疫学・生物統計学資料，Rev.4.1

Size: px

Start display at page:

Download "疫学・生物統計学資料，Rev.4.1"

としみしどり
5 years ago
Views:

3 (epidemiology) (classification of diseases)

4 (confounding) Fisher (Screening) (Wilson JMG and Jungner G, 1968) ROC

5 64 5

7 (2012) Rothman KJ (2012) Epidemiology: An Introduction 2nd Ed. Oxford Univ. Press J.L. KR (2004) * 1 (2010) (2010) (2005) 1.2 (epidemiology) a b (frequency) (distribution) c (causal relationship) a specified population b c *1

8 W-Bridge (When) (Where) (Who) (What) (Why)

9 1.4 (classification of diseases) (classification of diseases) ICD ICD (International Classification of Diseases) a ICD ICD-11 b a b ICD-10 ICD WHO ICD ICD-9 WHO (population at risk)

10 10 1

11 Descriptive epidemiology

12 12 2 Last JM [Ed.] A Dictionary of Epidemiology 4th Ed. descriptive study Natural Experiment Last JM ed., A Dictionary of Epidemiology 4th ed. John Snow Snow Natural Experiment Ecological Study ecological fallacy

13 ecological fallacy 1 2 A B X=1 X=0 X=1 X=0 A B Y=1 Y= Y= X=1 X=0 Y=1 2 X=1 X=0 A B Y=1 7/8 Y= Y= X=1 X=0 Y=1 1/2 X=1 X=0 A B Y=1 8/7 Y= Y= Greenland S (2001) Int. J. Epidemiol. 30:

14 cross sectional study (longitudinal study) case control study (case) (control) (retrospective study) cohort 1980 (followup study)

15 (nested case-control study) 2.4 ( clinical trial ) RCT (Randomized Controlled Trial; ) RCT

17 17 3 Random sampling a Cluster sampling a Stratified random sampling a (0,1) b c 1 N p

18 18 3 R *1 2 OK print(sapply(1:p, function(x, y) which(x==y), rank(i))) print(sample(1:n, p, replace=false)) Cluster sampling 50 10% 5 10% 3.4 Probability Proportionate Sampling (PPS) *1

19 R Vanderbilt Power and Sample Size *2 *2

20 20 3

21 21 4 a a (prevalence) (incidence rate) * 1 (cumulative incidence rate) *1

22 (prevalence) (odds) (incidence rate) 1 A Dictionary of Epidemiology, 4th Ed. incidence!!

23 (mortality rate) 1 1 (category-specific mortality rate) (disease-specific mortality rate) (cumulative incidence rate=risk) (case-fatality ratio) case-fatality rate rate ratio % * 2 100% 50 90% 60% % % * % 0.045% 0.005% CFR (confirmed case fatality ratio; ccfr) (symptomatic case fatality ratio; scfr)

24 (proportional mortality ratio; PMR) (competing risk) PMI (proportional mortality indicator) % (2011),, 32Special Issue: S35-S i P i

25 m i = i m i P i i P i (SMR) d i p i i M i SMR = d i M i p i = d/ i p i i M i p i / i p i SMR = SMR i M i P i i P i

27 (excess risk) (attributable risk)

28 28 5 (relative risk) (odds ratio) a b c d *1 (disease odds ratio) (a/b) (c/d) = ad bc (exposure odds ratio) (a/c) (b/d) = ad bc 5.3 (Attributable Proportion) (Attributable Population=Attributable Fraction) *1

29 X PYLL (Potential Years of Life Lost) (Loss of Life Expectancy at Age Zero) PYLL

31 * A B A B p p *1

32 Henle-Koch * 2 (1) Strength 0.07/1000/ /1000/ /1000/ (2) Consistency (3) Specificity (4) Temporality (5) Biological gradient (6) Plausibility *2 Sir Austin Bradford Hill The Environment and Disease: Association or Causation?, Proceedings of the Royal Society of Medicine, 58: , Hill

33 (7) Coherence (8) Experiment (9) Analogy 6.4!! (counterfactual) (causal inference) 6.5 A (exposure) A B A B A B *3 *3 1 1

34 cf B C cf cf (sufficient causes) (component causes) ; Rothman, 2002

35 B A E B A H C A J C D F G F I ひとつの因果メカニズムひとつの構成要因 * 4 =induction period *4

37 selection bias; information bias;

38 Neyman s (90/10) / (50/50)=9 9 90% 5% % (Graves et al., 1991) 2 Neyman s Bias Prevalence-incidence bias (Hill et al., JCE 56: 293-, 2003) Rothman, 2012 Boston Globe Otmar Suitner Leonard Bernstein Herbert von Karajan Karl Böhm

39 (RDD) Berkson

40 40 7 (Berkson s bias) (Healthy Worker Effect)

41 7.4 (confounding) 41 10% (maternal recall bias) A A A A 7.4 (confounding)

43 Panum PL ( ) Panum Gregg NM 1941 ( ) 1884 B1 Snow J ( ) 1854 John Snow (spot map) L 461/ S&V 4093/ S&V

44 Frammingham Heart Study , MD X X html 8075 : ph % 80% 99% 78%

45 INTERSALT INTERSALT BMI Na/K SMON *1 *2 * pdf * pdf

46 46 8

47 Fisher The lady tasting tea Cambridge R.A. Fisher!!

48 George Orwell 11 rules for perfect tea making 10 The Royal Society of Chemistry (2003) Dr. Andrew Stapley BBC RSC a ) 130cc 30cc b a b / % 9.3 (1) (2) (3) (4) 5% 80% 30% 15% 95 R

49 power.prop.test(p1=0.15, p2=0.3, sig.level=0.05, power=0.8, alternative="one.sided") Welch t dose-response relationship 9.6

50 I II III RCT (Randomized Controlled Trial) IV International Conference on Harmonization (ICH) (GCP) IRB (Instituitional Review Board)

51 GCP 46 ITT (intention to treat / intent to treat) * ITT ITT (2001) *2 A B 2 100% 4 Aa Ba Ab Bb A A B A A B B B Aa Bb Ba Ab Aa A Bb B Aa Ba A Ab Bb B Ab Bb B A B Aa Ab A Ba Bb B ITT Ab Ba A B *1 1 *2

52 52 9 A Aa+Ba A Aa+Ab A A Aa+Ba A Aa+Ab A ITT OK ITT ITT (Multiple Imputation) * = = (NNT ) *3 R mice

53 % (Quiz) (32/1000)/(41/1000) = 1 32/41 = 9/41 = % (41/1000) (32/1000) = 9/1000 = % NNT 1/(9/1000) = 1000/9 = % exp(ln(32/41) ± 1.96 (1/32 1/ /41 1/1000)) (0.495, 1.223) ( 0.223, 0.505) ( ) ± 1.96 ( ) 41 ( ) ( ) ( 0.008, 0.255) RCT RCT

55 (Screening) US PSTF (preventive services task force)

56 (lead time) (Wilson JMG and Jungner G, 1968) 1. eg. eg follow-up 9.

57 Andermann (2008) 2006 eg. DALYs Gold Standard (sensitivity) positive in disease (specificity) negative in health a b c d a (a+c) d (b+d) (a/(a+c)) (b/(b+d)) 1

58 58 10 (d/(b+d)) (c/(a+c)) 1 (positive predictive value) (negative predictive value) a (a+b) d (c+d) (a + c) << (b + d) % 10 10% 10% a/(a + b) = a/(2a + 9) a = 9/2 a/(a + b) 0 a 1 a = 1 1/11 10% (test-retest reliability) (interinstitute difference) (inter-rater difference) test-retest reliability ROC ROC Receiver Operating Characteristic ROC ROC ROC

59 10.5 ROC RH SW EH ROC ROC =1 1 0 ROC (0,1) ROC AUC Area under curve; ROC 10

60 /3 = 1 3/7 = R x <- c(20,22,28,13,19,21,11,25,16,19) y <- c(1,1,1,0,0,0,0,0,0,0) # Epi require(epi) ROC(x,y,plot="ROC") # fmsb require(fmsb) res <- roc(x,y) print(res) plot(res) AUC

61 % % kg VMA HVA % 5% (OAE) ABR % 60%

62 cm 90cm 100 cm 2 TG150mg/dL HDL 40mg/dL SBP/DBP 130/85mmHg 110mg/dL * 1 HPV X X CT MRI *1

63 X TG HDL LDL AST ALT GTP HbA1c

64 64 5W-Bridge, 8, 7, 22, 22, 22, 28, 27, 7, 7, 8, 21, 8, 28, 27, 8, 7, 7, 21, 21, 22, 28, 28, 21, 28, 28, 21

こんにちは由美子です

こんにちは由美子です Prevalence (proportion) 1. 1991 23.8 2. 1960 52 85 2477 310 cross sectional study prevalence Time referent: prevalence 1985 40 45 prevalence 0.5 80 85 43 Time referent 1985 time referent Risk and Cumulative