[Skip to content]

Sign up for our daily newsletter
The Actuary The magazine of the Institute & Faculty of Actuaries
.

Matching trend to trend

Patrick Carroll’s article on breastcancer increases drew my interestbut not my support. Two trendsare demonstrated and, I assume,supported by the data. Abortionshave increased significantly forwomen born since the 1930s, particularlywith a change in the lawto legalise them. Cancer rates havealso increased and the thrust ofthe article seems to be to try tograsp reasons (cause and effect)linking the two.

One should take care not to confusecorrelations with causation.Two trends may be well correlatedbut independently related to other causes. For instance, one mightlink these two issues to a greatercontrol of a woman’s own bodyand health taking place over thegeneration charted in Mr Carroll’sarticle. However, the first reasonhe gives for abortion to cause canceris that after an abortion breastcells suffer interrupted hormonaldevelopment and are moresusceptible to cancer. Yet the interruptionhappens generally in the20s and 30s of the woman’s life.But it is in her 50s that breastcancer has shown such aremarked-upon increase. It is difficultto see how a hormonal susceptibilitycreates cancer whichlies latent for 20 years and enduresbreast cancer screening throughouta woman’s 40s only to emergein her 50s.

Mr Carroll shows a surge inbreast cancer diagnosis in 1989with the introduction of screeningbut then dismisses the continuinghigh level from 1993 onwards.Screening is intended to provideearly detection (which it does) andso should surely continue to beeffective in elevating detectionwell beyond its initial introduction.We are not shown figures forwomen in their 60s, so I cannotcomment on what those figuresshow.

The reasons for changes indisease pattern are complex anddeserve much more study. It willbe of most help, I hope, ifresearchers can produce clearmedical reasons for cause andeffect that help not only in explanationbut also intervention thatcan prevent and cure the problemsooner.

However, as actuaries we canapplaud any paper that gives us clear correlations. If two items areclearly correlated with oneanother, one becomes a rating factorfor the other. We do not needto identify which is cause, whichis effect, or whether neither is – wecan still use them fairly in ourpricing structures.