Skip to main content

Table 1 Potential candidates to be involved as prognostic factors in the prognostic model

From: Predicting the treatment response of certolizumab for individual adult patients with rheumatoid arthritis: protocol for an individual participant data meta-analysis

 Age*, sex*, ethnicities
 BMI*, smoking history*
Clinical features
 Family history of first-degree relatives
 Length of time since first onset until the trial commencement, length of time since first onset until first treatment*
 Disease activity:
• Number of tender joints*, number of swelling joints*, self-report level of pain based on visual analogue scale (VAS)*
• Disease Activity Score (continuous)*, disease activity level (categorical)*
 Joint involvement:
• Large joint involvement: knee, hip joints, etc.
• Uncommon joint involvement
 Nonspecific systemic symptoms: fever, fatigue, etc.
 Comorbidities*: osteoporosis, osteoarthritis, etc.
 Functional/global quality of life (QoL) conditions at baseline*
 Prior treatment history: failure times, failed drug types, etc.
 Cointerventions decided before randomization:
• Steroids, nonsteroidal anti-inflammatory drug (NSAIDs)
Biochemical features
 Serum inflammatory factors*:
• Erythrocyte sedimentation rate (ESR), C-reactive protein (CRP); others such as TNF, IL-6, etc.
 Serum antibodies*:
• Rheumatoid factor (RF), anti-cyclic citrullinated peptide (anti-CCP), antinuclear antibody (ANA) spectrum
Radiographic features
 Joint fusion (already deformed), bone erosion*, synovitis*, early bone inflammation
 Radiographic scores*
 HLA (human leukocyte antigen) types and SNPs (single nucleotide polymorphisms) if they are tested.
  1. *Factors that have been proved to be a prognostic factor for any treatments in previous studies
  2. #Since genetic tests for RA are not routinely implemented in clinical practice, we anticipate that most studies will not report them. Although genetics are often considered critical in precision medicine, we will consider it justifiable if no genetic information is included in our model, because there is no single one that has been proven to be strongly associated with the prognosis or treatment responses, and two studies have indicated that genetic information barely contribute in predicting treatment effects [33]