ECOG Performance Status

ECOG Performance Status in Cancer Trials and Economic Evaluation

Oncologists grade how far cancer limits a patient's ordinary activity on the Eastern Cooperative Oncology Group scale, and the grade then travels through almost every stage of a cancer appraisal. Trial protocols use it to decide who may enrol, regression models use it to adjust survival, utility studies report EQ-5D scores by grade, and NICE committees use it to decide which patients a recommendation covers. In TA638 and TA818 NICE recommended the drugs only for people with an ECOG score of 0 or 1, because the trial evidence came from people in those grades. Abi Jaoude and colleagues reported in 2020 that 96.3% of patients enrolled in 600 phase III cancer trials had a score of 0 or 1, so the people who generate the evidence are on average fitter than many of those treated in the NHS.

Six grades and how a clinician assigns them

The scale was published by Oken and colleagues in 1982 in a paper on the toxicity and response criteria of the Eastern Cooperative Oncology Group, which set out standard measures for cancer trials. The ECOG-ACRIN Cancer Research Group reproduces the scale, credits it to that paper and states that it is in the public domain and free to use.

Each grade is a short description of activity, self-care and time spent in bed or a chair, and the treating clinician chooses the one that fits best, so the score is a clinician-reported outcome.

The scale therefore has six grades, of which only grades 0 to 4 describe living patients, so calling it a five-point scale is accurate only when death is left out, and a model that tracks death separately should not count grade 5 as a functional state. The grades are ordered categories, an instance of ordinal scaling, and economic analyses report the share of a cohort in each grade, or in groups such as 0 to 1 and 2 or more, instead of a mean grade.

Performance status as a gate to trials and its effect on generalisability

Most randomised cancer trials restrict entry by performance status. Abi Jaoude and colleagues reviewed 600 phase III trials and found that 527 (87.8%) specified a performance status cutoff and 237 (39.5%) admitted only patients with an ECOG score of 1 or less. Among all enrolled patients 96.3% had a score of 0 or 1, and even in trials that allowed a score of 2 or more, only 8.1% of those enrolled had a poor performance status. The authors concluded that the use of performance status restrictions in oncologic randomised trials is pervasive and that exceedingly few patients with poor performance status are enrolled.

For a health economist the restriction is a question of population. A cost-utility analysis built on such a trial estimates costs and QALYs for fit patients, and the decision may concern a wider group. An applicability assessment asks whether the trial population matches the people who would receive the treatment in practice, and performance status is one of the first characteristics checked.

How NICE committees have handled ECOG restrictions

In TA638, on atezolizumab with carboplatin and etoposide for untreated extensive-stage small-cell lung cancer, NICE recommended the combination only for adults with an ECOG performance status of 0 or 1, explaining that the clinical trial evidence was in people with that status. In its final appraisal document the committee added that healthcare professionals should take into account any physical, sensory or learning disabilities, or communication difficulties, that could affect ECOG performance status and make the adjustments they consider appropriate.

TA818, on nivolumab with ipilimumab for untreated unresectable malignant pleural mesothelioma, followed the same pattern. The recommendation covers adults with an ECOG performance status of 0 or 1, and the final appraisal document judged the trial population generalisable to people in UK clinical practice with those scores.

ECOG grade inside cost-effectiveness models

Performance status enters a model as a description of the baseline cohort, as a covariate in survival or progression equations, or as the basis for a utility value. In a partitioned survival model the curves for progression-free survival and overall survival come from a trial whose patients were mostly in grades 0 and 1, so the modelled survival already reflects that mix.

Utilities are the most direct link. Pickard, Neary and Cella analysed cross-sectional data from 534 cancer patients and used ECOG performance status as one anchor for the minimal clinically important difference in EQ-5D index scores, an anchor-based MID. With UK utility scores the estimates based on performance status lay between 0.10 and 0.12, both for all cancers and for the lung cancer subgroup, and with US utility scores between 0.07 and 0.09. The same respondents gave larger estimates under the UK set, so a utility gap attached to a change of ECOG grade belongs to a particular value set as well as to a population.

Worked example of an ECOG mix changing a QALY gain (illustrative)

All numbers in this example are illustrative. A new cancer drug extends progression-free survival by 0.5 years compared with standard care, at an incremental cost of GBP 15,100 per patient, with no discounting. The analyst holds an EQ-5D health state utility value for each ECOG grade in the progression-free state, 0.85 for grade 0, 0.75 for grade 1 and 0.60 for grade 2.

$$\bar{u} = \sum_{g} p_g u_g$$

where $\bar{u}$ is the mean utility of the cohort in the progression-free state, $p_g$ is the proportion of the cohort in ECOG grade $g$ and $u_g$ is the utility for that grade.

1. Trial cohort. Half the trial patients are in grade 0 and half in grade 1, so the mean utility is 0.5 × 0.85 + 0.5 × 0.75 = 0.80.

2. NHS cohort. Clinical advice suggests 35% in grade 0, 45% in grade 1 and 20% in grade 2. The products are 0.35 × 0.85 = 0.2975, 0.45 × 0.75 = 0.3375 and 0.20 × 0.60 = 0.12, which sum to 0.755.

3. QALY gains. The example treats the extra progression-free time as extra life, so the gain is the extra time multiplied by the mean utility. For the trial cohort the gain is 0.5 × 0.80 = 0.40 QALYs, and for the NHS cohort it is 0.5 × 0.755 = 0.3775 QALYs. If the extra progression-free time instead replaced time in progressed disease with no change in survival, each gain would be 0.5 multiplied by the difference between the progression-free and progressed-disease utilities.

4. Cost per QALY. The ICER is 15,100 ÷ 0.40 = 37,750 for the trial cohort and 15,100 ÷ 0.3775 = 40,000 for the NHS cohort, in GBP per QALY gained.

The fitter trial population makes the drug look about GBP 2,250 per QALY better value than it would be for the population treated in practice, even with the same survival gain in every grade. The example assumes the drug extends progression-free survival by 0.5 years in grade 2 patients too. In TA638 and TA818 NICE did not extend its recommendations to that group, because the trial evidence came from people with an ECOG score of 0 or 1. If grade 2 patients gain less, or suffer more toxicity, the ICER for the NHS cohort rises further, and a recommendation limited to grades 0 and 1 keeps the decision within the evidence.

How ECOG differs from nearby measures

The ECOG scale is one of several ways of recording how a person functions, and the differences decide where each fits in an analysis. The table compares ECOG with the measures most often confused with it.

The Karnofsky scale covers the same ground in finer steps, and Verger, Salamero and Conill warned of substantial error when an individual patient's score on one scale is predicted from the other, so trial evidence recorded on the two scales cannot simply be pooled. A performance outcome such as a timed walk asks the patient to do something measurable, while ECOG records a clinician's judgement of everyday life. ECOG grade can also be recorded at follow-up as an outcome measure, but like the baseline grade it needs utilities from a preference-based instrument before it can feed QALYs.

Limitations and common misreadings

Performance status is quick to record, and the single grade is easy to over-read. Several errors recur in economic work.

• Rater disagreement. Conill, Verger and Salamero reported a correlation of 0.75 between physicians rating the ECOG grade of the same 100 cancer patients, and Schnadig and colleagues found that patients and physicians disagreed on the grade for 56.6% of 1,636 patients with advanced lung or colorectal cancer. A one-grade shift at the boundary between 1 and 2 can move a patient across an eligibility rule or a utility group.
• Treating grades as numbers. Averaging grades or modelling them as a continuous score assumes equal steps that the scale does not claim, a source of measurement error in regression work.
• Assuming utilities transfer. A utility gap between grades measured in one population, instrument and value set does not carry over unchanged to another, as the UK and US differences in the Pickard study show.

Sources

• Abi Jaoude J, Kouzy R, Mainwaring W, et al. Performance status restriction in phase III cancer clinical trials. Journal of the National Comprehensive Cancer Network. 2020;18(10):1322-1326. doi:10.6004/jnccn.2020.7578
• Conill C, Verger E, Salamero M. Performance status assessment in cancer patients. Cancer. 1990;65(8):1864-1866. doi:10.1002/1097-0142(19900415)65:8<1864::AID-CNCR2820650832>3.0.CO;2-U
• ECOG-ACRIN Cancer Research Group. ECOG Performance Status Scale. https://ecog-acrin.org/resources/ecog-performance-status/
• National Institute for Health and Care Excellence. Atezolizumab with carboplatin and etoposide for untreated extensive-stage small-cell lung cancer (TA638). Published 1 July 2020. https://www.nice.org.uk/guidance/ta638
• National Institute for Health and Care Excellence. Nivolumab with ipilimumab for untreated unresectable malignant pleural mesothelioma (TA818). Published 17 August 2022. https://www.nice.org.uk/guidance/ta818
• Oken MM, Creech RH, Tormey DC, et al. Toxicity and response criteria of the Eastern Cooperative Oncology Group. American Journal of Clinical Oncology. 1982;5(6):649-655.
• Pickard AS, Neary MP, Cella D. Estimation of minimally important differences in EQ-5D utility and VAS scores in cancer. Health and Quality of Life Outcomes. 2007;5:70. doi:10.1186/1477-7525-5-70
• Schnadig ID, Fromme EK, Loprinzi CL, et al. Patient-physician disagreement regarding performance status is associated with worse survivorship in patients with advanced cancer. Cancer. 2008;113(8):2205-2214. doi:10.1002/cncr.23856
• Verger E, Salamero M, Conill C. Can Karnofsky performance status be transformed to the Eastern Cooperative Oncology Group scoring scale and vice versa? European Journal of Cancer. 1992;28A(8-9):1328-1330. doi:10.1016/0959-8049(92)90510-9