Table 6.

Baseline blood analyses as predictors of follow-up HAQ score

Study details			Predictor: Baseline ESR/CRP		Predictor: Baseline RF/ACPA
Study	N	Analysis method	Associated with HAQ	Effect sizea	Associated with HAQ	Effect sizea	Adjusted for
Multivariable analyses
Humphreys [40]	1995	Generalised estimating equations analysis	–	–	✗ (RF)	RF + vs. RF–: b –0.03 (–0.12, 0.05)	Age, gender, smoking status, polynomials of disease duration, year of recruitment
					✓ (ACPA)	ACPA+ vs. ACPA–: b 0.12 (0.02, 0.21)
Malm [32]	1387	Logistic regression (HAQ cut-off = 0.75)	✓ (ESR)	Per unit baseline ESR:	–	–	Age, gender, disease duration
				OR 1.01 (1.00, 1.01)
Nair [30]	1034	Linear mixed model	–	–	✗ (RF)	RF+ vs. RF–:	Age, gender, treatment, Sharp Score (van der Heijde modification), HAQ (t-1), DAS28, BMI
						Cohort 1 b 0.00, p = 0.99
						Cohort 2 b 0.00, p = 0.90
Burr [50]	640	Logistic regression (HAQ cut-off = 1)	–	–	✗ (ACPA)	Per unit baseline ACPA titre:	Baseline: age, gender, symptom duration, CRP, RF, HAQ, swollen joint count, tender joint count
						OR 1.00 (0.99, 1.01)
Kroot [38]	273	Multivariable linear regression	–	–	✓ (RF)	RF+ vs. RF–:	Age, gender, DAS28, HLA-DR4 gene,
					✗ (ACPA)	b 0.15, p < 0.05
						ACPA+ vs. ACPA–:
						b 0.00, p = NS
Combe [34]	191	Multivariable linear regression	✓ (CRP)✓ (ESR)	Per unit baseline CRP:	–	–	Baseline: DAS, swollen joint count, tender joint count, HAQ, pain VAS
				b 0.01 (p < 0.01)
				Per unit baseline ESR:
				b 0.01 (p < 0.01)
Bjork [49]	189	Projections to latent structure discriminant analysis (HAQ cut-off = 0.08)	✗ (CRP)	Per unit baseline CRP:	–	–	Baseline: Age, gender, HAQ, grip force, SOFI-lower limb, walking speed, GAT, wellbeing, swollen joint count, SOFI-hand, tender joint count, PGA, pain, SOFI-upper limb
			✗ (ESR)	VIP 0.63 (“not important”)
				Per unit baseline ESR:
				VIP 0.49 (“not important”)
Welsing [31]	185	General linear mixed model	–	–	✓ (RF)	RF+ vs. RF–:	Baseline: age, sex; time-varying: Sharp score, squared Sharp score, DAS28
						b 0.19 (0.03, 0.35)
Lindqvist [54]	183	Stepwise logistic regression (HAQ cut-off = 1.0)	✗ (ESR)	NS – coefficients and confidence intervals not reported	✗ (RF)	NS – coefficients and confidence intervals not reported	Age, gender, genotype, HAQ, active joint count
Verstappen [36]	112	Logistic regression (HAQ cut-off = 1)	✗ (ESR)	Per unit baseline ESR:	–	–	Disease duration (natural log transformed)
				OR 1.00 (0.99, 1.02)
Kuuliala [48]	85	Logistic regression (HAQ cut-off = 0.9)	–	–	✗ (RF)	RF+ vs. RF–:	Age, gender, shared epitope, tertiles of soluble E-selectin
						OR 1.09 (0.33, 3.57)
Benton [27]	42	Logistic regression (HAQ cut-off = 0.25) OR are the odds of being in the low HAQ group	✗ (CRP)	Per unit baseline CRP:	–	–	Baseline: DAS, Ritchie index, HAQ, Sharp score; one year: DAS, Ritchie index, CRP, HAQ
				OR 0.99 (0.96, 1.03)
Univariable analyses
Koevoets [39]	508	Generalised estimating equations analysis	–	–	✗ (RF)	RF+ vs. RF–:	–
					✗ (ACPA)	b –0.03 (–0.13, 0.08)
						ACPA+ vs. ACPA–:
						b –0.03 (–0.13, 0.07)
Thyberg [55]	251	Chi-Square (HAQ cut-off=1)	–	–	✗ (ACPA)	NS difference between proportion of ACPA+ patients between HAQ subgroups	–
Jäntti [33]	121	Somers’ d	✗ (ESR)	d = 0.12 (–0.04, 0.28)	✗ (RF)	d = –0.18 (–0.62, 0.26)	–
Contreras-Yanez [25]	107	Student’s t test Comparison groups: HAQ ≤ 0.2 at 5 years, yes/no	–	–	✗ (RF)	Proportion baseline RF+:	–
					✗ (ACPA)	HAQ ≤ 0.2 82.1%
						HAQ > 0.2 82.6%
						p = 1.00
						Proportion baseline ACPA+:
						HAQ ≤ 0.2 85.7%
						HAQ > 0.2 87.0%
						p=1.00
Woolf [58]	88	Calculated sensitivity and specificity of having HAQ > 0 at 5 years	✓ (“raised CRP”)	Specificity/sensitivity:	✓ (RF)	Specificity/sensitivity:	–
				93/74		64/37
Genevay [23]	25	Mann-Whitney	–	–	✗ (RF)	Mean HAQ:	–
						RF+ 0.78
						RF– 0.80
						p = NS

See Table 2 for acronym definitions: ACPA, BMI, CRP, DAS, ESR, HAQ, N, RF, VAS.

See Table 3 for acronym definitions: b,GAT,NS,OR,PGA, SOFI, VIP.

^aBrackets indicate 95% confidence interval unless otherwise stated.