Bias Tunis, 30th October 2014 Dr Sybille Rehmet Sources: T. Grein, M. Valenciano, A. Bosman Dr Sybille Rehmet

By the end of the lecture, you will be able to Define bias Identify different types of bias Explain how bias affects risk estimates Critique study designs for bias Develop strategies to minimise bias

If an association is observed, the first question asked must always be ... “Is it real?”

Should I believe the estimated effect? Gastroenteritis Mayonnaise RR = 4.3 True association? causal? non-causal? Chance? Confounding? Bias?

Definition of bias Any systematic error in the design or conduct of an epidemiological study resulting in a conclusion which is different from the truth

Errors in epidemiological studies Random error (chance) Systematic error (bias) Sample size Source: Rothman, 2002

Main sources of bias Selection bias Preferential selection or participation of subjects into the study according to exposure or outcome status Information bias Systematic error in the measurement of information on exposure or outcome

Selection bias association between exposure and disease differs between those who participate and those who don’t Selection of study subjects Allocation to study groups Factors affecting study participation

Selection of controls Association of alcohol intake and cirrhosis, hospital-based case control study OR = 6 How representative are hospitalised trauma patients of the population which gave rise to the cases?

Selection of controls OR = 6 OR = 36 Higher proportion of controls drinking alcohol in trauma ward than in non-trauma ward

e.g. association between haemorrhagic fever and stay in hospital Survival bias Only survivors of a highly lethal disease enter study e.g. association between haemorrhagic fever and stay in hospital This bias occurs when survivors of a highly lethal disease are more likely to enter a study than other cases. Exposure to hospital leads to death Underestimation of “a”  underestimation of OR

Diagnostic bias Oral contraceptives and uterine cancer OC use  can cause breakthrough bleeding  increased chance of testing & detecting uterine cancer Overestimation of “a”  overestimation of OR

Non-response bias Cohort study on lung cancer and smoking Scenario 1: all exposed and unexposed participate in the study Lung cancer No lung cancer Total Rate/1000 RR smoker 90 910 1000 9 non-smoker 10 990 reference

Non-response bias Scenario 2: Non-response among exposed (only 10% respond) Lung cancer No lung cancer Total Rate/1000 RR smoker 9 91 100 non-smoker 10 990 1000 1 reference

Non-response bias Scenario 3: Non-response among exposed cases (only 10% respond) Lung cancer No lung cancer Total Rate/1000 RR smoker 9 910 919 10 1 non-smoker 990 1000 reference

Loss to follow-up Difference in completeness of follow-up between comparison groups, esp. important in prospective cohort studies Study of risk of severe disease in migrants Migrants might be more likely to return to their place of origin when having a severe disease  lost to follow-up  lower disease rate among exposed (=migrant)

Not everything that looks like selection bias is really bias… Look around in the room – just by eyeballing, is there anything that might be a bias?? We count 9 men and 33 women – gender-bias??? How do we know? Is this distribution representative of the underlying population, or do we have selection bias?? What IS the underlying population? General population? Medical doctors? Public health doctors? Epidemiologists? …you have to thoroughly assess!

Information bias Systematic error in the measurement or classification of participants in a study due to inaccurate information on exposure or outcome Can be related to reporting or the observer

Information bias When? How? Consequences? During data collection Differences in accuracy of exposure data between cases and controls of outcome data between exposed and unexposed Consequences? Missclassification – study subjects are classified in the wrong category

Information bias: misclassification Measurement error leads to assigning wrong exposure or outcome category Non-differential Random error Missclassifcation of exposure SIMILAR between cases and controls Missclassification of outcome SIMILAR between exposed & nonexposed => Weakens measure of association Differential Systematic error Missclassification of exposure DIFFERS between cases and controls Missclassification of outcome DIFFERS between exposed & nonexposed => Measure of association distorted in any direction

Non-differential misclassification Cohort study: Alcohol  laryngeal cancer No misclassification exposure Total population cases Incidence/1.000.000 RR drinkers 1.000.000 50 5 Non-drinkers 500.000 10 50% of drinkers misclassified exposure Total population cases Incidence/1.000.000 RR drinkers 500.000 25 50 1.7 Non-drinkers 1.000.000 30

Cases remember exposure differently than controls Recall bias Cases remember exposure differently than controls Overestimation of “a”  overestimation of OR Mothers of children with fetal malformations remember past exposures better than mothers of healthy children

Cases report exposure differently than controls Prevarication bias Cases report exposure differently than controls e.g. isolation and heat related death Relatives of dead elderly may deny isolation Underestimation “a”  underestimation of OR

Interviewer bias Investigator asks cases and controls differently about exposure e.g: soft cheese and listeriosis Cases of Controls listeriosis Overestimation of “a”  overestimation of OR Eats soft cheese a b Does not eat c d soft cheese Investigator may probe listeriosis cases about consumption of soft cheese

Example – your evaluation form One group fills in the evaluation every day as requested One group fills in the evaluation form as a whole on the weekend because they did not have time before Will this influence the evaluation results for Monday? And if yes, why? What type of bias might occur?

How to prevent or minimize bias?

be aware of the potential of bias, Be creative in foreseeing potential biases, and be critical with your own study design!

Start at the beginning… Consider carefully your choice of study design and your protocol to minimize the chance for bias Clearly define study population, cases and exposures Choose the right comparison group Case Control studies: Cases and controls from same population to have the same possibility of exposure Cohort studies: selection of exposed and non-exposed without knowing disease status Randomization in intervention studies

…then keep going … Set up strict guidelines for data collection Standardized procedures, questionnaires, training for interviewers etc Administer instruments equally to cases and controls/ exposed and unexposed Use multiple sources of information Use direct measurements, registries, case records etc

…and if you could not completely avoid it … address it!!!! Discuss your potential bias in your study protocol Estimate the direction and magnitude of the distortion Take it into account when you interpret your results Discuss it in your report and publication

https://wiki.ecdc.europa.eu/fem/default.aspx

Thank you! Sybille Rehmet sybillerehmet@yahoo.com

Some more examples What potential bias could have been introduced if you found out that those who interviewed cases took 30 minutes longer on average than those who interviewed controls? Selection bias Information bias Volunteer bias Loss to follow-up http://epiville.ccnmtl.columbia.edu/bias/discussion_question.html

Healthy worker effect Source: Rothman, 2002

Healthy worker effect Source: Rothman, 2002

Differential misclassification OR = ad/bc = 3.0

What effect (if any) would you expect if the interviewers were aware of the disease status of the study subjects? It would benefit the validity of the results since the interviewers would understand more precisely how the exposure is related to the disease and collect better data for the cases. The results would likely not change. It could damage the validity of the results by introducing interviewer bias.

Volunteer or self-selection bias You want to determine the prevalence of HIV infection. You ask for volunteers for testing. You find no HIV. Is it correct to conclude that there is no HIV in this location? Volunteer or self-selection bias

Selection bias in case-control studies Risk factors for menstrual toxic shock syndrome: results of a multistate case-control study. Reingold AL et al. Rev Infect Dis. 1989 Jan-Feb;11 Suppl 1:S35-41,32 For assessment of current risk factors for developing toxic shock syndrome (TSS) during menstruation, a case-control study was performed. Cases with onset between 1 January 1986 and 30 June 1987 were ascertained in six study areas with active surveillance for TSS. Age-matched controls were selected from among each patient's friends and women with the same telephone area code. Of 118 eligible patients, 108 were enrolled, as were 185 "friend controls" and 187 telephone area code-matched controls

Selection bias in case-control studies Risk factors for menstrual toxic shock syndrome: results of a multistate case-control study Results for tampon use as a risk factor: OR when both control groups were combined = 29 OR when friend controls were used = 19 OR when neighborhood controls were used = 48 Why did use of friend controls produce a lower OR? Friend controls were more likely to have used tampons than were neighborhood controls (71% vs. 60%) Reingold AL et al. Rev Infect Dis. 1989 Jan-Feb;11 Suppl 1:S35-41

Bias in randomised controlled trials Gold-standard: randomised, placebo-controlled, double-blinded study Least biased Exposure randomly allocated to subjects - minimises selection bias Masking of exposure status in subjects and study staff - minimises information bias

Bias in prospective cohort studies Loss to follow up The major source of bias in cohort studies Assume that all do / do not develop outcome? Ascertainment and interviewer bias Some concern: Knowing exposure may influence how outcome determined Non-response, refusals Little concern: Bias arises only if related to both exposure and outcome Recall bias No problem: Exposure determined at time of enrolment

Bias in retrospective cohort & case-control studies Ascertainment bias, participation bias, interviewer bias Exposure and disease have already occurred  differential selection / interviewing of compared groups possible Recall bias Cases (or ill) may remember exposures differently than controls (or healthy)