InaNationalEye Institutepress release1 issuedonMay5, 2013, with simultaneous online publication in JAMA,2 the results from the second phase of the Age-Related Eye Disease Study 2 (AREDS2)were announced. This press release was titled “NIH study provides clarity on supplements for protection against blinding eye disease,” and indeed in its first paragraph, the primary outcome was clearly stated: “The plantderivedantioxidants lutein andzeaxanthinalsohadnooverall effect on [age-related macular degeneration (AMD)] when added to the combination; however, theywere safer than the relatedantioxidantbeta-carotene.”This studyprovideduswith a striking example of a major clinical trial in which the primary outcome was negative, but yet its broader findings prompted substantial changes in the formulation of so-called eye vitamins that constitute amajormarket presence and are likely recommended by most eye care practitioners for a patient they determine to be at risk for developing neovascular AMD. The ability of companies to market their eye vitamins as providing an exact match to the AREDS2 formula is evidently important because we encounter such statements in venues ranging from our professional journals to television commercials.Whilesomemightsaythat thetrainhas longsince left thestation, thepurposeof thiseditorial is tocriticallyevaluate the evidenceunderlying formulation changes that are under way or have already taken place. In this issue’s article,3 theAREDS2 investigators provided uswith a detailed look at additional analyses that led them to support adding lutein and zeaxanthin to, and deleting beta carotene from, the previously recommended oral formulation for eyehealth, termed theAREDS formula. Like the analysespresented in their initial report, someof theseanalyseswere termed exploratory in nature and relied on results from inspecting subgroups that were not prespecified and thuswere post hoc in nature. Subgroup analyses have been viewed as a 2-edged sword by Wittes,4 who wrote: “If reporting on subgroups is tempting but treacherous, failing to report on them seems unscientific and incurious.” As Fleming and Watelet5 pointed out, the results of such analyses should usually be viewed as hypothesis generating rather than definitive in nature.Theyadvised thatdefinitiveconclusions fromsuchanalyses occur only in rare situations: when very strong associations are found that are biologically plausible and supported by results fromother sources. Even so, claimsof treatment efficacy that result from such findings are not uncommon. The subgroup analysis finding from the Steroids for Corneal Ulcers Trial that corticosteroidsmay be beneficial for treatment of bacterial keratitis when baseline visual acuity is count fingers or worse, or when the ulcer was completely centrally located,was advised to be viewedwith caution in a letter to the editor.6 In their reply, the trial’s investigators fully agreed that their findings, while based on prespecified subgroup analyses, should be subject to further confirmation.7When pivotal phase 3 trials of proposed treatments yieldnegativeor equivocal results, a commonly encountereddirective from the sponsor is to revisit the data in an attempt to identify subgroups that showefficacy; thus thederogatory termdatadredgingand the humorous, but sometimes real, claim that if you torture your data long enough, they will confess. Acritical lookattheevidencefromAREDS2foranefficacious effectof lutein/zeaxanthinsupplementationstartswith theauthors’JAMAarticle,2 inwhichaprespecifiedcomparisonofthose receiving lutein/zeaxanthinvs thosewhodidnot receive lutein/ zeaxanthin revealeda 10%reduction in the risk forprogression to lateAMD.A secondprespecified analysis showed that those in the lowestquintileofdietary intakeof lutein/zeaxanthinwho received lutein/zeaxanthin alongwith the original AREDS formulationhada26%reducedriskforprogressiontolateAMDrelative to participants receiving the original AREDS formulation without lutein/zeaxanthin. Finally, an exploratory look at the groupgiven lutein/zeaxanthin in its supplementshowedan11% reduced risk for developingneovascularAMDvs thosewho received only beta carotene in their supplement. The further analyses presented herein include an exploratory analysis of the 1114 participants who received lutein/ zeaxanthinadded toanAREDS formulationwithoutbeta carotene vs the 1117 participants who received only beta carotene in the AREDS formulation. This revealed that those receiving lutein/zeaxanthinwithoutbetacarotenesupplementhadan18% reduction in the risk for late AMD and a 22% reduction in the risk for neovascular AMD (both which met the P < .05 criterion), as well as a 6% reduction in the risk for geographic AMD (which yielded aP value of .67). A prespecified analysis of progressionon theAREDS severity scale forAMD8didnot reveal a beneficial effectof lutein/zeaxanthinsupplementation (withor withoutbeta carotene) on reducing the risk for a 2ormore step progression.However, inanexploratory lookat thesubsetwho only received lutein/zeaxanthin (withoutbeta carotene) vs the subset who received only beta carotene, lutein/zeaxanthin supplementation reduced the risk for a 2 ormore stepprogressionby 13%.The resultswere compelling for the subset of 1748 subjectswhohadbilateral largedrusenandwere evaluated for progression toneovascularAMD,whichdeveloped in 11.5%(99 Related article page 142 Opinion