Modification of Tamoxifen Response: What Have We Learned?

TO THEEDITOR:InarecentJournalofClinicalOncologyeditorial, Desta and Flockhart 1 asked a yes-no question about the germline pharmacogeneticsoftamoxifenresponse:“Havewelearnedenough?” They ultimately answer by writing that recommendations regarding genetic testing for CYP2D6 variants must await further data, but the list of open questions preceding this answer did not include the question of whether there is a main effect. We therefore raise this question about the main effect: “Has it been established that tamoxifen is a less effectiveadjuvanttherapyforestrogenreceptor‐positivebreastcancer patients with functionally variant CYP2D6 allele(s) than for those without functionally variant alleles?” Figure 1 depicts the results of the five epidemiologic studies of thisquestion. 2-6 Toexaminethedepartureofthedistributionofthese results from the null (ie, relative risk of 1.0), we first ranked the five relative risks of recurrence from lowest to highest (citations 2 to 6 in ascending order). We then plotted each study’s relative risk and its 95%CIagainsttheinversenormalofitsrankpercentile. 7 Thestippled lineshowstheregressionofinverse-varianceweightedlog-relativerisk against the inverse normal of rank percentile. If the accumulated evidenceisarandomsamplefromalognormaldistributionofrelative risks centered on the null, then the relative risks should fall along this line and the line should intersect the x-axis at zero. The results of the accumulated studies fit just such a pattern. Indeed, the inversevariance weighted geometric mean of the relative risks equals 1.04 (95%CI,0.77,1.41).Eachstudyhascertainstrengthsandlimitations, some of which were pointed out by Desta and Flockhart, and consideration of them against one another is of value. Nonetheless, the simplest explanation for the results to date is that they are sampled from an underlying null association. The integer above each interval in Fig 1 shows the number of women in the study (as best we can determine) with a recurrence and with the variant allele(s). Because both recurrence and the variant allele are rare, this number is the primary determinant of the study’s precision. Summed over all of the studies, there are only approximately sixty women with both a recurrence and the variant allele(s), about half from studies reporting a protective association and half fromstudiesreportingacausalassociation.Inourview,sixtywomen, evenlydistributedaboutthenull,istoosmallanumbertoconsiderthe association between CYP2D6-variant alleles and tamoxifen effective