logging in or signing up Risk factors for TB Brian Williams Wanderer Download Post to : URL : Related Presentations : Let's Connect Share Add to Flag Embed Email Send to Blogs and Networks Add to Channel Copy embed code: Embed: Flash iPad Dynamic Copy Does not support media & animations Automatically changes to Flash or non-Flash embed WordPress Embed Customize Embed URL: Copy Thumbnail: Copy The presentation is successfully added In Your Favorites. Views: 681 Category: Entertainment License: All Rights Reserved Like it (0) Dislike it (0) Added: October 26, 2007 This Presentation is Public Favorites: 0 Presentation Description No description available. Comments Posting comment... Premium member Presentation Transcript Risk factors for disease: Risk factors for disease Brian Williams, STB/TMESlide2: A custom loathsome to the eye, hateful to the nose, harmful to the brain, dangerous to the lungs, and in the black stinking fume thereof, nearest resembling the horrible Stygian smoke of the pit that is bottomless. James I and VI: A Counterblast to Tobacco 1604Slide3: Risk factors for Tuberculosis Infection, disease, death, relapse, drug-resistance, treatment failure Biological: genetic, diabetes, HIV, vitamin D deficiency Behavioural: smoking, alcoholism, drug addiction Social: housing, poverty, health workers, institutional Environmental: silica, asbestos, solid fuelSlide4: Relative risk Excess risk Population attributable fraction Prevalence of the risk factor P: in the population F: among those with disease Slide5: Smoking and lung cancer 1 Sample: 34k male, British doctors recruited in 1951 and followed up for 40 years. Doll, R., Peto, R., et al. Mortality in relation to smoking: 40 years' observations on male British doctors British Medical Journal (1994) 309 901-11. Death from: cancers of the mouth, oesophagus, pharynx, larynx, lung, pancreas, and bladder; COPD, other respiratory diseases; vascular diseases; peptic ulcer; cirrhosis, suicide, and poisoning (perhaps because of confounding by personality and alcohol use). Negative association with death from Parkinson's disease. Stopping smoking before middle age avoided almost all of the excess risk.Slide6: Case-Control Studies In a cohort study we choose some smokers and some non-smokers. Then we wait a very long time and see which of them gets TB. But what if we choose some people with TB and some without TB (which we can easily do) and then see how many in each group smoke? Need to match the controls for age, gender and anything else that might affect their chance getting TB. But can we do the analysis?Slide7: Exposed Non-exposed Risk Odds Sick a b a/(a+b) a/b Healthy c d c/(c+d) c/d Risk a/(a+c) b/(b+d) Odds a/c b/d Odds and Odds-ratios OR(Sick) = ad/bc = OR(Exposed) Independent of the number that are sick or healthy RR(sick) = OR if a and b are small RR(sick) = RR(Exposed) =Slide8: 1. The odds ratio is a measure of association and is independent of which way round it is measured We want to know the odds for getting lung cancer if you smoke. But lung cancer is a rare disease. So we measure the odds for smoking if you have lung cancer instead and the two are the same. What's so special about odds? The relationship between the dose and the response often follows a logistic curve: But if you plot the log-odds against the value of the risk factor you get a straight line: 1 P 0 0 2 4 6 8 10 R ln(P/Q) 0 2 4 6 8 10 R 2. A logistic regression gives you odds-ratiosSlide9: Smoking and lung cancer 2 1357 lung cancer patients and as many controls. Doll, R. & Hill, A. B. Smoking and carcinoma of the lung British Medical Journal (1950) 2 739. Establishes an association; not causality. Gives us the relative risk compared to the control group. Need to measure the prevalence of the outcome and the prevalence of the risk factor. Because the controls are matched to the cases on possible confounders we can say nothing about any of them.Slide10: Smoking and TB in Chennai 27k deaths and 16k controls, 1994-1997. 2k TB deaths Gajalakshmi, V., Peto, R., et al. Smoking and mortality from tuberculosis and other diseases in India: retrospective study of 43000 adult male deaths and 35000 controls Lancet (2003) 362 507–515. Smoker Non-smoker Odds TB deaths 1454 386 3.76 Controls 6430 10058 0.64 OR = 5.9 (4.5) F = 0.79 60% of all TB deaths among men in Chennai are attributable to smokingSlide11: Silicosis and HIV in South African gold mines Logistic regression: HIV, calendar period, silicosis grade, age, duration of employment, previous TB Corbett, E. L. et al. HIV infection and silicosis: the impact of two potent risk factors on the incidence of mycobacterial disease in South African miners AIDS (2000) 14 2759-2768. Relative Risk Incidence(%/yr) F PAF Neither 1 0.7 Silicosis 3.3 2.3 0.18 0.12 HIV 5.9 4.1 0.37 0.31 Both 18 13.6 0.27 0.25 Incidence of TB (per cent per year) Because logistic regression uses a log-odds transform it always assumes that the effects are multiplicative.Slide12: 20-44 45-64 65-90 Total Incidence rate ratio Age (years) Ponce-De-Leon, A., de Lourdes Garcia-Garcia, L., et al. Tuberculosis and diabetes in southern Mexico Diabetes Care (2004) 27 1584-90. Incidence Rate Ratio for TB in diabetics and non-diabetics Numerator: number of TB patients with or without diabetes; Denominator: no of diabetics and non-diabetics in the study area PAF = 25%Slide13: TB notification rates per 100k per year in Kisumu, Kenya TB HIV Year HIV prevalence (%) TB notifications (%/yr) Age, sex and HIV Age (years) Age (years) 1990-1993 1994-1997 1997-2001 Notifications/100k/yrSlide14: Risk factors for TB deaths in Orel, Russia Sample: medical records, autopsy reports, TB registry Cases: patients who died within 8 months after treatment initiation Controls: patients who survived 8 months Unemployed: 87% of cases, 44% of controls OR = (0.87/0.13)/(0.44/0.54) = 8.5. PAF = 7.5/8.5×0.87 = 77% Alcoholism: 68% of cases, 41% of controls OR = (0.68/0.32)/(0.41/0.59) = 10.6. PAF = 9.6/10.6×0.68 = 80% Dewan et al. Risk factors for death during tuberculosis treatment in Orel, Russia. International Journal of Tuberculosis and Lung Disease (2004) 8 598-602. Social and behavioural factorsSlide15: TB risk among health care workers in Samara, Russia Notification rate among health care workers ~7.5 Notification rate among general population Dimitrova et al. Increased risk of tuberculosis among health care workers in Samara oblast, Russia: analysis of notification data. International Journal of TB and Lung Disease 2004;9(1):43-8. OccupationalSlide16: Ecological studies Cross-sectional studies Retrospective cohort study Prospective cohort study Case control study Randomized controlled trial Statistically stronger More limited answers Statistically weaker Broader answersSlide17: Ecological studies Cross-sectional studies Retrospective cohort study Prospective cohort study Case control study Randomized controlled trial Statistically stronger More limited answers Statistically weaker Broader answers Diabetes in Mexico: Numerator and denominator from different populations. Endless possibilities for confounding. But the effect is substantial and the result is suggestive.Slide18: Ecological studies Cross-sectional studies Retrospective cohort study Prospective cohort study Randomized controlled trial Case control study Statistically stronger More limited answers Statistically weaker Broader answers Unemployment in Russia: OR = 8.5; PAF = 77%. But it may all be confounded by alcoholism.Slide19: Ecological studies Cross-sectional studies Retrospective cohort study Prospective cohort study Case control study Randomized controlled trial Statistically stronger More limited answers Statistically weaker Broader answers Corbett, E. L., Charalambous, S., et al. Stable incidence rates of tuberculosis (TB) among human immunodeficiency virus (HIV)-negative South African gold miners during a decade of epidemic HIV-associated TB Journal of Infectious Diseases (2003) 188 1156-63. HIV and TB incidence in HIV-negative miners: IRR constant over time.Slide20: Ecological studies Cross-sectional studies Retrospective cohort study Prospective cohort study Case control study Randomized controlled trial Statistically stronger More limited answers Statistically weaker Broader answers Smoking and lung cancer: Many diseases, age, stopping smoking, causal, very big.Slide21: Ecological studies Cross-sectional studies Retrospective cohort study Prospective cohort study Case control study Randomized controlled trial Statistically stronger More limited answers Statistically weaker Broader answers Smoking and lung cancer: Clear demonstration of one relationship. But does not demonstrate causality. Controls need to be chosen very carefully.Slide22: Ecological studies Cross-sectional studies Retrospective cohort study Prospective cohort study Case control study Randomized controlled trial Statistically stronger More focussed answers Statistically weaker Broader answers Comparing drugs or vaccines: Stratify and then randomise. Limited, focussed question.Slide23: Max Born I believe that there is no philosophical high-road in science with epistemological sign-posts. No, we are in a jungle and find our way by trial and error, building our road behind us as we proceed. We do not find sign-posts at cross-roads, but our own scouts erect them to help the rest. … My advice to those who wish to learn the art of scientific prophecy is not to rely on abstract reason, but to decipher the secret language of Nature from Nature’s documents, the facts of experience. You do not have the permission to view this presentation. In order to view it, please contact the author of the presentation.