Biostat for Epidemiology
Biostat for Epidemiology EPP 247
Popular in Course
Popular in Med Epidemiology & Prev Med
Virgie Eichmann DDS
verified elite notetaker
verified elite notetaker
verified elite notetaker
CHEM 1040 - 002
verified elite notetaker
verified elite notetaker
This 21 page Class Notes was uploaded by Virgie Eichmann DDS on Tuesday September 8, 2015. The Class Notes belongs to EPP 247 at University of California - Davis taught by Staff in Fall. Since its upload, it has received 60 views. For similar materials see /class/187517/epp-247-university-of-california-davis in Med Epidemiology & Prev Med at University of California - Davis.
Reviews for Biostat for Epidemiology
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 09/08/15
EPP 247 Biostatistical Methods in Epidemiology 462007 Kyoungmi Kim PhD Division of Biostatistics amp Rowe Program in Human Genetics UC Davis School of Medicine EPP 2477Biostatistical Methods in Epidemiology Biostatistics courses so far 1 Introduction to Medical Statistics Data summary and presentation Exploratory data analysis A bit of everything 2 Statistical Analysis of Laboratory Data Gene expression arrays and other highthroughput biological assay technologies Analysis of data and design of experiments for laboratory data 3 Biostatistics for Clinical Research Clinical trial design Analysis of clinical trial data EPP 2477Biostatistical Methods in 462007 Epidemiology Taz39x biostatistics course EPP 247 4 Biostatistical methods in epidemiology gt Analysis of data from observational or epidemiological studies gt In contrast to the randomized clinical trials EPP 2477Biostatistical Methods in 462007 Epidemiology Course outline gt gt Goal To better understand statistical methods for epidemiological observational data Specific Aim Test for the association between exposure eg environmental or genetic risk factors and disease outcome with or without controlling for confounding risk factors gt Categorical data analysis Contingency tables and factors 2 x 2 Tables Stratified methods MantelHaenszel gt Logistic regression gt Survival analysis for cohort data KaplanMeier curves Cox proportional hazards model Eventually Be able to interpret and critique the results of application of such statistical procedures as found in the health sciences literature 462007 EPP 2477Biostatistical Methods in Epidemiology Types of study design Experimental studies Randomized controlled Clinical trials Assign subjects at random to different procedures or treatments whose effects is desired to discover Observational epidemiological studies Cohort studies our focus Casecontrol studies our focus Crosssectional studies EPP 2477Biostatistical Methods in 462007 Epidemiology Types of study design used in epidemiologic work Casecontrol studies 1 Subjects are selected according to their disease status 1 Cases are defined as those affected with the particular disease of interest where controls although at risk for developing the disease are unaffected with the disease a Retrospective studies selecting study subjects on the basis of their disease status and looking backward to a possible cause EPP 2477Biostatistical Methods in 462007 Epidemiology Cohort studies El Study subjects are selected on the basis of their exposure status are diseasefree at the start of the followup period and are followed forward in time to onset of disease Prospective studies the exposure is defined before disease occurs Advantages the ability to establish a temporal relationship between exposure and disease the suitability for the study of rare exposures the ability to study multiple effects of a single exposure Drawbacks costly and timeconsuming and not efficient for the study of rare diseases such as cancer and diseases with late onset such as Alzheimer disease 462007 EPP 2477Biostatistical Methods in Epidemiology Crosssectional studies 1 Subjects are asked about their current disease status and their current or past exposure status 1 Prevalence studies the prevalence of disease at one point in time is compared between exposed and unexposed groups EPP 2477Biostatistical Methods in 462007 Epidemiology Measures of effect for categorical data I Relative risk RR El lOd El El nAtt El Measure of association between exposure risk factor and disease D the ratio ofthe incidence ofthe disease in the exposed and junexposed groups using prospectivecohort ata contigency table Disease RR IeIo aabccd where le is the incidence of D among the exposed and la is Exposure Yes N0 the incidence of D in the unexposed ds ratio OR Yes a b Measure of association between E and D the No c d ratio ofthe odds of E among affected subjects cases to the odds of E among unaffected 30 bd subjects controls using casecontrol or cross Total m1 m2 sectional study data OR acbdadbc ributable risk AR Measure of impact of the excess risk of D among susceptible individuals compared to those who are not The AR may be expressed as the risk difference or Ielo Table1 Hypothetical table depicting exposuredisease relationship 2x2 Total ab n1 cd n2 abcd 462007 EPP 2477Biostatistical Methods in Epidemiology Example Cancer A study by MacHahon et al Age at first birth and breast cancer risk Bull World Health Organ 1970432209 21 Background A hypothesis has been proposed that breast cancer in women is caused in part by events that occur between the age at menarche and the age at first childbirth In particular the hypothesis is that the risk of breast cancer increases as the length of this time interval increases If this theory is correct then an important risk factor for breast cancer is age at first birth I An international study was set up to test this hypothesis MacHahon et al 1970 Breast cancer cases were identified among women in selected hospitals in 6 countries Controls were chosen from women of comparable age who were in the hospital at the same time as the cases but who did not have breast cancer All women were asked about their age at first birth EPP 2477Biostatistical Methods in 462007 Epidemiology 10 Example Cancer I Data The set of women with at least one birth was arbitrarily divided into two categories 1 women whose age at first birth was 29 and 2 women whose age at first birth was 230The results are shown in Table 2 I 01 Provide a point estimate and cont d Table 2 Data for the international study in cancer example comparing age at first birth in breast cancer cases with comparable controls Cancer status Case Control a 95 CI for the relative risk RR total of breast cancer among women of age 230 at first birth compared to women of age 329 at first birth I 02 Provide a point estimate an a 95 CI for the odds ratio OR relating age at first birth to breas cancer incidence d t Age at first birth 230 529 683 2537 1498 8747 2181 11284 total 3220 1 O 245 13465 EPP 2477Biostatistical Methods in 462007 Epidemiology Cancer example cont d Relative risk Age at first birth Cancer status 230 529 total Case 683 2537 3220 Control 1498 8747 10 245 total 2181 11284 13465 The estimated RR is i 31 RRZ E439 n2 11284 We can obtain a confidence 1nRRiZM2 l interval for the RR by first obtaining a confidence interval 1ng39i196 1498 8747 j for the logRR using the normal 6832181 253711284 approximation and then Zlna39i007 exponentiate the endpoints of 4025804 the interval for the RR 36039258360394129i149 EPP 2477Biostatistical Methods in 462007 Epidemiology 12 Cancer example cont d Odds ratio Age at first birth Cancer status 230 529 total Case 683 2537 3220 Control 1498 8747 10 245 total 2181 11284 13465 The estimated OR is 0 2 683 X 8747 21572 be 1498 X 2537 We can obtain a confidence interval for the OR by first obtaining a confidence interval for the logOR using the normal 1 1 1 1 approxnmation and then exponentiate the 1n1572i196 wj endpoints of the interval for the OR 1nORJrZ1m2 j 0452i0 101 03520553 Interpretation the 0R greater than 1 3 90352990553 2 1429174 indicate a greater likelihood of breast cancer among women of age 23 0 at rst birth EPP 2477Biostatistic21 Methods in 462007 Epidemiology 13 Assessing confounding Bias Confounding occurs when an observed association is due to the mixing of effects between the exposure the disease and a third factor ie confounder that is associated with the exposure and independently affects the risk of developing the disease Confounding bias can be controlled through various statistical analytic strategies including a For casecontrol studies stratification for a few strata and variables logistic regression a For cohort data Survival analyses are often used EPP 2477Biostatistical Methods in 462007 Epidemiology 14 Do we really have to control for confounding factors 462007 EPP 2477Biostatistical Methods in Epidemiology No confounding effects AMEHIICAN JDILntrIAL or EJ IIFEHHJLUfH39 Vol 29 Na 1 Eupyrinhl ID 19 by The John Hupkir u e Unin mily Scrum mquot Hm h and Public Heslml l Printed In if 11 All rlgl lli rmrvcrd Original Contributions FAMIAL OVARIAN CANCER A PDPULATIONBASED CASECONTROL STUWIJ E l JDELLEN M SCHILDKHAUT AND W DOUGLAS THDMPSDN Schlldkraut IL M Yale IJ Sal1cm ul39 Madbclna NW Haven CT D5510 and W D Truman Familial ovarian cancer a populationbased casecontrol unruly Am J Epldammr 1983125455 53 Data I39mm a mul nerlmr populationham ciao mutual study maria analysed can assess 39lha dagm M aggrega39llon al39 mariaquot summer In 39larnilis Included as min were 453 wanban aged 2501 54 who had been newly dlagmsad will claimElia ovarian unclear The frequency url h whidh cases reinDried a Ianlily binary at ovarian cancer was campred Him he quot uency law a group of 21455 canl mls salaried by random digit dining The odds rating or marlin cancer in lim and muddegree relMikes were 38 55 mn denrze iMBWBl Cl 1811 and 29 95 Gil 15 53 respeclively came wilh warnan with no family history or ovarian mncer The null hypulihggis all no aa mia un ma Blackmail on both ma matemal and paternal sing of me familial studied Ovarian cancer in melaninas was quotsnarled by no an wlrh mallng leslms but not by women wilh borderline lasiansu II r L39 a quot Wu II 2 u an 1 clients of any u l39 several cant ames or In errors ln reporting lnmlly Mala m l l vlk 1M III I 39 nilg IJI IH 1 I iIIH EPP 2477Biostatistical Methods in 462007 Epidemiology I Confounding effects on genetic association studies Am J39 Hum Garrett 43520 526 1938 Gm35quot3quot4 and Type 2 Diabetes Mellitus An Association in American Indians with Genetic Admixture William C Knowler Robert C WilliamsT I David J Putinquot and Arthur G Steinberg Diabetes and Arthritis Epo39demiobgy Suction National Institute of Diabetes and Digestive and Kidney Diseases Phoenix YHistocompmiaility Laboratory Blood Systems Irv Bed Scottsrhle tDepamnem at Andlrnpaday Arimm State University We and Depnrtmant crf Biology Case Western Reserve University Cleveland Summary in a sample of 4920 Natlw Americans of the Pima and Papago tribes that is a vcry strong negative as sociation between the Gm haplotype GmJ5393 and type 2 01 noninsulindcpcndent diabetes mellitus prevalence ratio 027 95 con dence interval 018 040 One might conclude from this observation that the absence of this haplotype or the presence of a closely linked gene is a causal risk factor for the disease It is shown that Girl39s 339 is a marker for Caucasian admixturt and it is must lith the presence of Caucasian alleles and the concomitant decrease of Indian alleles that lowers the risk for diabetes rather than the direct action oi the haplotype or of a closely linked locus This study demonstrates both the Eutenrial confounding effect of admixture on the intergretalion of disease association studies and the im portance of considering genetic admixture or excluding individuals with genetic admixture in studies of genetic markers of disease The relationship between this admixture marker and the prevalence of diabetes also suggests a strong genetic component in the susceptibility to type 2 diabetes in Pima and Papago In diam EPP 2477Biostatistical Methods in 462007 Epidemiology 17 I Variable age of onset is also a confounding factor Am 1 Hum Genet 45521 529 1989 Evaluating Genetic Association among Ovarian Breast and Endometrial Cancer Evidence for a BreastOvarian Cancer Relationship Joellen M Schildkraut Neil RischT39I and W Douglas Thompson T 39Department of Epidemiology School of Public Health Universlty of North leirla Chapel Hill and Departments of 1Epidemiodogy and Public Health and Human Genetics Yale Universlty School of Medicine New Haven Cl39 Summary The possibility of a genetic relationship between ovarian breast and endometrial cancer was investigated in data from a large 39 39 r r 39 quotn 1 39 study the Cancer and Steroid Hormone Study conducted by the Centers for Disease Control CDC Ageadjusted relative rislG 1135 for mothers and sisters of 493 ovarian cancer cases 895 reast cancer casmms Wok were calculated Signi cantly elevated ageadiusted RRs were found for ovarian cancer RR 28 95 con dence interval CI 1649 and breast cancer RR 116 95 C1 11 21 among relatives of ovarian cancer probands and for breast cancer RR 21 95 CI 17 25 and ovarian cancer RR 17 95 CI 10 20 among relatives of breast cancer probands Relatives of cndometrial cancer probands had an elevated RR for endometrial cancer only RR 27 95 CI 16 48 The genetic relationship between ovarian breast and endometrial cancer was tested using a mul tivariate polygcnic threshold model developed by Smith 1976 which was modi ed to accommodate three classes of x L r 39 of 39 39 39 quot39 for ovarian breast and endometrial cancer were 40 56 and 52 respectively There was a signi cant genetic correlation between ovarian and breast cancer Ru 484 Evidence for signi cant genetic overlap between endometrial cancer and either ovarian or breast cancer was not found These results suggest the existence of a familial breastovarian cancer syndrome Endomctrial cancer while heritable appears to be genetically unrelated EPP 2477Biostatistical Methods in 462007 Epidemiology 18 The length of follow up in cohort studies is also a important confounding factor x 13916 mnmimif American 5mm 15 19mm 54 ND Lo I Mr I nil1h Journal 39739 W m mquot 7 maul H m j r i l l mm mmum L IdIOIQE u l lr H quot1 III III in In W Coronary Risk Associated with Age and Sex at Parental Heart Disease in the Framingham Jue en M Schildkraut PhD Rxhard H Myers PhD L Adrienne Cupptes PM Dan K rudely MA and WHYam B annel M3 junlnrlluallllIutInyunudi jm hmdmh it u hl uh Maril i uf l mw HR14 AHJMII IWIM EPP 2477Biostatistical Methods in 462007 Epidemiology Course project Identify a journal paper that describes an observationalepidemiological study in your field of specialty a The outcome variables should be either categorical or survival data a The methods should be either logistic regression or survival regression model Give a class presentation for 1015 minutes to talk about What was the purpose of the study What was the study design Was the study design appropriate to address the objective of the study What were the statistical methods used What were the outcome variables predictors and confounders Were the statistical models appropriate for the data collected for the study Is a clinical trial feasible or more suitable for the type of study If so are there any clinical trials that have been carried out DDDUDUD Note You will need to choose a date from either 61 or 68 to present your journal paper Team works are allowed EPP 2477Biostatistical Methods in 462007 Epidemiology 20 References I MacHahon et al Age at first birth and breast cancer risk Bull World Health Organ 1970432209 21 Knowler WC et al Gm351314 and type 2 diabetes mellitus an association in American Indians with genetic admixture Am J Hum Genet 19884345206 Schildkraut JM et al Coronary risk associated with age and sex of parental heart disease in the Framingham Study Am J Cardiol 198964105559 Schildkraut JM et al Evaluating genetic association among ovarian breast and endometrial cancer evidence for a breastovarian cancer relationship Am J Hum Genet 19894545219 Schildkraut JM Thompson WD Familial ovarian cancer a populationbased casecontrol study Am J Epidemiol 1988128345666 EPP 2477Biostatistical Methods in 462007 Epidemiology 21
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'