Myeloproliferative Neoplasms (MPN) harbor highly recurrent driver mutations affecting targetable kinases yet treatment options for these phenotypically diverse diseases are limited, and patients experience significant morbidity and shortened survival. The most important disease-related complications—thrombosis, transformation and death—are not used as clinical trial endpoints due to the long follow-up required to assess such disease modifying activity. A reliable monitoring biomarker linking MPN biology with these important clinical outcomes is missing. MPN driver mutation allele frequency (MAF) from whole blood or marrow (WB) does not faithfully predict MPN phenotype, clinical progression or response. This is likely because WB MAF is a composite measure of alleles from a heterogenous and variable mixture of mature leukocytes and, as such, does not report any information about the critical MPN stem and progenitor cells (MPN-SPCs). Driver mutations allow MPN cells to outcompete their normal hematopoietic counterparts and this competitive advantage—increased “fitness”—underlies core biology of MPN pathogenesis. We developed an approach to directly measure MPN fitness from samples. We measured fitness in 115 samples from 84 patients with JAK2V617F MPNs by quantifying MAF of 11 well-defined and strictly validated hematopoietic stem, progenitor and mature cell populations purified from routinely collected blood and marrow specimens. Unsupervised, hierarchical clustering of MPN fitness revealed 4 major fitness levels: F1, F2, F3, and F4 with significantly different but overlapping clinical features and diagnoses. Notably, these four fitness levels were associated with significantly different event-free survival (EFS): 95% (F1), 81% (F2), 73% (F3), 50% (F4) at 24 months (log-rank p=0.017). In contrast, WB MAF quartile failed to predict EFS. Multivariable models showed that fitness was associated with event risk independent of age, sex, duration of disease, MPN diagnosis and WB MAF. Principal component analysis allowed convenient projection of the 11-component MAF fitness measures to reduce dimensionality and develop a model for relative risk (RR) of event that could be used to assess individual or serial samples. Serial samples with more than a year of follow-up was available for 13 patients. We found that a reduction of this RR score was associated with a therapeutic response (p=0.045). In contrast, increasing RR overtime portended a disease-related event (p=0.045). Changes in WB MAF did not correlate with RR (r2=0.022) possibly explaining why WB MAF failed to predict events. These data demonstrate that fitness dynamics from serial blood samples can be used as a monitoring biomarker to assess changes in RR over time. Thus, fitness risk is a promising endpoint alongside corresponding clinical parameters such as blood counts, spleen size and marrow fibrosis grade. Our study offers a feasible approach to monitor the MPN biology central to disease progression and can be used in clinical trials to efficiently identify disease-modifying, potentially life-prolonging treatments.