Epidemiology A branch of science that investigates the
Epidemiology A branch of science that investigates the frequency and distribution of diseases in a defined population in an attempt to determine their causes, to discover ways to alleviate them, and to prevent their reoccurrences 1 © 2006
Features of epidemiologic studies • Observe people in their natural setting – Over a specific period of time – At one point in time – Retrospectively • The objective is to describe specific traits that may be present among members of a population Evidence-based Chiropractic 2 © 2007
Risk factors a. k. a. , predictor variables • Outcomes of groups of subjects are compared – One group was exposed and the other was not exposed to some risk factor – Did the exposed group develop a higher incidence of the disease • Valuable type of study because it is unethical for researchers to ask patients to do or take something known to be harmful Evidence-based Chiropractic 3 © 2007
Epidemiologic studies considered observational • Data are collected via surveys, review of medical records, etc. • Data analyzed using statistical tests designed to identify patterns and correlations – e. g. , people who lift heavy at work are more likely to develop low back pain Evidence-based Chiropractic 4 © 2007
The epidemiologic approach • Its focus is on prevention rather than treatment • It deals with populations rather than individual patients • The approach is to identify subgroups that are at high-risk of developing a disease and then find out what factors caused persons in the subgroups to be at high-risk Evidence-based Chiropractic 5 © 2007
The epidemiologic approach (cont. ) • Preventive measures can be developed to minimize risk factors in persons found to be at high-risk – e. g. , teaching construction workers to lift properly to prevent low back injuries • The effectiveness of the preventive measures can be monitored using epidemiologic methods Evidence-based Chiropractic 6 © 2007
Epidemiologic studies • Preferred to RCTs when questions are about diagnosis, prognosis or causation – Related to ethical concerns and feasibility • Example studies – The link between driving heavy equipment and the incidence of lower back pain – Persons who consume enough calcium and vitamin D are less likely to develop osteoporosis Evidence-based Chiropractic 7 © 2007
Epidemiologic studies (cont. ) • Two groups are selected – One group is exposed to some agent or event, while the other is not – Any difference in the rate of disease between the groups may be because of the exposure • Statistical tests are used to determine the probability that other persons exposed to the same risk factors would develop the disease Evidence-based Chiropractic 8 © 2007
More potential for bias in epidemiologic studies • Subjects are not randomly assigned to be in exposure versus no-exposure groups • Differences may exist in the baseline risk of disease between the groups • Random selection helps reduce the influence of many biases • Randomly selected samples are much more representative of the population Evidence-based Chiropractic 9 © 2007
Evidence-based Chiropractic 10 © 2007
Reliability of data sources • The validity of epidemiologic studies depends to a large extent on the reliability of the data that is collected • Often collected directly from people or by relying on others – e. g. , mailed surveys, hospital emergency room personnel, physicians, etc. – These are all potentially unreliable sources Evidence-based Chiropractic 11 © 2007
Reliability of data sources (cont. ) • Questionnaires need to be properly constructed to ensure accurate information – No “double-barreled” questions – No leading questions – Respondents should be competent to answer the questions • The next slide is an example of a biased questionnaire from the Canadian Stroke Consortium – SPONTADS Study Evidence-based Chiropractic 12 © 2007
There are many other daily activities that have been reported to be associated with cervical artery dissection. Why weren’t they listed as possible choices? Evidence-based Chiropractic 13 © 2007
Canadian Stroke Consortium results were biased • Reported that 28% of the cervical artery dissections in their group were the result of manipulation • This is significantly higher than what other authors have reported – Approximately 6% between 1994 and 2003 • Cervical trauma/manipulation as the only choice called for a highly biased response Evidence-based Chiropractic 14 © 2007
15 © 2006
Face to-face interviews • The potential for bias due to leading questions is even higher • Interviewer facial expressions, body language, and vocal inflections can influence responses Evidence-based Chiropractic 16 © 2007
Respondents should be competent to answer questions • Includes children, persons with dementia, etc. • Also automobile crash victims – They are typically quite upset and may be confused – Mild traumatic brain injury is fairly common following car crashes which may result in confusion Evidence-based Chiropractic 17 © 2007
The best measures for epidemiologic research • Are as free from error as possible • Inexpensive, readily available, easy to use, etc. • Ethically acceptable • The resulting information is important • Change in the predictor variable is linked to change in the outcome variable Evidence-based Chiropractic 18 © 2007
Common designs used in epidemiologic research • • Cross-sectional studies Case-control studies Cohort studies Which one to use depends on – The frequency of the disease or condition – The availability of human and economic resources Evidence-based Chiropractic 19 © 2007
Cross-sectional studies a. k. a. , prevalence studies • Assess the health status and exposure levels of persons in a population at one point in time • Cases must actively manifest the disease to be included • Cases with developing conditions that have not yet been diagnosed are not counted Evidence-based Chiropractic 20 © 2007
Cross-sectional study (cont. ) • Purpose – To determine if there is an association between a suspected causal factor and a condition • They are useful to discover associations, but incapable of determining if one factor caused the other • Case-control or cohort studies are often used to verify their results Evidence-based Chiropractic 21 © 2007
Cross-sectional study (cont. ) • They are attractive to researchers because they are relatively quick and easy to carry out; and are inexpensive – However, it may take a long time to gather all of the required information • Consequently, they are often the initial research tools used to investigate exposures to risk factors and their relationships to disease Evidence-based Chiropractic 22 © 2007
Cross-sectional study design At one point in time Exposure levels Health status Evidence-based Chiropractic 23 © 2007
Case-control studies • A study that starts off by identifying two groups of subjects – One group has a disease or condition (cases) – The other group is free from the disease (controls) • Prior exposures of the cases are compared with those of the controls to see if the exposures influenced the odds of developing the disease Evidence-based Chiropractic 24 © 2007
Subjects in case-control studies • Controls should be as similar to cases as is possible • Cases and controls are normally matched so that they are as alike as possible – Regarding variables such as age, gender, weight, occupation, etc. • Except for the presence of the disease under investigation Evidence-based Chiropractic 25 © 2007
Case-control study design Evidence-based Chiropractic 26 © 2007
Case-control studies are retrospective • Exposure levels are determined by looking back in time before the person became a case or a control subject • Because they are retrospective – Cannot determine the risk of developing a disease – But can estimate the odds of developing a disease given that a person was exposed to a risk factor; the odds ratio (OR) Evidence-based Chiropractic 27 © 2007
Odds ratio (OR) • A ratio of the odds of exposure among cases divided by the odds of exposure among controls Evidence-based Chiropractic 28 © 2007
2 X 2 contingency table used to calculate OR Cases Controls Exposed a b Not Exposed c d a/c Odds of exposure among cases OR = = b/d Odds of exposure among controls May also be expressed ad/bc Evidence-based Chiropractic 29 © 2007
Interpretation of ORs • OR = 1 – No association between an exposure and a disease • OR > 1 – A positive association between an exposure and a disease – The exposure is considered to be harmful • OR < 1 – A negative association between an exposure and a disease – The exposure is protective Evidence-based Chiropractic 30 © 2007
Bias in case-control studies • This design is vulnerable to a number of biases because cases and controls are selected after the disease outcome and the exposure have occurred • Recall bias – Systematic differences between the way cases and controls recall past exposures Evidence-based Chiropractic 31 © 2007
Bias in case-control studies (cont. ) • Berkson’s bias (a. k. a, admission rate bias) – In a hospital setting, cases and controls are systematically different from each other – This is related to cases being more prone to hospital admission simply because they have a higher rate of exposure and incidence of the disease – It is actually a type of selection bias Evidence-based Chiropractic 32 © 2007
Case-control studies and rare diseases • Case-control studies are usually the best way to study rare diseases – Definitive prospective studies typically require many subjects, which is not feasible if the disease is rare – It is much easier to select a group of subjects who already have a rare disease and then look back into their histories for clues Evidence-based Chiropractic 33 © 2007
Advantages of case-control studies • Good for investigating rare diseases • Can be performed quickly and inexpensively • Useful for studying diseases with long latency period between exposure and manifestation • Facilitates the study of multiple potential causes at once • Existing records can often be used Evidence-based Chiropractic 34 © 2007
Disadvantages of case-control studies • Typically rely on patients’ recall of past exposures • Does not permit calculation of true disease rates in the population • Difficult to validate information on exposure • Other variables that may be associated with the disease are not controlled Evidence-based Chiropractic 35 © 2007
Confounding in case-control studies • Studies have shown that coffee drinking is positively associated with lung cancer • However, coffee drinkers are more likely to be smokers • When considered separately, there is no relationship between coffee drinking and lung cancer – Smoking is the culprit Evidence-based Chiropractic 36 © 2007
Confounding in case-control studies (cont. ) • Stratified analysis – The effect each of the independent variables on the outcome is analyzed separately – Commonly used to distinguish confounding variables from true exposure-disease associations – Data is divided into strata of homogenous subgroups to see if an association observed in the undivided group data remains Evidence-based Chiropractic 37 © 2007
Categories of case-control studies • Prevalent case – All persons with the disease during the observation period are included in the study – For example, all low back pain patients at a chiropractic clinic between Jan. 1 and Dec. 31 • Cumulative incidence – Only new (incident) cases that occur during the period of observation are selected Evidence-based Chiropractic 38 © 2007
Cohort studies a. k. a. , longitudinal studies • Studies that follow groups of subjects forward in time and compare their outcomes • One group is exposed to a known or suspected cause of disease while the other group is not exposed – Or one group was exposed to a risk factor in the past Evidence-based Chiropractic 39 © 2007
Cohort studies (cont. ) • A population is identified, from which two or more groups of subjects are selected • The groups are called cohorts • Outcomes are assessed on both groups prospectively over a defined time period – The objective is to determine if there are differences of disease occurrence between the groups Evidence-based Chiropractic 40 © 2007
Cohort study design One cohort is exposed to a risk factor, while the other is not A population is identified Evidence-based Chiropractic 41 © 2007
They have major advantage over case-control studies • Able to establish the temporal precedence of an exposure in relation to a health outcome – Thus, they satisfy one of the main prerequisites in determining causation • They are much less subject to bias because the exposure level is evaluated before disease develops Evidence-based Chiropractic 42 © 2007
Advantages over case-control studies (cont. ) • The best design to determine risk of a harmful substance • Still considered observational because random assignment to groups is not used • More useful investigating common diseases because of the complexity and expense of assembling cohorts that are large enough to investigate rare diseases Evidence-based Chiropractic 43 © 2007
Advantages over RCTs • Cohort studies are easier to administer and less costly than RCTs • Usually more acceptable than RCTs from an ethical perspective – Potentially harmful treatments are not utilized – Treatment is not withheld from one of the groups Evidence-based Chiropractic 44 © 2007
Advantages of cohort studies • Portrays the natural history of disease • Does not rely on patient recall • Better to establish a cause-effect relationship than case-control • Less vulnerability to bias or chance • Permits calculation of true disease rates in the population Evidence-based Chiropractic 45 © 2007
Disadvantages of cohort studies • Typically very expensive • Many people must be followed to obtain enough with the disease • Very time consuming, given that one must wait for people to develop the disease • Subjects frequently drop out in time • Difficult to generate a control group to study very common conditions Evidence-based Chiropractic 46 © 2007
Cohort studies are capable of determining risk • Relative risk (RR) is calculated in cohort studies • RR is the probability of disease in the exposed group, divided by the probability of disease in the unexposed group Evidence-based Chiropractic 47 © 2007
RR is calculated using a 2 X 2 contingency table Disease Yes No Exposed a b Not Exposed c d a/a+b Probability of disease in exposed RR = = c/c+d Probability of disease in unexposed Evidence-based Chiropractic 48 © 2007
Interpretation of relative risk • If RR is different from one, the risk factor is considered to be associated with the risk of disease – RR>1, the association is positive – RR<1, there is a negative or protective association • If RR=1, there is no association Evidence-based Chiropractic 49 © 2007
Relative Risk example • Cohort study where 50 subjects exposed to dust (work in a coal mine) and 50 subjects work in offices – followed for 20 years – Exposed: 20/50 develop lung cancer – Not exposed: 5/50 develop the condition – RR=4 • Exposed were 4 times more likely to develop lung cancer Evidence-based Chiropractic 50 © 2007
Lung cancer example Lung cancer Coal dust exposure Present Evidence-based Chiropractic Present Absent 20 30 5 45 51 © 2007
RR example 2 • Brandt et al conducted a 1 year prospective study of computer users and reported that the RR for new neck pain was – 1. 8 for keyboard use ≥ 15 hrs/week – 2. 4 for keyboard use ≥ 30 hours/week Evidence-based Chiropractic 52 © 2007
Other measures of risk • Attributable risk (AR) – The probability of disease in the exposed group minus the probability of disease in the unexposed group – Represents the excess risk due to exposure to the factor under investigation – AR formula a c a+b c+d Evidence-based Chiropractic 53 © 2007
Other measures of risk (cont. ) • Attributable risk reduction (ARR) – Used mainly in RCTs – Is the difference in the probability of disease between the treatment and control groups – Thus, the formula for calculating ARR is the same as for attributable risk a c a+b c+d Evidence-based Chiropractic 54 © 2007
Other measures of risk (cont. ) • Relative risk reduction (RRR) – The comparative reduction in rates of bad outcomes between the experimental and control groups – Commonly reported, but not a very good way to compare outcomes • It magnifies small differences; hence clinically insignificant findings appear significant Evidence-based Chiropractic 55 © 2007
Other measures of risk (cont. ) – RRR formula (a/(a+b))-(c/(c+d)) c/c+d Evidence-based Chiropractic = Absolute risk reduction Probability of disease in control group 56 © 2007
Other measures of risk (cont. ) • Number needed to treat (NNT) – The number of patients who would need to be treated in order to prevent one additional bad outcome – NNT formula 1 (a/(a+b))-(c/(c+d)) Evidence-based Chiropractic = 1 Absolute risk reduction 57 © 2007
Other measures of risk (cont. ) • NNT example – Fritz et al reported that low back pain patients who had lumbar spine hypomobility had better outcomes after manipulation plus exercise, while patients with hypermobility did better with exercise alone – Hypomobility patients NNT = 2. 1 – Hypermobility patients NNT = 1. 6 Evidence-based Chiropractic 58 © 2007
Single group cohort studies • a. k. a. , longitudinal studies or inception cohort studies • Involve one group of subjects who have some factor in common – e. g. , had a positive screening test or were exposed to a risk factor – The inception cohort is regularly evaluated to monitor the evolution of the disease Evidence-based Chiropractic 59 © 2007
Use of contingency tables Case Control Cohort Diseased Not Diseased Exposed a b Not Exposed c d RCT Evidence-based Chiropractic 60 Cases Controls Exposed a b Not Exposed c d Diseased Not Diseased Treatment a b Control c d © 2007
Appraising epidemiologic articles • What is the study’s main objective? – Articles reporting on epidemiologic studies should specify why it was conducted in the first place – Otherwise, one can never tell whether the authors actually had a purpose or if they were conducting a “fishing expedition” – For example, “To identify risk factors for degenerative disc disease in factory workers” Evidence-based Chiropractic 61 © 2007
Appraising epidemiologic articles (cont. ) • Was the study population clearly identified? – It is important for the population to be adequately described in order to assess the external validity of the study – It amounts to a flaw when missing or incomplete Evidence-based Chiropractic 62 © 2007
Appraising epidemiologic articles (cont. ) • Did cases and controls originate from the same population? – Other than different exposure to the risk factors, groups should be as equivalent as possible – If dissimilar populations are involved, differences in outcomes could merely be the result of core differences in the populations Evidence-based Chiropractic 63 © 2007
Appraising epidemiologic articles (cont. ) • Did sufficient numbers of subjects complete the study? – A high drop-out rate may produce a biased study since drop-outs may be unlike those who followed through • For example, someone who drops out of a study because they are having too much pain – If drop-outs are lost to follow-up, the resulting estimate of risk is likely to be inaccurate Evidence-based Chiropractic 64 © 2007
Appraising epidemiologic articles (cont. ) • Was the appropriate study design used? – It depends primarily on the rarity of the disease being studied – Case-control designs • Best utilized with rare conditions • Sometimes used with common diseases • Suitable when there is a very long interval between the exposure and the outcome Evidence-based Chiropractic 65 © 2007
Appraising epidemiologic articles (cont. ) – Cohort designs • Best utilized with common diseases • Not feasible with rare conditions • The potential for bias in case-control and cross sectional studies is much greater than cohort studies Evidence-based Chiropractic 66 © 2007
Appraising epidemiologic articles (cont. ) • Were the exposures adequately described? – Exposures should be listed along with information on how they are measured – Accurate and precise measures of exposures should be used – Better studies describe the magnitude and duration of the exposures Evidence-based Chiropractic 67 © 2007
Appraising epidemiologic articles (cont. ) • Was the outcome clearly defined? – Authors should explain how the presence or absence of the outcome was determined – The same method of measurement should be used in both groups – The method of measurement should be valid and reliable Evidence-based Chiropractic 68 © 2007
Appraising epidemiologic articles (cont. ) • Is it readily apparent that the exposure preceded the outcome? – Especially of concern in case-control studies where temporality is assessed retrospectively – For example, it is difficult to tell whether chronic pain following whiplash causes patients to become depressed or if depression causes them to develop chronic pain Evidence-based Chiropractic 69 © 2007
Appraising epidemiologic articles (cont. ) • Was an association observed? If so, how strong is it? – If an association is found, an appropriate RR or OR will be reported – Consider the strength of the association and whether it is statistically significant • Confidence interval should not include 1, the value representing no association • P value ≤ 0. 05 Evidence-based Chiropractic 70 © 2007
Appraising epidemiologic articles (cont. ) • What biases potentially influenced the results of the study? – Selection bias – Observation bias • • Recall bias Use of measures that lack reliability and validity Evaluator bias Follow-up bias Evidence-based Chiropractic 71 © 2007
Appraising epidemiologic articles (cont. ) • Was the statistical analysis appropriate? – RR should be used with cohort studies – OR with case-control studies – All variables mentioned in the article should be included in the statistical analysis Evidence-based Chiropractic 72 © 2007
Appraising epidemiologic articles (cont. ) • Were the study’s conclusions supported by the evidence that was presented? – Findings should agree with and logically follow what was presented in the article – No speculative or unsupported statements – Authors should not suggest cause-and-effect relationships, over-generalizations, or expect patients to change behaviors based on weak associations Evidence-based Chiropractic 73 © 2007
Appraising epidemiologic articles (cont. ) • Should patients modify their behaviors to avoid the risk factors that were presented? – Evidence from a single study is rarely sufficient – Several well done studies that present strong enough relationships and include an adequate number of subjects should cause patients to modify or restrict their behaviors – Consistency of other studies on the topic Evidence-based Chiropractic 74 © 2007
- Slides: 74