Table 1.
Baseline demographic, clinical and SF-36 characteristics of the total analysis population and each statistical analysis population.
Fig 1.
Change in SF-36 mental component score (MCS) and physical component score (PCS) over the two-year follow-up.
(A) Number of individuals providing evaluable SF-36 data in each visit window. (B) Population-level change in median (Inter-Quartile Range) MCS (blue) and PCS (red). (C) Individual-level change in PCS. (D) Individual-level change in MCS. Points in (C) and (D) coloured by visit window assigned in data processing: Red = Baseline, Yellow = 3 months, Green = 6 months, Turquoise = 12 months, Blue = 18 months, Pink = 24 months.
Table 2.
Paired Wilcoxon Rank Sum Test change in median mental and physical component scores.
Table 3.
Weighted generalised estimating equation regression output for the Mental Component Score model with time modelled using a fractional polynomial.
Table 4.
Weighted Generalised Estimating Equation regression output for the Physical Component Score model with time modelled with a three-degree polynomial.
Fig 2.
Adjusted mental and physical component scores, estimated using the best-fitting non-linear weighted generalised estimating equation models.
(A) Mental component score (MCS). (B) Physical component score (PCS). MCS is modelled with time with a fractional polynomial [((time + 0.1)/10)-2], adjusting for sex, age and presentation with advanced HIV. PCS is modelled with a 3-degree polynomial [time3], adjusting for sex, age, physical comorbidities, presentation with advanced HIV and baseline HIV RNA count.
Table 5.
Summary of the strengths and limitations of the four statistical approaches.