ACR20, ACR50 and ACR70 Response Calculator
ACR20, ACR50 and ACR70 Response Calculator
A conjunction, not a score: both joint counts must improve by the stated percentage AND three of the remaining five core-set measures. Improvement in the joints alone is not an ACR response.
ACR response
9 inputs → highest level metTender joint count 10 to 4, swollen joint count 8 to 3, and improvements of 60%, 55%, 52%, 30% and 10% in patient pain, patient global, physician global, the disability index and the acute-phase reactant
A conjunction over seven measures, at one of three levels
AND ≥ 20% improvement in the SWOLLEN joint count
AND ≥ 20% improvement in at least THREE of the following five:
patient pain · patient global assessment · physician global assessment · patient-assessed physical function (a disability index) · an acute-phase reactant (ESR or CRP)
ACR50 and ACR70 are the same construction with 20% replaced by 50% and 70%
- there is no total
- the ACR response is not a score and produces no number. Seven measures improving is the same verdict as five. Anything reporting an ‘ACR score’ has invented it
- the two joint counts are a GATE
- not two of seven votes. Ninety per cent improvement in all five remaining measures with one joint count improving by 19% is no ACR response at all. This is the single most misread feature of the definition
- three of five, not a majority of seven
- exactly three of the remaining five suffices; two does not. Which three is irrelevant
- the disability index
- in practice the HAQ disability index, which is copyright Stanford University per LOINC’s own record, with terms of use ‘used with permission’. THIS PAGE CANNOT SUPPLY IT and takes the percentage from the reader. This page reproduces a formula and its published cut-offs, not a questionnaire. Felson’s own derivation hit the same wall and substituted grip strength, because four of its five source trials had no functional measure
- the acute-phase reactant
- ESR or CRP — the core set names either and does not choose. They move on different timescales and respond to different things, so the choice can decide whether this measure counts among the three
- a baseline of zero
- makes the definition inapplicable, not negative. Percentage improvement from zero does not exist, which is why trials set a minimum joint count at entry
Worked example
Tender joint count 10 to 4, swollen joint count 8 to 3, and improvements of 60%, 55%, 52%, 30% and 10% in patient pain, patient global, physician global, the disability index and the acute-phase reactant
Tender joints: (10 − 4) / 10 = 60% improvement
Swollen joints: (8 − 3) / 8 = 62.5% improvement
At the 70% level both joint counts fail, so ACR70 is not met whatever the other five do
At the 50% level both joint counts pass. Of the remaining five, pain 60%, patient global 55% and physician global 52% each reach 50% — that is three of five, which is exactly enough
The disability index at 30% and the acute-phase reactant at 10% do not reach 50%, and it does not matter: three is the requirement, not four
So this patient is an ACR50 responder, and therefore also ACR20
Set all five remaining measures to 0% and the same joint counts give no ACR response at all — improvement in the joints alone is never an ACR response
Set all five to 90% and change the tender count to 10 → 9 (10% improvement): still no ACR response, because one failing joint count closes the gate
The seven core-set measures, and what each one is for
| Measure | Role in the definition | Can this site supply it? |
|---|---|---|
| Tender joint count | Gateway — must improve | Counted at the bedside |
| Swollen joint count | Gateway — must improve | Counted at the bedside |
| Patient pain assessment | One of five; three needed | A visual analogue scale |
| Patient global assessment | One of five; three needed | A visual analogue scale |
| Physician global assessment | One of five; three needed | A visual analogue scale |
| Patient-assessed physical function | One of five; three needed | No — the HAQ-DI is copyright Stanford University |
| Acute-phase reactant (ESR or CRP) | One of five; three needed | A laboratory result |
Why improvement in the joint counts alone is not a response
| Tender | Swollen | Of the five remaining | Verdict at the 20% level |
|---|---|---|---|
| 50% better | 50% better | none reaches 20% | No ACR response |
| 50% better | 50% better | two reach 20% | No ACR response |
| 50% better | 50% better | three reach 20% | ACR20 |
| 10% better | 50% better | all five reach 90% | No ACR response |
| exactly 20% better | exactly 20% better | exactly three reach exactly 20% | ACR20 |
A composite definition, a trial endpoint, and the questionnaire it needs that this site will not reproduce
The ACR response is not a score and it never was. Felson and colleagues set out in 1995 to find a single primary endpoint for rheumatoid arthritis trials, tested 40 candidate definitions, and published one: at least 20% improvement in the tender joint count and in the swollen joint count, and at least 20% improvement in at least three of the five remaining core-set measures. Nothing is added up. A patient meeting six of seven criteria has the same verdict as one meeting five, and a patient meeting all five of the non-joint measures while one joint count improves by 19% has no ACR response at all. The two joint counts are a gate rather than two of seven votes, and that is the feature the definition is most often misread on.
The derivation is worth knowing because it explains the modest threshold. The candidates were ranked first against the impressions of 68 responding rheumatologists rating 43 paper patients, then tested in five placebo-controlled trials of second-line agents — 320 of 508 enrolled patients had complete data — and then in a methotrexate against auranofin comparative trial with 274 complete cases. The selected definition discriminated active drug from placebo while identifying few placebo patients as improved, and cut the sample size needed for 80% power from at least 80 patients per group to between 20 and 32. Twenty per cent was chosen for statistical efficiency in a trial, not because it describes a good outcome. ACR50 and ACR70 came afterwards as the same construction with the percentage raised; the 1995 paper defined one level.
On what this definition is for, the honest answer is that the source says more than the folklore does. The paper describes both uses: ‘It would be a single primary end point for analysis’ and ‘The definition of improvement can characterize the response of individual patients to therapy’. What is nonetheless true is that trials report the proportion of patients achieving ACR20, that treat-to-target recommendations set targets as states — remission, or low disease activity — rather than as percentage changes, and that a patient entering with a low joint count cannot generate an ACR response at all. For an individual patient the state instruments are the better instruments: CDAI, SDAI, DAS28-ESR and Boolean and index-based remission criteria. EULAR DAS28 response criteria is the European answer to the same question and takes a different shape — it needs both the change and the current value.
One of the seven measures cannot be supplied here. The core-set’s physical-function measure is in practice the HAQ disability index, and LOINC — an independent third party — records the Health Assessment Questionnaire as copyright Stanford University with terms of use ‘used with permission’. So neither the questionnaire nor a paraphrase of it appears anywhere on this site, and this page takes the percentage improvement in the disability index from the reader, who has the scored instrument. This page reproduces a formula and its published cut-offs, not a questionnaire. Felson’s own derivation ran into the same gap from the other side: four of its five source trials had no functional measure, so grip strength was substituted for functional status throughout the validation. A composite index is not a diagnosis, and a cut-off chosen in one cohort is not a probability for one patient. This states what the instrument measured in a named cohort; the clinician in front of the patient decides what follows.
Frequently asked questions
What is the ACR20 definition?
At least 20% improvement in the tender joint count AND at least 20% improvement in the swollen joint count AND at least 20% improvement in at least three of the five remaining core-set measures: patient pain, patient global assessment, physician global assessment, patient-assessed physical function, and an acute-phase reactant (ESR or CRP). ACR50 and ACR70 are the same construction at 50% and 70%.
Is improvement in the joint counts enough for an ACR20?
No. The two joint counts are a gate and not two of seven votes: both must improve by the stated percentage, and separately at least three of the remaining five must. A patient whose joint counts halve while nothing else changes has no ACR response, and a patient who improves by 90% on all five remaining measures with one joint count improving by 19% also has none.
Is the ACR20 a score?
No, and nothing reporting an ‘ACR score’ is reporting the published definition. It produces a yes or a no at each of three levels and no total. Six measures improving is the same verdict as five.
Should I use ACR20 to judge my own patient’s response?
It is the endpoint trials report, as the proportion of patients achieving it, and treat-to-target recommendations set their targets as states — remission or low disease activity — rather than as percentage changes. The source paper does say the definition can characterise an individual patient’s response, so this page states both positions rather than choosing. For an individual, the state instruments are easier to act on.
Why does the page ask me for the disability index percentage instead of calculating it?
Because the core set’s physical-function measure is in practice the HAQ disability index, which LOINC records as copyright Stanford University with terms of use ‘used with permission’. Neither it nor a paraphrase appears on this site, so the percentage comes from your own scored instrument.
What happens if the baseline joint count was zero?
The definition does not apply, and this page says so rather than returning a verdict. Percentage improvement from a baseline of zero is undefined. It is one reason trials set a minimum joint count at entry, and one reason a percentage-change endpoint is a poor instrument in a patient already close to remission.
Related calculators
References
- Felson DT, Anderson JJ, Boers M, et al. American College of Rheumatology preliminary definition of improvement in rheumatoid arthritis. Arthritis Rheum. 1995;38(6):727–35.
- Regenstrief Institute. LOINC 75874-8, Health Assessment Questionnaire. Copyright notice recorded as “Copyright © Stanford University”, terms of use “Used with permission”. (Read for the licensing position on the HAQ-DI, which this page therefore does not supply.)
- MSB11456 in participants with moderately to severely active rheumatoid arthritis — statistical analysis plan, NCT04512001. ClinicalTrials.gov document archive. (States the SDAI and CDAI definitions, the CRP unit as mg/dL, both full cut-off sets and the ACR response definition.)
- A study of LY3462817 in participants with rheumatoid arthritis — protocol, NCT04634253. ClinicalTrials.gov document archive. (Independently corroborates SDAI ≤ 3.3 and ≤ 11, CDAI ≤ 2.8, and states the SDAI CRP input range as 0.1 to 10.0 mg/dL.)
- Felson DT, Smolen JS, Wells G, et al. American College of Rheumatology/European League Against Rheumatism provisional definition of remission in rheumatoid arthritis for clinical trials. Ann Rheum Dis. 2011;70(3):404–13.
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/