The Smoke Filled Debate Clearing the air 106195006

The Smoke Filled Debate: Clearing the air 羅維浩 106195006經濟 21 林韋如107048223科管 22

Statisticians vs Doctors • 1950 -1960 • Does smoking cause lung cancer? • Scientists, and even families, were divided.

A Family Debates: Smoking-cancer hypothesis • Jacob Yerushalmy (雅各布·耶魯沙爾米) A biostatistician at the University of California “Yerushalmy opposed the notion that cigarette smoking caused cancer until his dying day, ” -David Lilienfeld(Jacob’s nephew). • Abe Lilienfeld An epidemiologist at Johns Hopkins University(約翰 霍普金斯大學)

A Family Debates: Smoking-cancer hypothesis Scientific Argument: “possible existence of unmeasured factors that cause both craving for nicotine and lung cancer” The Subject of Debates: “caused”

Casual Question In the Past Mid-1700 s �James Lind 發現柑橘類水果可以預防壞血病 Mid-1800 s �John Snow 發現被糞便汙染的水會引起霍亂 All of them are one-to-one relation between cause and effect.

Impossible Triumph for Public Health • Prove no Confounder & Prove Negative • 1964, "Cigarette smoking is causally related to lung cancer in men. “ • “Hill's criteria”, an informal series of guidelines --(Austin Bradford Hill)

Rates of lung cancer and tobacco consumption

CANCER EPIDEMIC: CASE-CONTROL STUDY �Compare patients(Doll&Hill) 已罹患肺癌的病人& 健康的志願受試者(control group) ❖ 649人中，647人都曾吸菸 �Obvious Drawbacks: Retrospective(回顧性) Probability Logic (概率性邏輯的問題) Recall bias(回憶偏見) Selection bias(選擇偏見)

Prospective Study �吸菸者的致癌死亡率，是不吸菸者的24倍 Strong Piece of Evidence： 一致性(Consistency) & “Dose-response effect“—劑量反應效應 懷疑論者的觀點： Failed to compare smokers to otherwise identical nonsmokers Genetically or "constitutionally" different from nonsmoker

Prospective Study 1959 Jerome Cornfield & Abe Lilienfeld 駁回Fisher的觀點。 吸菸基因的假設： 吸煙者罹患肺癌的風險是九倍，混雜因素(confounder factor)也要在吸菸者 中普遍存在九倍。 11% nonsmoker have “smoke gene” → 99%smoker 必須要有 “smoke gene” Cornfield's inequality 12% nonsmoker have “smoke gene” → Mathematically impossible

CORNFIELD'S INEQUALITY ❑Cornfield‘s inequality 導出了因果論正

CORNFIELD'S APPROACH • Sensitive Analysis • Extend Cornfield’s approach Without graphical tools→難以想像的算法&估計技術

THE END OF DEBATES • 1950 s Epidemiologists: “Only statistical” , no “laboratory proof” • If “Laboratory Proof” applied in past … • 1953, RJ雷諾茲公司，化學家克勞德·蒂格（Claude Teague）寫給高層的信 tobacco was “an important etiologic factor in the induction of primary cancer of the lung”

希爾準則 BRADFORD HILL'S CRITERIA

Definitive statement by Health Authorities � 1962 The Royal College of Physicians (UK) “Cigarette smoking was a causative agent in lung cancer” �US Surgeon General Appointment a special advisory committee to study matter Put aside 19 th deterministic causality and statistic “statistical methods can not establish proof of a casual relationship in an association”

Special Advisory Committee I. Consistency II. Strength of association III. Specificity of association IV. temporal relationship V. coherence

Hill’s Ctiteria I. Consistency II. Strength of association III. Specificity of association IV. temporal relationship V. coherence VI. Dose-response VII. Biological plausibility VIII. Analogy IX. Reversibility

VIEWPOINTS , NOT REQUIREMENT “easy to find arguments against each of the criteria”

Perspective of casuality ✔Established the gravity of casual questions and that data alone couldn’t answer them ✔As roadmap for future discovery , it was uncertain flimpsy

吸菸對新生兒的影響 BIRTH-WEIGHT PARADOX

Birth-weight paradox Background ❑Babies of smoking mothers weighted less at birth on average than the babies of nonsmokers �母親吸菸易生出體重過輕新生兒 ❑Death rate of Low-birth-weight infants was more than 20 times than that of normal-birth-weigh �過輕新生兒死亡率較高 Smoking �� low birth weight �� mortality

ADD SOMETHING TO IT • Certain other causes of low weight Serious or life-threatening genetic abnormalties �� Much more harmful

衝突偏差Collider bias