patients treated with TKIs for at least 3 years and with sustained DMR for at least 1 year prior to discontinuation. From a statistical perspective, interpretation of duration of MR ⁴ is constrained by the observed range in the training data (median 5.9 years; range 1–18.6 years). Extrapolation beyond this range is possible but should be interpreted with caution.

Regarding the β coefficient from the Cox model, the proportional hazards assumption was satisfied, indicating that the effect of duration of MR ⁴ on the hazard remained constant over time. However, this does not preclude a non‐linear association between duration of MR ⁴ and pTFR. Visual inspection of Figure 1 suggests a possible attenuation of benefit at longer DMR durations, although this warrants cautious interpretation.

FIGURE 1.

Predicted probability of treatment‐free remission (pTFR) according to duration of MR ⁴ in the training set. ¹ The horizontal axis represents the duration (in years) of sustained deep molecular response (DMR; ≥MR⁴) before TKI discontinuation (duration of MR ⁴ ), and the vertical axis shows the estimated pTFR. The solid line depicts model‐based estimates from a Cox proportional hazards model fitted to the training set, ¹ with the shaded area indicating the 95% confidence interval. The probability of maintaining major molecular response after TKI discontinuation increases with longer DMR durations. [Colour figure can be viewed at wileyonlinelibrary.com]

Future research incorporating longitudinal data and flexible modelling approaches will be required to determine whether dynamically updated prediction tools can reliably inform the optimal timing of TKI discontinuation.

We would welcome continued discussion on optimizing predictive strategies for TFR in chronic myeloid leukaemia and, for accessibility, we have provided an easy‐to‐use online TPS calculator at the following link: https://www.imperial.nhs.uk/tfr‐cml‐calculator.

AUTHOR CONTRIBUTIONS

Simone Claudiani: Conceptualization; methodology; data curation; writing – original draft; writing – review and editing; formal analysis; project administration; supervision; visualization. Silvia Metelli: Writing – review and editing; visualization; methodology; software; formal analysis; data curation. Dragana Milojkovic: Writing – review and editing; supervision; resources.

CONFLICT OF INTEREST STATEMENT

The authors have no conflicts of interest to disclose.

DATA AVAILABILITY STATEMENT

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.