OBSERVERENDE UNDERSØGELSER Kim Overvad Institut for Epidemiologi og Socialmedicin Aarhus Universitet Forår 2002
Epidemiologisk design Observerende undersøgelser beskrivende: Undersøgelsesenheden er populationer regional variation migrationsundersøgelser korrelationsundersøgelser tidsrækker analytiske: Undersøgelsesenheden er individer tværsnitsundersøgelser follow-up undersøgelser case-kontrol undersøgelser Eksperimentelle undersøgelser
Korrelationsundersøgelser Biologi rationale tid Ikke nødvendigvis personsammenfald niveau fejlslutning Confounding variation i andre risikoindikatorer
Populationserfaring Eksponering Syge Raske I alt Obs. Tid + a b n + t + - c d n - t - I alt a+c b+d N T Sampling tværsnitsundersøgelse N follow-up undersøgelse eksponerede og ikke-eksponerede (n + og n - ) case-kontrol undersøgelse oftest syge og raske (a + c og b + d)
Cross Sectional Study, 1 A cross-sectional study was carried out among women attending a university health service to investigate the determinants of cervical human papillomavirus (HPV) infection. A sample of 467 women were asked to complete a self-administered questionaire on socio-demographic variables and sexual behaviour at the time of their visit to the clinic. The prevalence of HPV infection was then examined in relation to marital status and lifetime number of male sexual partners (Ley et al., 1991).
Cross Sectional Study, 2 Lifetime no. of sexual partners No. of women % positive for HPV (prevalence proportion) Prevalence proportion ratio (95% CI) 1 90 21.1 1.0 2-3 101 32.7 1.5 (0.9-2.4) 4-5 93 54.8 2.6 (1.7-4.0) 6-9 66 56.1 2.7 (1.7-4.3) 10+ 102 68.6 3.3 (2.1-4.9)
Cross Sectional Study, 3 The prevalence proportion ratio for each exposure level was calculated by forming 2 x 2 tables as illustrated below for women with 10+ partners Lifetime no of HPV- HPV- No of sexual partners negative positive women 1 71 19 90 10+ 32 70 102 Prevalence proportion among women with 10+ partners = 70/102 = 68.6% Prevalence proportion among women with one partner = 19/90 = 21.1% Prevalence proportion ratio = 68.6% / 21.1% = 3.3
Follow-up Study, 1 The Nurses Health Study is a cohort study of 121,700 US Female registered nurses aged 30-55 years when the cohort was established in mid-1976. A total of 1,799 newly diagnosed breast cancer cases were identified during the first 10 years of follow-up from mid-1976 to mid-1986. Analyses were then conducted to investigate the relationship between oral contraceptive use and risk of breast cancer. (Romieu et al., 1989).
Follow-up Study, 2 The incidence rates of breast cancer among nurses aged 45-49 years at the time of their entry into the cohort was examined in relation to use of oral contraceptives Oral contraceptive use Cases Person-years at risk Incidence rate per 100,000 pyrs Ever (current or past use) 204 94,029 217 Never 240 128,528 187 Total 444 222.557 199 Data from Romieu et al. (1989) Rate ratio = 217 per 100,000 pyrs/187 per 100,000 pyrs = 1.16 95% confidence for the rate ratio = 0.96 to 1.40 Rate difference = 217 per 100,000 pyrs 187 per 100,000 pyrs = 30 per 100,000 pyrs 95% confidence interval for the rate difference = -8 to 68 per 100,000 pyrs
Fordele ved observerende follow-up undersøgelser Fordele direkte identifikation af studiebasen indflydelse på eksponeringsfordelingen kvalitetskontrol af data eksponeringsoplysninger før sygdom flere sygdomsudfald mulighed for beregning af flere associationsmål
Ulemper ved observerende follow-up undersøgelser Ulemper ressourcer tid penge enkelt eksponeringer påvirkning af den diagnostiske proces eller sygdomsregistreringen etik
Oplysninger om eksponering Hypoteser Muligheder specifik (tid og sted) gennemsnit max/min kumuleret
Associationsmål Follow-up undersøgelse, komplet follow-up Eksponering Antal pers. Syge Obs. Tid + n + a t + - n- c t- Relativt associationsmål relativ risiko RR = a/n + / c/n - Absolut associationsmål risiko differens RD = a/n + c/n -
Associationsmål Follow-up undersøgelse, inkomplet follow-up Eksponering Antal pers. Syge Obs. Tid + n + a t + - n - c t - Relativt associationsmål incidens rate ratio IRR = a/t + / c/t - Absolut associationsmål incidensrate differens IRD = a/t + c/t -
Valg af associationsmål i follow-up undersøgelsen Sygdomshyppigheden blandt ikke-eksponerede, et relativt og et absolut associationsmål supplerer hinanden.
Follow-up undersøgelse Eksponering Syge Raske I alt Obs.tid + a b n + t + - c d n - t - Ved sjældne sygdomme er n + og n - samt b og d meget større end a og c
Case-Control Study, 1 A population-based case-control study was carried out in Spain and Colombia to assess the relationship between cervical cancer and exposure to human papillomavirus (HPV), selected aspects of sexual and reproductive behaviour, use of oral contraceptives, screening practices, smoking, and possible interactions between them. The study included 436 incident cases of histologically confirmed invasive squamous-cell carcinoma of the cervix and 387 control of similar age randomly selected from the general Population that generated the cases (Muñoz et al., 1992a).
Case-Control Study, 2 Risk of developing cervical cancer in relation to the lifetime number of sexual partners. Number of sexual partners Cervical cancer cases Controls Odds ratio (95% confidence interval) 0-1 265 305 1.0 2-5 125 74 1.94 (1.39-2.70) 6+ 46 8 6.62 (3.07-14.27)
Case-Control Study, 3 The odds ratios for each category of exposure were calculated in the following way: Number of sexual partners 0-1 (unexposed) 2-5 (exposed) Cervical cancer cases 265 125 Control 305 74 Odds ratio = (125/265) / (74 x 305) = 1.94
Fordele ved case-kontrol undersøgelser Fordele ressourcer tid penge flere eksponeringer hypotese genererende etik
Ulemper ved case-kontrol undersøgelser Ulemper afgrænsning af studiebasen kvalitetskontrol af data eksponering effekt enkelt sygdomme ikke økonomisk ved studiet af sjældne eksponeringer associationsmål
Studiebasen Studiebasen er den populationserfaring, vi ønsker at beskrive på baggrund af undersøgelsen Studiebasen er de personer, for hvem det gælder, at de ville blive cases, givet de udviklede den sygdom, der studeres
Eksponering Eksponeringssandsynlighed Muligheder for at øge eksponeringssandsynligheden restriktion geografi køn alder andre specielle studiebaser
Associationsmål Case-kontrol undersøgelse, komplet follow-up Eksponering Syge Raske I alt Obs. Tid + a b n + t + - c d n - t - Stikprøve af raske case non-case design Associationsmål OR = a/c / b/d OR ~ RR sygdommen sjælden RR tæt ved 1,0
Associationsmål Case-kontrol undersøgelse, komplet follow-up Eksponering + Syge a Raske - c d n - Stikprøve af studiebase personer i alt case-base design kumulativ incidens proportions sampling b I alt n + Obs.tid t + t - Associationsmål OR ~ RR uafhængig af sygdomshyppighed uafhængig af størrelsen af RR
Associationsmål Case-kontrol undersøgelse, komplet follow-up Eksponering Syge Raske I alt Obs.tid + a b n + t + - c d n - t - Estimerings- og vurderingsproblemer cases i referencegruppen statistiske metoder Referencegruppen kan bruges i flere undersøgelser med forskellige endpoints
Associationsmål Case-kontrol undersøgelse Inkomplet follow-up (censureringer) Eksponering Syge Raske I alt Obs. Tid + a b n + t + - c d n - t - Stikprøve af studiebasen observationstid case-base design incidence density sampling hver gang en case opstår samples reference personer fra den aktuelle studiebase
Eksponering + - Associationsmål Case-kontrol undersøgelse Inkomplet follow-up (censureringer) Syge a c Raske b d I alt n + n - Obs. Tid Associationsmål OR ~ IRR uafhængig af sygdomshyppighed og størrelse af IRR Estimerings- og vurderingsproblemer cases i referencegruppen statistiske metoder Referencegruppen er specifik for den pågældende undersøgelse t + t -
Case-kontrol undersøgelse i follow-up undersøgelse Ikke økonomisk at gennemføre en follow-up undersøgelse, men heller ikke mulig at indhente relevante eksponeringsoplysninger på det tidspunkt, hvor cases og referencepersoner kan identificeres. N-3 fedtsyrer i 30. graviditetsuge