Sokal Score for Chronic Myeloid Leukemia (CML)

The Sokal score is a prognostic tool used to estimate how aggressively chronic myeloid leukemia (CML) will behave in a given patient, sorting people into low-, intermediate-, and high-risk categories based on a handful of clinical measurements taken at the time of diagnosis. Developed decades ago from data on patients treated with older chemotherapy drugs, it remains one of several risk-stratification systems that oncologists use when deciding how to treat CML. Its relevance has shifted considerably since the arrival of targeted therapies, and newer scoring systems now outperform it in certain patient groups, but it continues to appear in clinical guidelines and research publications worldwide.

How the Score Was Built

The Sokal score traces back to work carried out between 1962 and 1981 on 813 patients with CML in its chronic phase. After multivariate analysis of the clinical features that predicted survival, a formula was published in 1984 by an international study group. The vast majority of those patients were treated with busulphan, an alkylating chemotherapy agent that was standard at the time but has long since been replaced by more effective drugs.¹ The study was retrospective, meaning it looked backward through patient records rather than following patients forward through treatment, and it relied entirely on disease features present at diagnosis.²

That origin matters because the patients used to build the score were treated under conditions that look nothing like modern CML care. Busulphan controlled white blood cell counts but did not target the underlying genetic driver of the disease. Survival outcomes for those patients were dramatically worse than what patients experience today. A scoring system calibrated on that population was, in effect, measuring how quickly CML would overwhelm a treatment that could slow it but not fundamentally change its course.

What the Score Measures

The Sokal formula takes four clinical variables available from a standard workup at diagnosis: the patient’s age, spleen size (measured by physical exam as centimeters below the rib margin), platelet count, and the percentage of blasts (immature white blood cells) circulating in the blood. Each variable is weighted differently. The formula produces a number, and that number places the patient into one of three risk groups.

Patients with a low Sokal score are expected to have the most favorable course, while those in the high-risk group face a greater chance of the disease transforming into a more dangerous accelerated or blast phase. The intermediate group falls in between. In principle, this stratification lets clinicians make earlier, more informed decisions about how intensively to treat someone from the very start. Age pulls the score upward, which means older patients tend to land in higher risk categories even when their other numbers look similar to those of younger patients. This built-in age weighting becomes a real problem in certain populations, as we will see.

How Risk Scores Guide Treatment Choices

Modern CML treatment revolves around tyrosine kinase inhibitors, a class of targeted drugs that block the abnormal protein produced by the Philadelphia chromosome, the genetic hallmark of CML. Imatinib was the first of these drugs and remains widely used as a first-line option. Second-generation TKIs like nilotinib and dasatinib are more potent and can be used upfront or after imatinib fails to achieve adequate response.

Risk scores including the Sokal score factor into which TKI a patient receives first. Higher-risk patients may benefit from starting on a second-generation TKI rather than imatinib because of the more aggressive biology of their disease.³ This is not a trivial distinction. One study found that patients with high Sokal scores who were started on imatinib failed to achieve early molecular response (a key milestone measured at three months) in about 56% of cases, compared with just 14 to 18% of high-risk patients started on nilotinib.⁴ That gap suggests high-risk patients in particular stand to gain from a stronger initial drug.

In practice, the choice of TKI also depends on the patient’s other health conditions, side-effect profile, age, and preferences. The Sokal score is one input among several, not a stand-alone prescription. But the fact that risk stratification at diagnosis can meaningfully change the likelihood of hitting treatment milestones means these scores carry real clinical weight.

The Competing Scoring Systems

The Sokal score is not the only game in town, and it is arguably no longer the best one. Three other scoring systems have been developed to address its limitations, each reflecting a different treatment era or analytic approach.

Hasford (Euro) score: Added eosinophil and basophil percentages to the variables used by Sokal. It was designed to better discriminate between risk groups in the interferon-alpha era, which preceded TKIs.
EUTOS score: Developed specifically during the imatinib era, using only spleen size and basophil percentage. It is simpler than Sokal and was designed to predict cytogenetic response to imatinib.
ELTS score: The EUTOS long-term survival score, the newest of the four, was built to predict the probability of dying specifically from CML rather than from other causes. It uses age, spleen size, platelet count, and blood blast percentage.

Despite being developed when standard treatments for CML were very different, all four scores remain in use for predicting survival outcomes.⁵ The EUTOS score predicts event-free survival under imatinib therapy better than either the Hasford or the Sokal systems.⁶ One head-to-head comparison found that patients with a low EUTOS score achieved complete cytogenetic response within 18 months at significantly higher rates than those with a high EUTOS score, whereas the Sokal-based distinction between groups did not reach clear statistical significance for that same endpoint.⁷

Why the ELTS Score Is Gaining Ground

The strongest challenger to the Sokal score is the ELTS score. In a validation study of nearly 3,000 patients who were not part of any score’s development cohort, the ELTS score showed the best ability to distinguish long-term survival across all comparisons. About 57% of patients fell into the ELTS low-risk group, 29% into intermediate risk, and 14% into high risk. Both the intermediate and high-risk groups had significantly higher probabilities of dying from CML than the low-risk group, with the high-risk group facing roughly five and a half times the hazard.⁸

A key finding from the same study was that the Sokal score classified too many patients as high risk and placed only about 40% of patients in its low-risk category, compared with 60% by the ELTS score. That difference is not just academic bookkeeping. Inappropriate risk classification can push patients toward more aggressive (and more toxic) treatments they do not need, or conversely fail to flag patients who would benefit from stronger therapy upfront.⁹ When the stakes involve years on a daily medication with real side effects, getting the risk category right matters.

The Age Problem at Both Ends

The Sokal score’s age weighting creates particular issues at the extremes of the age spectrum. Because the formula treats age as a continuous variable that pushes the score higher, older patients are almost mechanically sorted into higher risk groups. In patients 65 and older, the Sokal score and the ELTS score diverge dramatically. One analysis found that only 8% of elderly patients classified as low risk by the ELTS score also received a low Sokal score. The agreement was much better in younger adults: among patients under 30, the two scores agreed on low-risk classification about 87% of the time.¹⁰

For elderly patients, only the ELTS score was able to predict the achievement of deep molecular responses. In that group, ELTS-defined low-risk patients hit the deeper response milestone (known as MR4) in about 82% of cases, versus 61% for intermediate and 50% for high-risk patients. The Sokal score did not make similarly useful distinctions in this age group for molecular response, though both scores could still predict overall survival.¹¹ This is why some researchers argue strongly that the ELTS score should replace the Sokal score as the default tool in older patients.

At the other end of the spectrum, pediatric CML poses its own challenges. Children diagnosed with CML tend to present with higher white blood cell counts and larger spleens than adults, features that inflate the Sokal score. In one pediatric cohort, about 62% of children were classified as low risk by Sokal, but neither the Sokal nor the Hasford score correlated with early treatment outcomes. Only the EUTOS score showed a statistically meaningful relationship with early response in children.¹² The takeaway is that a score developed on middle-aged adults treated with 1970s-era chemotherapy does not necessarily translate to a nine-year-old on a modern TKI.

Beyond the Baseline Score

All of the scoring systems discussed so far share one limitation: they are snapshot tools. They take measurements at diagnosis and produce a number that is supposed to characterize the disease’s behavior going forward. But CML is a disease whose trajectory becomes much clearer once treatment starts. How quickly a patient’s leukemia burden drops in the first three to six months of TKI therapy, measured as early molecular response, has turned out to be a powerful predictor in its own right.

A patient classified as high risk by Sokal who nonetheless achieves a rapid molecular response may do just as well as a low-risk patient. Conversely, a low-risk patient whose molecular levels plateau early may need a treatment change sooner than their reassuring baseline score suggested. This dynamic reassessment has become central to modern CML management. Studies examining early molecular response found that it could refine or even override baseline risk stratification, particularly for patients started on second-generation TKIs where response rates are higher across all risk groups.¹³

In practice, this means the Sokal score (or whichever baseline score is used) matters most for the initial treatment decision. Once therapy is underway, molecular monitoring takes over as the primary guide. The score does not disappear from the clinical picture, but its influence fades as real-world treatment response data accumulates.

When Advanced Disease Enters the Picture

The Sokal score was designed for chronic-phase CML, the earliest and most treatable stage. It was never intended to predict outcomes once the disease transforms into accelerated phase or blast crisis, the more dangerous stages where immature blood cells flood the bone marrow. In advanced-phase disease, the blast count itself becomes the dominant prognostic factor, dwarfing the variables used in baseline scoring systems.¹⁴ Age, hemoglobin levels, and the presence of additional chromosomal abnormalities also matter in advanced phases, but the percentage of blasts is what drives prognosis most.

There are occasional case reports where a high Sokal score at diagnosis seems to foreshadow rapid transformation. One report described a patient with a high Sokal score who also showed elevated expression of a gene called HOXA9 and entered blast crisis within months, dying five months later.¹⁵ But individual case reports do not establish causation. The Sokal score captures some of the same biology that drives transformation (large disease burden, high blast percentages), so a high score and rapid progression can co-occur without the score being the mechanism for prediction. The score is a correlate of disease aggressiveness, not a crystal ball for blast crisis specifically.

Common Misconceptions About Risk Scores

People encountering the Sokal score for the first time, often after a CML diagnosis, tend to misunderstand what it does and does not tell them. One frequent error is treating the risk category as a survival prediction. A high Sokal score does not mean someone will die from CML. It means the statistical likelihood of certain treatment milestones being missed is higher in a group of patients with similar numbers. With modern TKIs, even many high-risk patients achieve deep and durable responses.

Another misconception is that the score is static and final. As discussed above, treatment response data quickly becomes more informative than any baseline number. A patient’s risk category at diagnosis is a starting point for a conversation about treatment intensity, not a permanent label. Clinicians who manage CML regularly reassess response at defined time points and adjust therapy accordingly, regardless of the initial score.

A third misunderstanding relates to score selection. Because four different scoring systems exist and they can assign the same patient to different risk categories, patients sometimes assume one system must be “right” and the others wrong. The reality is that each system was optimized for a different treatment era and a slightly different outcome. The Sokal score predicts overall survival reasonably well, the EUTOS score was designed around cytogenetic response to imatinib, and the ELTS score focuses on CML-specific death. Choosing among them depends partly on what question the clinician is asking and what treatment the patient is receiving.

Where the Sokal Score Still Shows Up

Despite the growing evidence that newer scores outperform it, the Sokal score persists in clinical practice for several practical reasons. It has been validated in more datasets and across more populations than any competing system. Many clinical trials still stratify patients by Sokal risk, making it a common language across research publications. International treatment guidelines from organizations like the European LeukemiaNet reference it alongside other scores, and some clinicians who have used it for decades continue to rely on its familiar categories.

In resource-limited settings where access to molecular monitoring may be inconsistent, the Sokal score’s reliance on simple, widely available clinical measurements gives it an edge over systems that require more specialized laboratory infrastructure. A physical exam for spleen size, a basic blood count, and the patient’s age are available virtually everywhere. Whether that simplicity is worth the trade-off in accuracy is debatable, but it explains part of the score’s staying power in global practice.

Research papers still frequently report Sokal categories even when using newer scores as the primary analytic tool, partly to allow historical comparisons. If a trial from 2008 reported outcomes by Sokal risk and a trial from 2023 wants to compare its cohort, both need the Sokal data to line up. This backward compatibility keeps the score embedded in the CML literature long after its predictive edge has been surpassed by alternatives designed for the TKI era.