CML Risk Scores: Sokal, Hasford, EUTOS, ELTS
CML Risk Scores: Sokal, Hasford, EUTOS, ELTS
Four published scores on one patient. They use different variables, different formulas and different scales, they classify the same patient into different risk groups, and the published agreement between Sokal and the other two is poor. The ELN 2020 panel recommends the ELTS score instead of the older three.
Sokal, Euro, EUTOS and ELTS
4 published regressions, side by sideAge 45; spleen 10 cm below the costal margin; peripheral blasts 4%; platelets 400 ×10⁹/L; basophils 4%; eosinophils 3%
Four published formulas, their cut-offs, and which one the guidelines use
low <0.8 · intermediate 0.8–1.2 · high >1.2
Euro / Hasford (1998) = [0.6666 × (age ≥ 50) + 0.0420 × spleen + 0.0584 × blasts + 0.0413 × eosinophils + 0.2039 × (basophils ≥ 3%) + 1.0956 × (platelets ≥ 1500)] × 1000
low ≤ 780 · intermediate 781–1480 · high >1480
EUTOS (2011) = 7 × basophils% + 4 × spleen
low ≤ 87 · high >87
ELTS (2016) = 0.0025 × (age / 10)3 + 0.0615 × spleen + 0.1052 × blasts + 0.4104 × (platelets / 1000)−0.5
low ≤ 1.5680 · intermediate >1.5680–2.2185 · high >2.2185
- which one the guidelines use
- the European LeukemiaNet 2020 recommendations state that ‘The panel recommends the use of the new ELTS survival score’, instead of the older ones, because ‘most patients now die from causes other than leukemia while still in remission’. ELTS predicts the probability of dying FROM CML, which is why it is the headline here
- the same four variables, four different shapes
- ELTS uses the same raw measurements as Sokal with different weights, and the recommendations say so explicitly. But the SHAPES differ: Sokal squares the platelet ratio while ELTS takes it to the power −0.5; Sokal makes age linear while ELTS cubes it; the Euro score reduces age, basophils and platelets to three on/off steps; and EUTOS drops age, blasts, eosinophils and platelets altogether
- why they disagree so often
- look at what each one ignores. Below 50 years the Euro score’s age term is zero and Sokal’s is not; below 1,500 ×10⁹/L the Euro score’s platelet term is zero and Sokal’s and ELTS’s are not; EUTOS uses only two variables. Published pairwise concordance in 114 patients: 53% Sokal against Hasford, 64% Sokal against EUTOS, 98% Hasford against EUTOS
- Sokal’s era, the oldest here by twenty years
- 813 patients in six series (435 Europe, 378 USA) diagnosed between 1962 and 1981 on CONVENTIONAL CHEMOTHERAPY — median age 43, overall median survival 47 months, 5 years in the lower-risk group against 2.5 in the high-risk group. Imatinib was licensed in 2001. The ELN 2020 recommendations note that Sokal ‘apportions more patients to the intermediate- and high-risk groups than ELTS’
- what a median is
- a median is the middle of the derivation cohort and not a forecast: half of that cohort reached the figure and half did not, and a figure measured in a cohort treated decades ago is a historical measurement rather than an expectation
- the Euro score’s era
- 1,303 patients in 14 series diagnosed between 1985 and 1996 and treated with INTERFERON-ALFA-based regimens, median age 49. Also pre-imatinib, and derived in a treatment that is no longer first-line anywhere
- EUTOS and ELTS, the two imatinib-era scores
- EUTOS was derived in 2,060 patients on imatinib and predicts the probability of NOT being in complete cytogenetic remission at 18 months. ELTS was derived in 2,205 patients starting imatinib within six months of diagnosis, validated in an independent 1,120, and predicts CML-related death. Both are about patients on a tyrosine kinase inhibitor, which neither Sokal nor the Euro score is
- two printed disagreements between sources
- first, the EUTOS boundary. The European LeukemiaNet’s own page defines low risk as 87 OR LESS and high risk as above 87; a widely read review defines low risk as BELOW 87. This page uses the authors’ own site. Second, the ELTS boundary: the ELN 2020 table prints low risk as below 1.5680 while the ELTS page itself prints 1.5680 or less, and again the authors’ own page governs here. Both disagreements affect only a patient whose score lands exactly on the boundary
- two transcription errors found and not propagated
- one European LeukemiaNet teaching document prints Sokal’s age coefficient as 0.116 rather than 0.0116, which would multiply a 60-year-old’s relative risk about sixfold; three other renderings agree on 0.0116. A review’s score table adds ‘a constant of 100’ to the Euro score, which would put every patient near 100,000 on a scale that runs to about 2,000, and renders the Sokal platelet term’s square as a trailing ‘2’. Neither error is reproduced here
Worked example
Age 45; spleen 10 cm below the costal margin; peripheral blasts 4%; platelets 400 ×10⁹/L; basophils 4%; eosinophils 3%
ELTS = 0.0025 × (45/10)3 + 0.0615 × 10 + 0.1052 × 4 + 0.4104 × (400/1000)−0.5 = 0.2278 + 0.6150 + 0.4208 + 0.6490 = 1.9125 — above 1.5680 and below 2.2185, so intermediate risk
Sokal = exp[0.0116 × 1.6 + 0.0345 × 2.49 + 0.188 × (0.3265 − 0.563) + 0.0887 × 1.9] = exp(0.2285) = 1.257 — above 1.2, so HIGH risk
Euro (Hasford) = [0 (age under 50) + 0.0420 × 10 + 0.0584 × 4 + 0.0413 × 3 + 0.2039 (basophils 4%, which is 3% or more) + 0 (platelets under 1,500)] × 1000 = 981 — between 781 and 1,480, so intermediate risk
EUTOS = 7 × 4 + 4 × 10 = 68 — 87 or less, so LOW risk
So: Sokal high, Euro intermediate, EUTOS low, ELTS intermediate. Three different risk groups on one patient, spread 2 — the maximum possible. It is the default on this page because the disagreement is the content
Where it comes from is visible in the arithmetic: at 45 years the Euro score's age term is zero and Sokal's is not; at 400 ×10⁹/L the Euro score's platelet term is zero and Sokal's and ELTS's are not; EUTOS never sees the age, blasts, platelets or eosinophils at all
The ELN 2020 recommendation settles which to report: the panel recommends ELTS instead of the older three
Which variables each score uses, and how
| Variable | Sokal 1984 | Euro / Hasford 1998 | EUTOS 2011 | ELTS 2016 |
|---|---|---|---|---|
| Age | Continuous, centred on 43.4 | Step at 50 years | Not used | Cubic in age/10 |
| Spleen (cm below costal margin) | Continuous, centred on 7.51 | Continuous | Continuous, × 4 | Continuous |
| Peripheral blasts (%) | Continuous, centred on 2.10 | Continuous | Not used | Continuous |
| Platelets | Squared, as (platelets/700)2 | Step at 1,500 ×10⁹/L | Not used | (platelets/1000)−0.5 |
| Basophils (%) | Not used | Step at 3% | Continuous, × 7 | Not used |
| Eosinophils (%) | Not used | Continuous | Not used | Not used |
| What it predicts | Overall survival on chemotherapy | Overall survival on interferon-alfa | Complete cytogenetic remission at 18 months on imatinib | Death from CML on imatinib |
| Derivation cohort | 813 patients, diagnosed 1962–1981, conventional chemotherapy | 1,303 patients, diagnosed 1985–1996, interferon-alfa | 2,060 patients on imatinib | 2,205 patients starting imatinib, validated in 1,120 |
ELTS and Sokal on the same patients, from the ELN 2020 comparison
| Risk group | Share of patients, ELTS | Share, Sokal | 10-year overall survival, ELTS | 10-year OS, Sokal | 6-year leukaemia-related death, ELTS | 6-year LRD, Sokal |
|---|---|---|---|---|---|---|
| Low | 55% | 38% | 88% | 89% | 2% | 3% |
| Intermediate | 28% | 38% | 79% | 81% | 5% | 4% |
| High | 13% | 23% | 68% | 75% | 12% | 8% |
Four scores, three answers, and why the guidelines picked one
Chronic myeloid leukaemia has four widely used prognostic scores and they do not agree. Sokal (1984) is a four-variable exponential in age, spleen size, peripheral blasts and the square of the platelet ratio. The Euro or Hasford score (1998) adds eosinophils and basophils and reduces age, basophils and platelets to three on/off steps. EUTOS (2011) is 7 times the basophil percentage plus 4 times the spleen size and nothing else. ELTS (2016) uses the same raw measurements as Sokal with a cubic age term and an inverse-square-root platelet term. Put one patient through all four and three different risk groups is a routine result rather than a contrived one — the default inputs on this page do exactly that.
The disagreement is not a defect. It falls out of what each score ignores. Below 50 years the Euro score’s age term is zero while Sokal’s is not; below 1,500 ×10⁹/L platelets — nearly every patient — the Euro score’s platelet term is zero while Sokal’s and ELTS’s are not; EUTOS never looks at age, blasts, platelets or eosinophils at all. A published comparison in 114 chronic-phase patients found pairwise concordance of 53% between Sokal and Hasford, 64% between Sokal and EUTOS and 98% between Hasford and EUTOS.
The eras explain more of it than the arithmetic. Sokal’s 813 patients were diagnosed between 1962 and 1981 on conventional chemotherapy, with an overall median survival of 47 months — 5 years in the lower-risk group and 2.5 in the high-risk group. Imatinib was licensed in 2001, two decades later. The Euro score’s 1,303 patients were diagnosed between 1985 and 1996 on interferon-alfa, which is no longer first-line anywhere. EUTOS and ELTS are the two derived in imatinib-treated patients and they predict different things: EUTOS the probability of NOT being in complete cytogenetic remission at 18 months, ELTS the probability of dying from CML.
That distinction is why the European LeukemiaNet 2020 panel recommends ELTS instead of the older scores: most patients on a tyrosine kinase inhibitor now die of something other than their leukaemia while still in remission, so a score fitted to all-cause survival in a chemotherapy-era cohort answers a question nobody is asking. On the same patients ELTS and Sokal give nearly identical overall survival by group and quite different leukaemia-related death. The ELN also treats response milestones — BCR::ABL1 transcript levels at three, six and twelve months — as the dominant prognostic information once treatment has started, with a high-risk ELTS score at baseline listed only as a warning flag. Every outcome figure here describes the cohort the index was derived in and not the patient in front of you; the spread within one stratum is wider than the gap between strata. Every laboratory threshold here is method- and laboratory-dependent, so the reader’s own laboratory’s reference interval takes precedence. A risk stratum is not a diagnosis and not a plan. This page computes the index, names the stratum and reports what it predicted in the derivation cohort; what follows is a decision for the treating team with the patient.
Frequently asked questions
Which CML risk score should I use?
The European LeukemiaNet 2020 recommendations state that the panel recommends the ELTS score, instead of the older ones. The reasoning is that Sokal and the Euro score were fitted to all-cause survival in patients treated before tyrosine kinase inhibitors, while most patients now die of something other than their CML; ELTS was fitted to CML-related death in imatinib-treated patients. All four are computed on this page because trials and records quote all four.
Why do Sokal, Hasford and EUTOS disagree?
Because they use different variables, different functional shapes and different endpoints, and were derived two decades apart. EUTOS uses only basophils and spleen size. The Euro score reduces age, basophils and platelets to steps whose thresholds most patients fall below. Sokal squares the platelet ratio where ELTS takes its inverse square root. A published 114-patient comparison found concordance of 53% between Sokal and Hasford, 64% between Sokal and EUTOS, and 98% between Hasford and EUTOS.
How is spleen size measured for these scores?
As the maximum distance in centimetres from the costal margin to the palpable splenic tip, on clinical examination. All four were derived on the clinical measurement, so an ultrasound craniocaudal length is not interchangeable with it. Enter 0 for an impalpable spleen.
Is the EUTOS low-risk boundary 87 or less, or below 87?
87 or less, on the European LeukemiaNet’s own page, which states high risk as greater than 87 and low risk as less than or equal to 87. A widely read review prints low risk as below 87 instead. The two differ only for a patient scoring exactly 87, and this page uses the authors’ own definition. The same kind of boundary disagreement exists for ELTS at 1.5680.
Related calculators
References
- Sokal JE, et al. Prognostic discrimination in “good-risk” chronic granulocytic leukemia. Blood. 1984. 813 Philadelphia-positive patients, overall median survival 47 months.
- Hasford J, et al. A new prognostic score for survival of patients with chronic myeloid leukemia treated with interferon alfa (the Euro or Hasford score). J Natl Cancer Inst. 1998.
- Hasford J, et al. Predicting complete cytogenetic response and subsequent progression-free survival in 2,060 patients with CML on imatinib treatment: the EUTOS score. Blood. 2011.
- Pfirrmann M, et al. Prognosis of long-term survival considering disease-specific death in chronic myeloid leukemia (the ELTS score). Leukemia. 2016. 2,205 imatinib-treated patients, validated in 1,120.
- European LeukemiaNet. The EUTOS long-term survival (ELTS) score and Online calculation of the EUTOS score — formulas, units and cut-offs as published by the group that derived them.
- European LeukemiaNet. European LeukemiaNet 2020 recommendations for treating chronic myeloid leukemia: the recommendation of the ELTS score, and Table 2’s ELTS and Sokal comparison.
- European LeukemiaNet. Relative risk in CML: the Sokal and Euro scores and their derivation cohorts — Sokal 813 patients, 1962–1981, conventional chemotherapy; Euro 1,303 patients, 1985–1996, interferon-alfa.
- Prognostic models and front-line treatment options for chronic-phase CML. The ASCO Post, 25 May 2015 — the Sokal, Euro and EUTOS formulas in full.
- Comparison of Sokal, Hasford and EUTOS risk scores in 114 patients with chronic-phase CML. Cureus, 2020 — all three formulas, and pairwise concordance of 53%, 64% and 98%.
Not medical advice. For healthcare professionals and education. Reference intervals vary by laboratory and assay — always use your own laboratory's. Never base a dose or a treatment decision on this page alone. Full disclaimer at calcengines.com/disclaimer/