The Most Readable Signal of All

Multiple myeloma is diagnosed in approximately 35,000 Americans each year. The five-year survival rate has improved substantially with modern therapy, reaching 62 percent overall. However, the disease is almost universally diagnosed at an advanced systemic stage, typically after the complete blood count (CBC) and comprehensive metabolic panel (CMP) have been showing abnormalities for years.¹ Of all 13 cancers in this series, myeloma has the most unambiguous and earliest pre-diagnostic blood signature.

Multiple myeloma announces itself in the CBC two to three years before diagnosis. The signal is among the clearest in all of oncology.

The Signature in the Blood

Myeloma arises from plasma cell proliferation in the bone marrow, suppressing normal hematopoiesis. A United Kingdom primary care case-control study of 2,703 myeloma cases found that hemoglobin decreased three years before diagnosis; erythrocyte sedimentation rate (ESR) was elevated two years before (85 percent abnormal in the year before diagnosis, odds ratio [OR] 5.7); hypercalcemia carried an OR of 11.4; raised creatinine OR 1.8; raised mean corpuscular volume (MCV) OR 3.1.²

A landmark New England Journal of Medicine study of 1,384 monoclonal gammopathy of undetermined significance (MGUS) patients, followed for over 11,000 person-years, confirmed that essentially all multiple myeloma is preceded by MGUS, which itself produces a detectable protein gap on the CMP (an elevation in total protein disproportionate to albumin) years before progression.³ A Scandinavian study of 1.1 million blood donors confirmed hemoglobin decline begins three years before myeloma diagnosis, the earliest pre-diagnostic decline among all cancers studied.⁴

The Machine Learning Case

An AdaBoost model trained on routine serum biomarkers achieved an area under the curve (AUC) of 96.8 percent and an accuracy of 92.6 percent for myeloma detection from standard blood inputs.⁵ A random forest model trained on CBC and cell population data parameters, with features selected via LASSO, achieved an AUC of 0.956 in training and 0.875 in the test cohort in 465 myeloma patients and 150 controls.⁶

What This Proves

Multiple myeloma has arguably the strongest pre-diagnostic blood signal of any cancer in this series. Hemoglobin decline, ESR elevation, hypercalcemia, creatinine rise, and protein gap abnormalities are all documented in the CBC and CMP years before clinical diagnosis. Machine learning models achieve near-diagnostic AUC from these standard inputs. The science is established, and the algorithmic performance is exceptional. However, the population-scale screening tool does not yet exist.

Endnotes

1.National Cancer Institute. “Cancer Stat Facts: Myeloma.” SEER, 2024.

SEER data reporting overall five-year relative survival of 62% for multiple myeloma, with the vast majority of cases presenting at an advanced systemic stage. Annual incidence of approximately 36,000 and 12,000 deaths. Establishes the disease burden and staging context.

2.Koshiaris C, Van den Bruel A, Oke JL, et al. “Early detection of multiple myeloma in primary care using blood tests: a case-control study in primary care.” Br J Gen Pract. 2018;68(674):e586–e593.

Case-control study of 2,703 myeloma cases and 12,157 matched controls from UK CPRD. Hemoglobin declined three years before diagnosis; ESR elevated two years before (OR 5.7); hypercalcemia OR 11.4. Establishes the multi-year pre-diagnostic CBC and CMP timeline with specific odds ratios.

3.Kyle RA, Therneau TM, Rajkumar SV, et al. “A Long-Term Study of Prognosis in Monoclonal Gammopathy of Undetermined Significance.” N Engl J Med. 2002;346(8):564–569.

Mayo Clinic study of 1,384 MGUS patients over 11,009 person-years confirming 1% annual progression rate to myeloma or related malignancy. Establishes the MGUS-to-myeloma continuum and the protein gap on CMP as the primary pre-clinical blood signal detectable years before progression.

4.Edgren G, Bagnardi V, Bellocco R, et al. “Pattern of declining hemoglobin concentration before cancer diagnosis.” Int J Cancer. 2010;127(6):1429–1436.

Nested case-control study within 1.1 million Scandinavian blood donors showing hemoglobin decline beginning three years before myeloma diagnosis—the earliest pre-diagnostic hemoglobin change among all cancer types studied. Confirms that the CBC carries a myeloma signal years before clinical suspicion.

5.Fan G, Cui R, Zhang R, et al. “Routine blood biomarkers for the detection of multiple myeloma using machine learning.” Int J Lab Hematol. 2022;44(3):558–566.

AdaBoost model trained on routine serum biomarkers, achieving AUC 96.8% and an accuracy 92.6% for myeloma detection. Among the highest-performing machine learning models across any cancer in this series, confirming that the CMP and CBC signal for myeloma is not only detectable but highly discriminative when processed by machine learning.

6.Cai J, Liu Z, Wang Y, Yang W, Sun Z, You C. “Construction of the prediction model for multiple myeloma based on machine learning.” Int J Lab Hematol. 2024;46(5):918–926.

Random forest model on 465 myeloma patients and 150 controls using CBC and cell population data parameters selected by LASSO. Training AUC 0.956, test AUC 0.875. Demonstrates reproducibility of high-performance machine learning myeloma detection from standard blood inputs across independent validation cohorts.