Apr 16

Diagnostic utility of rheumatoid factor in low risk population

Journal reference: Nishimura K, Sugiyama D, Kogata Y, Tsuji G, Nakazawa T, Kawano S, et al. Meta-analysis: diagnostic accuracy of anti-cyclic citrullinated peptide antibody and rheumatoid factor for rheumatoid arthritis. Ann Intern Med 2007; 146(11): 797-808; and others.

Link: http://dx.doi.org/10.7326/0003-4819-146-11-200706050-00008

Published: 5 June 2007

Evidence cookie says…

There is no role for diagnostic rheumatoid factor (RF) testing in someone without inflammatory arthritis.

An elevated RF increases the odds of rheumatoid arthritis about 5-fold.  The post-test risk, however, depends on the pre-test risk.


  • Dr Michael Tam, GP and Conjoint Senior Lecturer at UNSW Australia and GP Unit Fairfield Hospital
  • Dr Joel Rhee, GP and Senior Lecturer in General Practice at UNSW Australia

A draft of this article was originally written for Australian Doctor.

Clinical case

James is a fit 70-year-old man with what I had assumed to be hand osteoarthritis.  Recently, he saw our GP registrar who ordered a rheumatoid factor (RF) along with other pathology tests.  This came back elevated – 50 IU/mL.  As the registrar informed me of the pathology result, we both shared a “d’oh!” moment – the registrar had expected the result to be normal, and I wasn’t sure how to interpret the finding.

On his next visit, I took the opportunity to conduct a clinical assessment.  James had but a few moments of stiffness in the morning.    His hands were not tender, swollen, “boggy”, or inflamed – there was no clinical evidence of synovitis.  It appeared to be nodal osteoarthritis – there were bilateral bony lumps around his DIP and PIP joints (Heberden’s and Bouchard’s nodes).  James was able to do most things, including playing a weekly round of golf.

Clinical question

As a result of the elevated RF, what is the likelihood that James has rheumatoid arthritis?

What is the evidence?

To answer this question, we need to know two things:

  1. The baseline likelihood, or “pre-test probability” that James has RA
  2. The test characteristics of elevated RF for RA diagnosis (sensitivity and specificity)

Pre-test probability of RA

Ideally, we want to know the prevalence of undiagnosed RA in Australian men aged 70 years.  I wasn’t able to find this specific data (it probably doesn’t exist), but we can make some inferences.  In an American population, the residual lifetime risk of developing RA for men at age 70 was 0.73% (about 1 in 140). [1]  It is improbable that the prevalence of undiagnosed RA is higher than the residual lifetime risk.  As such, this value can be considered the upper boundary of undiagnosed RA prevalence in men this age.

We can estimate James’ pre-test probability of RA by taking his clinical presentation, along with the population prevalence into consideration.  Likelihood ratios (LRs) are the odds that a clinical feature (e.g., a symptom or a sign, or a test result) would be expected in someone with the condition of interest, as compared to someone who does not have the condition.  Values for LRs can range from zero to infinity:

  • LR < 1 – the feature  reduces the likelihood of the condition
  • LR = 1 – the feature  has no effect on the likelihood of the condition
  • LR > 1 – the features increases the likelihood of the condition

The LRs of James’ signs and symptoms (or lack thereof) for RA: [2]

  • No morning stiffness (LR = 0.5)
  • No active swelling in three or more joint areas (LR = 0.5)
  • No active symptoms in wrist, MCP or PIP joints (LR = 0.4)
  • Presence of symmetrical arthritis (LR = 1.2)
  • No rheumatoid nodules (LR = 0.98)
  • All together for John: LR ≈ 0.12 (0.5 × 0.5 × 0.4 × 1.2 × 0.98)

John’s clinical picture reduces his odds of RA around 8-fold.  Factoring this into the previous value of 0.73%, his pre-test probability of RA is less than 0.1% (< 1 in 1000).

Characteristics of rheumatoid factor

The sensitivity and specificity of elevated RF for RA diagnosis varies between studies, but reasonable estimates are: [3]

  • sensitivity:          69%
  • specificity:           85%
  • positive LR:         4.9

The elevated RF increases the odds of RA about 5-fold, which raises James’ post-test probability of RA to an uninspiring 0.5%.


“Positive” test results must be interpreted in the clinical context.  For James, the probability that he had RA was no higher than 1 in 200 despite the elevated RF.  The important determinant was his tiny pre-test risk.  James’ negative clinical features had a larger impact on the probability of RA than the RF result!  This demonstrates the minimal utility of RF testing in someone without inflammatory arthritis and highlights the importance of history and examination in the diagnostic assessment.

Although the maths is unlikely to be routinely used by clinicians, the underlying idea can be conceptualised in an intuitive qualitative manner. [4]  Post-test probability depends on pre-test probability.  Most test results change the probability of diagnosis by only one qualitative category (very unlikely ↔ unlikely ↔ uncertain ↔ likely ↔ very likely).

In this case, the pre-test probability for RA was “very unlikely”.  An elevated RF might change this probability to “unlikely” – which isn’t clinically meaningful.


  1. Crowson CS, Matteson EL, Myasoedova E, Michet CJ, Ernste FC, Warrington KJ, et al. The lifetime risk of adult-onset rheumatoid arthritis and other inflammatory autoimmune rheumatic diseases. Arthritis Rheum 2011; 63(3): 633-639. DOI: 10.1002/art.30155
  2. Rindfleisch JA, Muller D. Diagnosis and management of rheumatoid arthritis. Am Fam Physician 2005; 72(6): 1037-1047
  3. Nishimura K, Sugiyama D, Kogata Y, Tsuji G, Nakazawa T, Kawano S, et al. Meta-analysis: diagnostic accuracy of anti-cyclic citrullinated peptide antibody and rheumatoid factor for rheumatoid arthritis. Ann Intern Med 2007; 146(11): 797-808
  4. Medow MA, Lucey CR. A qualitative approach to Bayes’ theorem. Evid Based Med 2011; 16(6): 163-167. DOI: 10.1136/ebm-2011-0007
Print Friendly, PDF & Email

Leave a Reply

%d bloggers like this: