Dietary lutein and zeaxanthin: authors' response

We were interested to see the letter by Mitchell,¹ written in response to our earlier publication.²

Mitchell raises several points, the first of which is the way one handles missing data. Age related macular degeneration (AMD) can be assessed either clinically or photographically. Rather than throw away data and run the potential of bias by excluding patients who did not have a retinal photograph, we chose to include in our analysis those people for whom we did not have a retinal photograph for one reason or another but for whom we had clinical macular grading. In separate analyses the data were not materially different if the analysis was confined to only those with photograding. In the data Mitchell et al report they excluded 15% of their sample as they lacked data and there is no evidence of what bias this may have induced.

We were very pleased to see the data from the Blue Mountains Eye Study (BMES) and recognise the advantage of incidence data over cross sectional data. However, both cross sectional and longitudinal studies are susceptible to recall bias in the ascertainment of historical dietary intake. In nutritionally non‐deficient populations people do alter their diets in response to the available commercial information. However, the bias caused by public awareness of hypothesised protective associations applies equally to both cross sectional and longitudinal observational studies, as many know about either their disease or family predisposition to AMD before participation. However, our data were collected before the recent upsurge in publicity given to the use of nutritional supplement for macular degeneration.

The most striking thing from the data presented by Mitchell is that they show no evidence of a protective effect of dietary or supplement intake of lutein and zeaxanthin. The consistent finding of these two population based studies must seriously challenge the association reported by previous case control studies. Case‐control studies are always strongly influenced by the possibility of bias in selection of controls. The fact that the dietary intakes in the two population based studies is so much lower than the volunteer case‐control studies clearly indicates “healthy volunteer bias” as we point out in our paper.²

The lack of effect modification by linoleic acid found in the BMES data is interesting and possibly not surprising. We found a complex U‐shaped interaction between dietary intake and the risk of AMD and the decision as to where cut points are taken will strongly influence the outcome (fig 1). It would be more informative if Mitchell presented his data either with the same cut points we had used, or presented his data in a continuous form.

Figure 1 Estimated odds ratios derived from the analyses, including all participants.

Nevertheless, these two studies confirm the lack of evidence to support recommendations for a specific increase in dietary intake of lutein and zeaxanthin to protect against macular degeneration. At the same time of course no one would encourage people to deviate from a normal healthy diet including fresh fruit and vegetables.¹