Table 3.

Baseline demographic predictors of follow-up HAQ score

Study details			Predictor: Age		Predictor: Gender
Authors	N	Analysis method	Associated with HAQ	Effect sizea	Associated with HAQ	Effect sizea	Adjusted for
Multivariable analyses
Camacho [24]	3666	Multivariable linear random effects model (<55 years used as the reference category)	✓	Men:55–74 b 0.19 (–0.01, 0.39)≥75 b 1.81 (1.25, 2.36)	✓	Women vs. men:	Age at final follow-up, year recruited to the study
						b 0.24 (0.20, 0.29)	*Further adjustment: baseline disease duration and DMARDs within 6 months of symptom onset.
						Further adjustment*: b 0.29 (0.25, 0.34)
				Women:
				55–74 b 0.26 (0.12, 0.40)
				≥75 b 0.51 (0.05, 0.98)
Malm [32]	1387	Logistic regression	✓	Age at onset (years):	✓	Women vs. men:	Disease duration
		(HAQ cut-off = 0.75)		OR 1.03 (1.02, 1.04)		OR 2.53 (1.85, 3.46)
Nair [30]	1034	Linear mixed model	✓	Age at onset (years):	✓	“sex”:	Treatment, SHS, HAQ (t-1), DAS28, BMI, RF
				Cohort 1 b 0.00, p < 0.01		Cohort 1 b 0.08, p < 0.01
				Cohort 2 b 0.00, p = 0.01		Cohort 2 b 0.19, p < 0.01
Combe [51]	813	Logistic regression (HAQ cut-off = 0.3)	✓	“older age”:OR 1.91 (1.32, 1.77)	✓	Women vs. men:OR 1.60 (1.02, 2.50)	Baseline: HAQ, pain
Wiles [37]	684	Generalised estimating equations analysis (HAQ cut-off = 1)	✓	<47 years at onset (ref cat):	✓	Women vs. men:	Duration from symptom onset to baseline, Time-varying: morning stiffness, RF, rheumatoid nodules, number of deformed joints
				47–63 OR 1.45 (1.06, 2.00)		OR 1.70 (1.29, 2.24)
				≥64 OR 3.21 (2.33, 4.42)
Ahlmen [4]	549	ANCOVA	–	–	✓	Mean (SD) HAQ:	Age
						Men 0.51 (0.56)
						Women 0.73 (0.68)p < 0.01
Wiles [57]	528	Logistic regression	✓	<47 years (ref cat):	–	–	Year 1 HAQ, Nodules, Knee involvement factor, Tenderness factor (factors created using principal component analysis)
		Model 1 HAQ cut-off = 1,		47–63:
		Model 2 HAQ cut-off = 1.5		Model 1 OR 2.06 (1.11, 3.83)
				Model 2 OR 1.62 (0.79, 3.32)
				≥ 64:
				Model 1 OR 3.46 (1.77, 6.76)
				Model 2 OR 2.70 (1.29, 5.67)
Welsing [56]	378	General linear mixed model	✓	Age at onset (years): b 0.01 (0.01, 0.20)	✓	Women vs. men: b 0.22 (0.08, 0.36)	Baseline: RF; time-varying: SHS, squared SHS, DAS28
Kroot [38]	273	Multiple regression	✓	Age at entry (years): b 0.01, p < 0.01	✓	Female gender: b –0.128, p < 0.05	RF, DAS28, HLA-DR4 gene, ACPA
Hallert [35]	251	Generalised estimating equations analysis	✗	NS – coefficients and confidence interval not reported	✗	NS – coefficients and confidence interval not reported	DMARD use, biologic use, grip force, SOFI-hand, SOFI-upper extremity, SOFI-lower extremity, GAT, pain, walking time
Bjork [49]	189	Projections to latent structure discriminant analysis (HAQ cut-off = 0.08)	✗	Baseline age: VIP 0.22 (“not important”)	✓	“sex”: VIP 1.39 (“important”)	Baseline: HAQ, grip force, SOFI-lower limb, gender, walking speed, GAT, wellbeing, CRP, SOFI-hand, ESR, tender joints, PGA, pain, SOFI-upper limb, swollen joints
Welsing [31]	185	Mixed model (HAQ was log transformed)	✓	Age at onset per year: b 0.02, p < 0.01	✓	Women vs. men: b 0.38 p = 0.02	DAS28, Modified SHS, Modified SHS squared, age*modified SHS
Lindqvist [54]	183	Stepwise logistic regression (HAQ cut-off = 1.0)	✗	NS – coefficients and confidence interval not reported	✗	NS – coefficients and confidence interval not reported	Genotype, RF, HAQ, ESR, active joint count
Kapetanovic [53]	183	Hierarchical linear regression	–	–	✗	HAQ at final follow-up:	CCI, DAS, joint damage
						“sex” b –0.095, p = NS
						HAQ over time (AUC):
						“sex” b –0.20, p < 0.01
Verstappen [36]	112	Logistic regression (HAQ cut-off = 1)	✓	Age at onset: OR 1.05 (1.01, 1.09)	✗	Women vs. men: OR 0.90 (0.37, 2.17)	Disease duration (natural log transformed)
Contreras-Yanez [25]	107	Multivariable linear regression	✓	Age at baseline (years): b 0.10, p = 0.001	✗	NS – coefficients and confidence intervals not reported	Variables tested in univariable analysis: age, gender, disease duration, DAS28, persistence of DMARDs, comorbidity
Kuuliala [48]	85	Logistic regression (HAQ cut-off = 0.9)	✗	Age at entry (years): OR 1.02 (0.97, 1.07)	✓	Women vs. men: OR 5.51 (1.81, 16.8)	RF, Shared epitope, tertiles of soluble E-selectin
Eberhardt [28]	63	Logistic regression (HAQ cut-off = 1)	–	–	✓	Women vs. men: OR 1.02 p < 0.01	“[demographic,] clinical, radiographic and laboratory data”
Univariable analyses
Koevoets [39]	508	Generalised estimating equations analysis	–	–	✓	Women vs. men: b 0.14 (0.05, 0.24)	–
Kuiper [45]	332	Student’s t test	✓	Older men had higher HAQ scores than younger men (p < 0.01)	✓	Women had higher HAQ scores than men (p < 0.05)	–
Combe [34]	191	Spearman’s test	✗	NS – coefficients and confidence interval not reported	✗	NS – coefficients and confidence interval not reported	–
Jäntti [33]	121	Somers’ d	✓	Age at entry:	✗	“sex”:	–
				Somers’ d 0.30 (0.16, 0.45)		Somers’ d 0.01(–0.31, 0.33)

See Table 2 for acronym definitions: ACPA, BMI, DAS28, ESR, HAQ, N, RF.

^aBrackets indicate 95% confidence interval unless otherwise stated; ANCOVA = analysis of covariance, AUC = area under the curve, b = regression coefficient, BL = baseline, CCI = Charlson Comorbidity Index, DMARDs = disease modifying anti-rheumatic drugs, FU = follow-up, GAT = Grip Ability Test, NS = non-significant, OR = odds ratio, PGA = patient global assessment, RA = rheumatoid arthritis, SD = standard deviation, SE = standard error, SHS = Sharp score, SOFI = Signals of Functional Impairment, VIP = variable influence on projection.