Verteporfin photodynamic therapy for neovascular age-related ...

70 Discussion of resultsOur analyses revealed that the decrease in generic HRQoL for a clinically significantdeterioration in vision, as measured by BCVA, was small. The predicted decreases for a fiveletterdrop in BCVA in the better-seeing eye were 0.0058, 0.245 and 0.546 for the SF-6D, PCSand MCS respectively (all p < 0.0001). We also had CS data from a subset of participants andtherefore were able to estimate changes in HRQoL for both BCVA and CS and relate these tothe findings observed in the TAP trial. However, when interpreting these associations, it is veryimportant to recognise that the predicted HRQoL changes do not describe the average benefit ina representative sample of patients, given that a proportion (47% of the VPDT cohort) will havehad treatment to the first affected eye and had a normally sighted fellow eye.The gradient of change in visual functioning measured by the NEIVFQ instrument was notinfluenced by the duration of follow-up. Models predicting distance, near and compositeNEIVFQ scores from BCVA were quadratic. The predicted decreases for a five-letter drop inBCVA in the better-seeing eye were 5.08, 5.48 and 3.90 for the distance and near domains andfor the composite instrument respectively. The Submacular Surgery Trials Research Groupquantified the association between BCVA and NEIVFQ scores using the change in BCVA. 65 Theobserved change in NEIVFQ composite score per 100 letters was 18 points, very much less thanour estimate of 55 letters. Their lower estimate may be due, in part, to the assumption of a linearrelationship or to their much smaller sample size. The VPDT cohort study has enabled moreprecise quantification of these relationships, particularly those between BCVA and HRQoL.These relationships constitute an important source of robust information for modelling the costeffectivenessof current and future interventions for nAMD.The shape of the relationship between visual function and HRQoL had also not been previouslyinvestigated. Even in a study as large at the VPDT cohort study, the analyses had limited powerto detect departures from a linear relationship. The tendency for some functions to plateau withsevere loss of visual function is clearly not a floor effect and supports the prior hypothesis thatHRQoL decreases less steeply when visual function is very poor. Our failure to observe a plateauwhen vision is excellent may have arisen because few patients achieved BCVA < 0.0 logMAR(6/6 Snellen) in the best-seeing eye, either because patients could not achieve better acuity orbecause they were not encouraged to do so. However, this explanation is not consistent withthe observation that visually demanding tasks such as fluent reading and driving ability can beperformed with BCVA = 0.3 logMAR (6/12 Snellen) in the best-seeing eye.Previous researchers have investigated which of BCVA or CS in the better-seeing eye is thestronger predictor of HRQoL. Based on a multivariable regression, Bansback et al. 66 reported thatCS was a better predictor of HRQoL than BCVA and, using the Health Utilities Index mark 3 67(HUI3), estimated a change in utility of 0.14 per log unit. In contrast, we found that, in this largedata set, BCVA was a consistently stronger predictor of HRQoL than CS.Patients’ adaptation to loss of visual function was investigated by Brown. 31 In a sample of 237patients with mixed causes of visual loss there was a tendency for those who had had visual lossfor longer (over a time frame of 5 years) to report better HRQoL. 31 In a sample of 72 patientswho had had AMD for up to 20 years (35 for < 1 year, 19 for 1–3 years and 18 for 3–20 years),patients with durations of visual loss ≥ 1 year had better HRQoL than those with durations< 1 year (p < 0.005). 18 This association was potentially affected by the small number of patientswith very longstanding disease. Also, these studies did not assess HRQoL longitudinally in thesame patients, so the observed finding is less confidently attributed to the duration of disease andit is not clear whether or not the differences in HRQoL with duration of visual loss were adjusted