Gr8 @EpiEllie “Sick individuals and sick populations” (SISP) a classic; along with his many other writings (including his valedictory book “Strategy of Preventive Medicine”) it should have remade the epidemiological landscape. A thread here on a few interpretive differences
@EpiEllie 1/n His views don’t fit within the “well defined interventions” envelope. His central idea is that the causes of health differences between populations and between individuals can be different. Consider his oft-used figure on circulating cholesterol levels in Japan and Finland 
@EpiEllie 2/n Between Japan and Finland dietary intake is Key (pun intended), but within Japan and Finland, where the variation in diet is relatively small, different factors – including genetics and other individual-level factors - will influence cholesterol level ahajournals.org/doi/pdf/10.116…)
@EpiEllie 3/n Among the individual level factors he was concerned with was illness – early stages of cancer could influence cholesterol level; indeed he led the first study suggesting low cholesterol level within a given population predicted raised colon cancer risk thelancet.com/journals/lance…
@EpiEllie 4/n Across his writings he distinguished between factors that lay within the usual influences on cholesterol level, such as diet, exercise and obesity & those that lay outside of the se ..
@EpiEllie 5/n such as use of the cholesterol lowering drug clofibrate, very high levels (outside the range of historical primate consumption that we evolved against) of polyunsaturated fatty acid intake, many different acute and chronic illnesses, etc
@EpiEllie 6/n In his 1974 colon cancer paper he noted that a positive correlation across populations is seen between colon cancer incidence and blood cholesterol level, with a negative association seen within populations. The *central* point of SISP was already here
@EpiEllie 7/n One of the papers I co-wrote with Geoffrey was a detailed analysis of the higher risk of some adverse outcomes for those with low circulating cholesterol; we frequently discussed these issues (he was the gentlest and most non-dogmatic teacher) jamanetwork.com/journals/jama/…
@EpiEllie 8/n Geoffrey also studied the U-shaped curve between body weight and mortality (similar to that of cholesterol) and the influence of factors such as illness on the shape of this association ncbi.nlm.nih.gov/pmc/articles/P…
@EpiEllie 9/n He would have considered it bizarre that anyone would consider that illness induced cholesterol lowering or weight reduction would be beneficial. His population approach meant he thought that what led to differences between populations were the intervention targets.
@EpiEllie 10/n These intervention targets might contribute only a modest proportion to differences within populations (especially when estimation of such is influenced by measurement error)
@EpiEllie 11/n Cholesterol level and weight were definite “causes” in his world, as familiarity with his broader writings makes clear. But factors that don’t contribute substantially to the weight or cholesterol level differences between populations – such as illness, genetics, clofibrate,
@EpiEllie 12/n other “non-naturals” (such as @_MiguelHernan cute having an arm cut off to reduce BMI) – are not intervention targets. A reasonable starting hypothesis – and one you can get evidence on – is that the between-population influences on causes ..
@EpiEllie @_MiguelHernan 13/n (and yes, Rose thought they were causes) such as cholesterol and weight would have roughly similar effects on outcomes, given their effect on the exposure. For LDL cholesterol we have good evidence that this is largely true.
@EpiEllie @_MiguelHernan 14/n Many targets working through different pathways (including cholesterol absorption across the gut) produce the same effect on coronary heart disease (CHD) – the graph shows comparison of drug trials vs specific genetic influences on these targets
ncbi.nlm.nih.gov/pmc/articles/P…
ncbi.nlm.nih.gov/pmc/articles/P…
@EpiEllie @_MiguelHernan 15/n with the expected larger predicted effect from genetic modification because of the lifelong influence on cholesterol of genetic differences compared to the few years within a trial
@EpiEllie @_MiguelHernan 16/n The long term effect of cholesterol was demonstrated by Geoffrey using serial population data in 1982 ncbi.nlm.nih.gov/pmc/articles/P…
@EpiEllie @_MiguelHernan 17/n The lifetime vs short term trial ratio of effects of cholesterol estimated using the tiny amount of Mendelian randomization data available in 2004 is pretty much as it is now, with orders of magnitude more trial and MR data academic.oup.com/ije/article/33…
@EpiEllie @_MiguelHernan 18/ nStatin trials demonstrate remarkably stable relative risk reductions in CHD across subgroups with greatly distinct baseline absolute risk; other ways of modifying cholesterol through different approaches including the wide range of drugs targeting different steps ..
@EpiEllie @_MiguelHernan 19/n in the cholesterol metabolic pathways produce the same effects as statins
thelancet.com/pdfs/journals/…
thelancet.com/pdfs/journals/…
@EpiEllie @_MiguelHernan 20/n Genetic differences between individuals in these pathways produce small robust effects on cholesterol and through this CHD; whilst the genetic variants have tiny individual level effects, they point to interventions that bring meaningful benefit for individuals
@EpiEllie @_MiguelHernan 21/n ... and substantial population benefits. Rose’s “prevention paradox” follows from his understanding of the generally modest between-individual effects (and low benefit to risk ratio) verses substantial between population differences.
@EpiEllie @_MiguelHernan 22/n My first grant (with an anthropologist) was for a qualitative study of how the public understand this apparent paradox
onlinelibrary.wiley.com/doi/abs/10.111…
onlinelibrary.wiley.com/doi/abs/10.111…
@EpiEllie @_MiguelHernan 23/n Rose’s conclusion to “Strategy of Preventive Medicine” (if you haven’t read this, do) “The primary determinants of disease are mainly economic and social, and therefore its remedies must also be economic and social. Medicine and politics cannot and should not be kept apart”
@EpiEllie @_MiguelHernan 24/n For a long reflection on how the causes of within and between population disease can be different, see academic.oup.com/ije/article/40…
