Introduction Prognostic models such as the Revised International Prognostic Scoring System (IPSS-R) and the Molecular IPSS (IPSS-M) have been widely used in MDS pts. However, frailty, which is essential for clinical outcomes, is not considered in these prognostic tools. A 15-item frailty score (FS-15) specific to MDS pts was developed, including laboratory indicators, self-care, physical status, fatigue assessment, tumor history, and BMI. We aim to validate the FS-15 and further explore its potential to improve the predictive ability of IPSS-R and IPSS-M.

Methods In total, 812 consecutive subjects with newly diagnosed de novo MDS from 08/2016 to 06/2023 were enrolled. Frailty index (FI) was calculted according to the FS-15; all pts miss <20% of the 15 elements. To determine the significant FI cut-off score related to the overall survival (OS), we constructed a receiver-operating characteristic (ROC) curve. We identified the point with the highest Youden index as the cut-off score (0.44). Pts were stratified into risk categories according to the IPSS-R and IPSS-M separately. Survival analysis was then performed to evaluate the impact of reclassification by FS-15 within each IPSS-M and IPSS-R risk group, and the C-index was calculated for each model to compare their predictive accuracy.

Results The median age was 55 (IQR: 42-64) years, and 65% were male. The median follow-up time was 22.5 (20.2-24.9) months (mo), and the median OS was 43.3 (95%CI: 36.8-49.8) mo. The mean FI score was 0.43 (SD = 0.13), and the median was 0.42 (IQR: 0.35-0.54), with a range of 0.08-0.81. FI score differed significantly by gender (p = 0.006), age (p = 0.014), IPSS-R (p = 0.001), and IPSS-M (p = 0.014). Using the FS-15, pts were divided into two groups based on the identified cut-off value 0.44 (FS-15 score ≤0.44 was 452 pts, and >0.44, 360 pts). Subjects with cut-off values higher than 0.44 had lower hemoglobin (P = 0.005) and higher IPSS-M categories (P = 0.010). Kaplan-Meier OS curves using the cut-off value of 0.44 showed a highly significant difference (p < 0.0001) with a median survival of 54.7 mo (95%CI: 47.5-NA) and 31.5 mo (95%CI: 22.9-41.0), respectively.

Using the IPSS-M model, 20 subjects (2.5%) were classified as very low-risk, 122 (15.0%) as low-risk, 107 (13.2%) as moderate low-risk, 99 (12.2%) as moderate high-risk, 185 (22.8%) as high-risk and 200 (24.6%) as very high-risk, 3-year overall survivals were 88.4%, 80.5%, 68.0%, 62.7%, 40.0% and 32.2% respectively(P < 0.0001). We further stratified different IPSS-M risk groups by FS-15 frailty score (≤0.44 vs. >0.44 ). The frailty index cut point further refined the projected survivals of the IPSS-M risk groups. Subjects with FS-15 scores ≤0.44 had significantly higher 3-year survival rates compared to those >0.44 in the IPSS-M low-risk group (88.7% vs. 68.2%, P = 0.026) and very high-risk group (40.3% vs. 26.4%, P = 0.036). In other IPSS-M risk groups, the 3-year survival rates of the FS ≤0.44 group were lower than that of the FS >0.44 group, but the differences were not statistically significant. The same method was used for analysis in each risk group of IPSS-R, and the differences in the very low-risk group(100.0% vs. 66.7%, P = 0.008) and the high-risk group(50.7% vs. 27.8%, P = 0.019) were statistically significant. So, a patient classified as having a higher risk disease according to the IPSS-M/IPSS-R but exhibiting a low degree of frailty as measured by the FS-15 might have a longer-than-expected lifespan.

The C-index for the IPSS-M model was 0.682 (95% CI: 0.649-0.716); after incorporating FS-15, the C-index increased to 0.699 (95% CI: 0.663-0.733). Moreover, the integrated model of FS-15 and IPSS-R also had a higher C-index than IPSS-R (0.691 [95% CI: 0.657-0.727] vs. 0.666 [95% CI: 0.626-0.733]). The addition of FS-15 significantly improved the predictive performance of the IPSS-M and IPSS-R models.

Conclusion This study firstly validated the feasibility of FS-15 evaluation and its potential use in prognostic assessment and providing clearer clinical guidance for newly diagnosed de novo MDS pts. In addition, we demonstrated that FS-15 could refine the risk stratification of IPSS-R and IPSS-M. The incorporation of the frailty score significantly enhances the predictive accuracy of the IPSS-R and IPSS-M models, suggesting that in addition to clinical, pathological, and cytogenetic data, frailty factors also need to be taken into account in outcome prediction for pts with MDS.

Disclosures

No relevant conflicts of interest to declare.

This content is only available as a PDF.
2024
Sign in via your Institution