2026 ASH®
Annual Meeting & Exposition December 12-15, 2026 ▪ New Orleans, Louisiana

Paper Details

Abstracts, authors, affiliations, and disclosures.

Saturday, December 12, 2026
05:30 p.m. - 07:30 p.m.
MCCNO - Hall F
Technology and Procedures, Assays, Imaging, Molecular Testing, Omics Technologies, Bioinformatics, Computational Biology, Emerging Technologies, Profiling, Treatment Considerations, Combination Therapy, Therapy Sequence, Biological Therapies, Transplantation, Bispecific Antibody Therapy, Chimeric Antigen Receptor, (CAR)-T Cell Therapies, Artificial Intelligence (AI), Computer Vision, Machine Learning, Biological Processes, Pathogenesis, Genomics, Molecular Biology, Diseases, Lymphoid Malignancies, Plasma Cell Disorders, Myeloid Malignancies, Chronic Myeloid Malignancies, Measurable Residual Disease, Research, Translational Research, Clinical Research, Clinical Trials, Health Outcomes Research, Real-World Evidence, Study Population, Human, Adult, Elderly
PAPER

2306 Early serial laboratory changes predict deep response and nonresponse at cycle 4 in multiple myeloma: Development in routine care and independent clinical-trial validation

Vaishnavi Kumbhar, MS, Moritz Binder, MD, MPH2, Wilson Gonsalves, MD3, Saurabh Zanwar, MBBS, MD4, Nadine Abdallah I, MD5, Joselle M Cook, M.D.6, Rahma Warsame, MD7, Taxiarchis Kourelis, MD8, Yi Hwa, DNP9, Amie Fonder, PA-C10, Lee Her Lee, APRN, C.N.P., D.N.P.11, Miriam Hobbs, CNP12, Francis Buadi, MD13, David Dingli, MD, PhD14, Suzanne Hayman, MD15, Morie Gertz16, Angela Dispenzieri, MD17, Prashant Kapoor, MD18, S. Vincent Rajkumar, MD19, Shaji Kumar, MD20
1 Mayo Clinic, Rochester, Minnesota, United States
2 Mayo Clinic, Rochester, Minnesota, United States
3 Mayo Clinic, Rochester, Minnesota, United States
4 Mayo Clinic, Rochester, Minnesota, United States
5 Mayo Clinic, Rochester, Minnesota, United States
6 Mayo Clinic, Rochester, Minnesota, United States
7 Mayo Clinic, Rochester, Minnesota, United States
8 Mayo Clinic, Rochester, Minnesota, United States
9 Mayo Clinic, Rochester, Minnesota, United States
10 Mayo Clinic, Rochester, Minnesota, United States
11 Mayo Clinic, Rochester, Minnesota, United States
12 Mayo Clinic, Rochester, Minnesota, United States
13 Mayo Clinic, Rochester, Minnesota, United States
14 Mayo Clinic, Rochester, Minnesota, United States
15 Mayo Clinic, Rochester, Minnesota, United States
16 Mayo Clinic, Rochester, Minnesota, United States
17 Mayo Clinic, Rochester, Minnesota, United States
18 Mayo Clinic, Rochester, Minnesota, United States
19 Mayo Clinic, Rochester, Minnesota, United States
20 Mayo Clinic, Rochester, Minnesota, United States
Background
Early identification of patients (pts) with multiple myeloma (MM) who are unlikely to achieve a deep response to therapy could support timely treatment reassessment. The end of cycle 4 (C4) is a clinically relevant time point for stem-cell transplantation and other treatment decisions. We evaluated whether serial changes in routinely monitored MM markers through the end of cycle 2 (C2) could predict ≥very good partial response (VGPR) or better and nonresponse (<PR) at C4, and whether laboratory patterns during the first 4 months could predict eventual best response (BR).

Methods
The dataset included 3,080 routine-care pts and 1,123 independent clinical-trial pts. Serial features were derived from serum M-protein, kappa and lambda free light chains, and quantitative IgA, IgG, and IgM, summarizing baseline and recent values, magnitude and rate of change, and variability.

For the C4 analyses, 912 pts had at least three serial laboratory assessments through the end of C2 (approximately 2 months). Requiring an evaluable response at C4 (approximately 4 months) and all predictor measurements to precede response assessment yielded 381 endpoint-eligible pts: 210 routine-care training, 60 internal validation, and 111 clinical-trial external validation. The same cohort and predictor window were used for complementary endpoints: VGPR or better and nonresponse (<PR; MR, SD, PD, or NR), with PR treated as an intermediate response.

For the eventual BR analysis, 2,011 pts had at least three serial laboratory assessments during the first 4 months. Requiring an evaluable eventual BR and adequate follow-up yielded 1,705 pts: 755 routine-care training, 200 internal validation, and 750 clinical-trial external validation. Feature selection was performed using routine-care data only, and the selected models were refit using all eligible routine-care pts before being applied unchanged to the corresponding clinical-trial cohort.

Results
Among the 111 C4 external-validation pts, 33 achieved VGPR or better, 41 achieved PR, and 37 remained below PR. For prediction of VGPR or better at C4, a gradient-boosting serial model achieved an AUC of 0.806 versus 0.541 for baseline measurements (ΔAUC, 0.265; 95% CI, 0.142–0.384; P=0.001), with 81.8% sensitivity, 70.5% specificity, 76.2% balanced accuracy, and 54.0% positive predictive value.

Using the same cohort and observation window, a separately optimized gradient-boosting model predicted nonresponse (<PR). The serial model achieved an AUC of 0.775 and a precision-recall AUC of 0.653 versus 0.337 and 0.292, respectively, for baseline measurements. The improvement in AUC was 0.438 (95% CI, 0.279–0.593; P=0.001), and the improvement in precision-recall AUC was 0.361 (95% CI, 0.194–0.486; P=0.002). At the validation-selected operating point, sensitivity was 67.6%, specificity 70.3%, positive predictive value 53.2%, and negative predictive value 81.3%.

Among the 750 eventual BR external-validation pts, 271 ultimately achieved VGPR or better. A random-forest serial model achieved an AUC of 0.865 versus 0.540 for baseline measurements (ΔAUC, 0.325; 95% CI, 0.285–0.364; P=0.001), with 81.5% sensitivity and 74.5% specificity.

Conclusions
Serial laboratory changes improved prediction of both deep response and nonresponse at C4 compared with baseline measurements, while laboratory patterns during the first 4 months also strongly predicted eventual best response. These models may facilitate earlier treatment reassessment, but prospective validation and further calibration are required before clinical implementation.
 
Vaishnavi Kumbhar, MS: No relevant conflicts of interest to declare.