There were 141 persons who had baseline visits. Results of the baseline visit indicated that 31 of these persons did not meet eligibility criteria and so were excluded (three had acuity worse than 20/80 in the better eye, eight had advanced ARM, four did not pass the liver and/or hypercalcemia screening tests, and 16 could not reliably make psychophysical judgments). No persons were ineligible based on serum vitamin A deficiency. Six persons withdrew from the study after baseline and randomization (five because of family illness or crisis and one because of migraine). Thus, the final sample consisted of 104 participants who were randomized to the vitamin A (n = 52) and placebo groups (n = 52) and who completed both the baseline and 30-day follow-up visits. During the course of the study, six persons in the treatment group reported transitory side effects (one had hot flashes, four nausea, one headache, two eye pain, one lethargy, and one blurry vision). Two persons in the placebo group reported side effects (one headache and one increased blood pressure).
Table 1provides descriptive information at baseline for the two groups, with respect to demographic variables, visual acuity, contrast sensitivity, early ARM presence, serum vitamin A levels, general health, and compliance in taking tablets. There were no statistically significant differences in baseline variables between the two groups. Participants were, on average, in their early 70s, in large part white, split evenly between males and females, and found to have three to four comorbid conditions. The vitamin A group tended to have less severe ARM as measured by the AREDS fundus grade compared with the placebo group; however, the observed difference was not statistically significant. Acuity in the tested eye was modestly impaired, on average, in both groups. Compliance in taking the tablet each day was excellent in both groups, with approximately one pill missed by each subject during the 30-day period. Though the vitamin A group tended to be more compliant, the difference was not statistically significant.
Table 2presents descriptive information at baseline for dark-adaptation parameters and subscale scores on the LLQ. The distributions of these variables did not differ between the two groups. LLQ subscale scores averaged for the most part in the 60s to the 80s, indicating moderate to serious problems in visual activities in low-lighting conditions.
Dark-adaptation parameters at the 30-day visit are shown in
Table 3 . Listed in the table for each group separately are the mean of each dark-adaptation parameter adjusted for that parameter’s baseline value. The groups did not significantly differ in cone time constant, cone threshold, rod-cone break or rod threshold. The vitamin A group had significantly larger rod slopes, indicating a faster sensitivity recovery, than did the placebo group (
P < 0.0419). There were some differences in baseline demographic, visual function, and medical characteristics that may explain this difference, despite the lack of statistically significant group differences in the baseline variables. However, after adjustment for all the variables in
Table 1 , the vitamin A group demonstrated larger rod slopes than the placebo group (0.16 vs. 0.14, respectively;
P = 0.08). That the probability associated with this comparison increased was not unexpected, given the number of variables included in the model. The magnitude of the association remained unchanged (i.e., 0.02), suggesting a lack of confounding by the characteristics in
Table 1 . To evaluate whether any observed association between dark-adaptation parameters and treatment group differed according to ARM status, an interaction term was introduced into the model. The interaction term was not statistically significant in any of the models (all
P > 0.5), suggesting that any observed differences between treatment groups were similar in those with and without ARM.
Figure 2shows composite dark adaptation at day 30 for the vitamin A group and the placebo group for cone-mediated responses (top) and rod-mediated responses (bottom). Visual acuity and contrast sensitivity in the tested eye were not different in the two groups at 30 days (
P = 0.949 and
P = 0.621, respectively).
With respect to the LLQ
(Table 3) , there were no group differences in scores on the subscales of driving, extreme lighting, emotional distress, general lighting, and peripheral vision. The vitamin A group had a higher score by 5 points on the mobility subscale compared with the placebo group (
P < 0.0141). This 5-point difference between the vitamin A and placebo groups (88.4 vs. 83.6, respectively) and the statistical significance (
P = 0.0224) remained after adjustment for the variables in
Table 1 . Change from baseline to day 30 in the mobility subscale score on the LLQ was significantly associated with changes in the rod slope (Pearson
r = 0.24,
P = 0.0141). Those who had the most self-reported improvement on the mobility subscale at day 30 tended to have a greater increase in the rod slope parameter
(Fig. 3) . Adjustment for the variables in
Table 1had little influence on this association (Pearson
r = 0.22,
P = 0.0333).
Serum vitamin A significantly increased over the 30-day period in the vitamin A group, from a mean at baseline of 62.47 ± 20.10 to 68.47 ± 19.18 μg/dL (SD) at 30 days (P = 0.0106). There was no change in serum vitamin A in the placebo group (baseline mean, 59.58 ± 14.17 μg/dL; 30-day mean, 59.59 ± 15.37 μg/dL). In the vitamin A group, the change in serum vitamin A over the 30-day period was not associated with the change in rod slope (P = 0.3764). there was a borderline association between change in serum vitamin A and change in scores on the mobility subscale of the LLQ (P = 0.0873).