CrossSectional Studies 2 Akbar soltani Assistant Professor of

Cross-Sectional Studies (2) Akbar soltani Assistant Professor of Medicine and Endocrinology Evidence-Based Medicine Working Team (EBMWT) Tehran University of Medical Sciences (TUMS) Shariati Hospital EBMWT A. SOLTANI TUMS

Critical appraisal: Crosssectional Studies (7) n In Cross-sectional studies think of: n n n n EBMWT Sampling Procedures. Clear definition of Target Population. Clear definition of outcome. Clear definition of risk factors. Statistical analysis Remember bias. Remember Confounders. Causality? A. SOLTANI TUMS

Cross-Sectional Study MRFIT: Association of left ventricular hypertrophy (LVH) by ECG and hypertensive status at baseline LVH Hypertensive Normotensive No LVH 231 7, 781 8, 012 43 4, 811 4, 854 274 12, 592 12, 866 Prevalence of LVH among hypertensives = 231 / 8, 012 = 2. 9% Prevalence of LVH among normotensives = 43 / 4, 854 = 0. 9% EBMWT A. SOLTANI TUMS

Cross-Sectional Study Association of PCP and gender at baseline in the CPCRA PCP Men Women No PCP 371 1, 189 2, 260 58 493 551 429 2, 382 2, 811 History of PCP among men = 371 / 2, 260 = 16. 4% History of PCP among women = 58 / 551 = 10. 5% P-value = 0. 001 for difference EBMWT A. SOLTANI TUMS

Critical appraisal: Crosssectional Studies (7) n In Cross-sectional studies think of: n n n n EBMWT Sampling Procedures. Clear definition of Target Population. Clear definition of outcome. Clear definition of risk factors. Statistical analysis Remember bias. Remember Confounders. Causality? A. SOLTANI TUMS

Statistical Analysis and Interpretation Calculating Prevalence P 1= a/a+b= 50/250 = 20% prevalence of CHD among people who are not active. P 0= c/c+d = 50/750 = 6. 7% prevalence of CHD among people who are active. EBMWT A. SOLTANI TUMS

PR = (a/N 1)/(c/N 0) PR= (50/250)/(50/750)=3. 0 EBMWT A. SOLTANI TUMS

Prevalence Ratio n n n EBMWT The prevalence ratio (PR) is analogous to the cumulative incidence ratio (CIR) of cohort studies. The denominators for both ratios are fixed populations – fixed at the start of the study in the case of a cohort study, and fixed at the point or period of time for the case-control study. The prevalence ratio is calculated when the outcome occurs over a short period of time. For example, one would calculate a prevalence ratio for an acute outbreak of tuberculosis in a prison population. This is in contrast to calculating the overall prevalence of positive tuberculin skin tests among the prisoners. A. SOLTANI 40 TUMS

Incidence: three types n Takes time at risk into account n Cumulative Incidence/(Risk) n n Incidence Density/Rate n n Denominator is number of people at risk at the beginning Denominator is the sum of the lengths of time people were at risk “person-time” The only measure that take account the time of disease onset Odds of disease n Denominator is number of people at risk at the end EBMWT Numerators are same A. SOLTANI TUMS

Prevalence Odds Ratio n The POR is calculated by dividing the odds of having CHD among inactive persons by the odds for those who are physically active. POR is 50 x 700/200 x 50, which is equal to 3. 5. so odds of having CHD is 3. 5 times greater if a person is not physically active than if a person is physically active or estimates the IRR and is interpreted to mean that the estimated incidence rate among the exposed is 3. 5 times greater than that among the unexposed. EBMWT A. SOLTANI TUMS

prevalence odds ratio n n n The POR estimates the incidence ratio (IRR) that would be obtained from a cohort study if the risk factor occurs over an extended period of time and if the duration of the outcome is not affected by exposure status. (duration of disease is the same in both groups) and also if presence of disease does not affect the exposure status. prevalence among unexposed is low<10% EBMWT A. SOLTANI TUMS

POR and PR at a Glance POR n Estimates the IRR n Best for chronic diseases n Measure of association n EBMWT A. SOLTANI PR n Estimates the CIR n Best for acute diseases n Measure of occurrence n TUMS

Which Measure of Association is Best? n n If the prevalence of disease is low, i. e. 10% or less in exposed and non-exposed populations, POR= PR. Since cross-sectional studies are particularly useful for investigating chronic diseases (e. g. prevalence of AIDS) where the onset of disease is difficult to determine, or for studying long lasting risk factors (such as smoking, hypertension, VD, and high fat diets), the prevalence odds ratio will generally be the preferred measure of association. EBMWT A. SOLTANI TUMS

Critical appraisal: Cross-sectional Studies (7) POR>1 <1 n Have some statistically "significant" results been obtained by "trawling" the data n EBMWT A. SOLTANI TUMS

Self-evaluation n n EBMWT A hypothetical study of the effect of alcohol intake on the risk of gallstones in Mexican. Americans aged 20 -74 was conducted. Ultrasonography of the gallbladder of 2200 study subjects identified 152 subjects with gallstones, 42 of whom listed their alcohol consumption level as "high". Of the men without gallstones, 220 listed their alcohol intake level as "high". Interviews were conducted at the same time as the examinations to determine the exposure level of study participants. A. SOLTANI TUMS

Self-evaluation n n EBMWT Which measure of association is most appropriate in this cross-sectional design? a. Prevalence ratio b. Prevalence odds ratio c. Cumulative incidence d. None of the above A. SOLTANI TUMS

Self-evaluation 2. Correct answer: b. n The prevalence odds ratio (POR) is the correct measure of association because POR is preferred in studies of chronic diseases or studies of long-lasting risk factors, such as alcohol intake. n EBMWT A. SOLTANI TUMS

Self-evaluation n EBMWT a. Incorrect. The PR is calculated when the outcome occurs over a short time, such as an acute outbreak of an infectious disease. Gallstones develop more slowly, so the POR would be more appropriate. A. SOLTANI TUMS

Self-evaluation n n EBMWT c. Incorrect. The cumulative incidence (CI) is a measure of association that can only be calculated from a cohort study. It requires a count of total new cases of a disease divided by the population at risk. This is not possible from a cross-sectional study since subjects are not observed over time. d. Incorrect. A. SOLTANI TUMS

Uses of Descriptive/Analytical Studies n n EBMWT Survival bias may be minimized if information can be obtained on exposures that clearly preceded the first symptoms or signs of a chronic disease such as arthritis, diabetes, chronic bronchitis. The ability to accomplish this depends on access to medical records documenting the first visits or examinations for the chronic disease, and possibly on historical records on the exposure of the individual prior to these first visits, e. g. where the person lived; where the person was employed A. SOLTANI TUMS

Critical appraisal: Crosssectional Studies (8) n In Cross-sectional studies think of: n n n n EBMWT Sampling Procedures. Clear definition of Target Population. Clear definition of outcome. Clear definition of risk factors. Statistical analysis Remember bias. Remember Confounders. causality? A. SOLTANI TUMS

Other Issues in Analysis n Subject to: n Selection and measurement bias n Confounding n Interactions n Sampling error EBMWT A. SOLTANI TUMS

Selection bias n n Sampling bias Ascertainment bias n n Participation bias n n n EBMWT Surveillance Referral, admission Diagnostic Self-selection (volunteerism) Non-response, refusal (BMD, DM in young) Survival A. SOLTANI TUMS

Selection bias CCS>Cross>Cohort n n Appropriate sampling does not guarantee representativeness: failure of participation EX: BMD in Tehran (time and money are limitations of recalls) EBMWT A. SOLTANI TUMS

Measurement-Information bias-Misclassification n n EBMWT Disease or exposure 1. Observer (interviewer Bias) 2. Responder (recall , memory bias) 3. Instruments (differential and nondifferential misclassification) A. SOLTANI TUMS

Misclassification of a single attribute (E or O)& observed prevalence n n n n n EBMWT EX: X-Sectional, prevalence of osteoporosis risk factors Questioner without refusals: exercise, sun exposure P*=observed proportion P=true prevalence=0. 1% P*=p. sensitivity+(1 -p). (1 -spesifity) Sen=90% Sp=90% P*=10. 1% p*+sp-1 p= sen+sp-1 A. SOLTANI TUMS

Misclassification of a single attribute (E or D)& observed prevalence n n The bias in overestimation is severely influenced by losses in SP (especially with true prevalence <50%) In contrast losses in SEN have at most moderate effects on the observed prevalence 20 EBMWT A. SOLTANI TUMS

Misclassification Measurement error leads to assigning wrong exposure or outcome category Non-differential Differential 6 Due to random error? ? 6 Due to systematic error 6 Unrelated to exposure or outcome status 6 Related to exposure or outcome status 6 Not a bias? ? 6 Bias 6 Weakens measure of association 6 Measure of association distorted in any direction EBMWT A. SOLTANI TUMS

Misclassification & E-D relationships 1. Non differential E misclassification: exercise, VD n (Lead to null only when E is a binary variable : E& UNE) EBMWT A. SOLTANI TUMS

Misclassification & E-D relationships 2. Differential E misclassification: recall, observer bias or… n not a problem for osteoporosis (asymptomatic disease and blinding) EBMWT A. SOLTANI TUMS

3. Non differential disease misclassification n n n n True True risk in the new drug r 1=25/10000 risk in placebo r 0=50/10000 risk ratio=r 1/r 0=0. 5 risk difference =r 0 -r 1=25/10000 prevented fraction=(r 0 -r 1)/r 0=0. 5=50% Prevention of lung cancer SEN=80% Observed r 1=20/10000 Observed r 1=40/10000 Observed risk ratio=0. 5 Observed risk difference= r 0 -r 1=20/10000 Observed prevented fraction=50% SP=90% Observed Observed EBMWT r 1=1022/10000 r 1=1045/10000 risk ratio=0. 98 risk difference= r 0 -r 1=23/10000 prevented fraction=2. 2% A. SOLTANI TUMS

3. Non differential disease misclassification n n n True True prevalence, without risk factor r 1=25/10000 healthy prevalence, with risk factor r 0=50/1000 osteoporotics prevalence ratio=20 difference =r 0 -r 1=25/10000 prevented fraction=(r 0 -r 1)/r 0=0. 5=50% SEN=80% Observed r 1=25/10000 Observed r 0=50/8000=64/1000 Observed risk ratio=25. 6 SP=90% Observed r 1=1025/10000 Observed r 1=50/1000 Observed risk ratio=0. 487 EBMWT A. SOLTANI TUMS

Misclassification & E-D relationships 3. Non differential disease misclassification n Either does not affect the estimates of relative effect (if SP=100% & SEN<100%) n Bias towards a relative effects of 1 (if SP<100% & SEN<=100%) n Decrease in SP resulted in risk ratio close to one and in marked underestimation of the prevalence fraction EBMWT A. SOLTANI TUMS

Misclassification & E-D relationships 4. Differential disease misclassification n Bias toward an under or overestimation of true EX effect n Observer bias if stuff of BMD center know risk factors (questioner) EBMWT A. SOLTANI TUMS

Issues in Designing X-Sectional Survey n Confounding n n Can occur in all observational studies Studies comparing two groups n n n cases (exposed), controls (not exposed) Could use matching, but can only match on known confounders Matching is rarely used in cross-sectional studies n n EBMWT No information on subjects with regard to their status on potential confounders or exposure before the study is initiated Confounding often handled through statistical analysis (multivariable strategies) A. SOLTANI TUMS

Strengths and Weaknesses of Observational Analytic Study Designs Ecological Crosssectional Casecontrol Cohort Selection bias NA Medium High Low Information bias ( Recall) NA High Low Loss to follow up NA NA Low High Medium Low Time required Low Medium High Cost Low Medium High Probability of: n n Confounding NA EBMWT = Not applicable A. SOLTANI TUMS

Assessing the relationship between factors in epidemiology EBMWT A. SOLTANI TUMS

Critical appraisal: Crosssectional Studies n In Cross-sectional studies think of: n n n n EBMWT Sampling Procedures. Clear definition of Target Population. Clear definition of outcome. Clear definition of risk factors. Statistical analysis Remember bias. Remember Confounders. causality? A. SOLTANI TUMS

Measures of disease occurrence and exposure effect in analytic study designs Type of Analytical Study Measure of disease (outcome) occurrence Measure of exposure effect Ecological Rate, Risk, Prevalence Correlation or Regression Coefficient Cross-sectional Prevalence ratio, Prevalence difference, Odds ratio Cohort Rate, Risk, Odds Rate Ratio, Risk Ratio, Odds Ratio, Rate Difference, Vaccine efficacy Case-Control None 1 Odds Ratio, Vaccine efficacy Intervention Rate, Risk, Odds Rate Ratio, Risk Ratio, Odds Ratio, Rate Difference, Vaccine efficacy 1 Unless and the sampling fraction is known for both cases and controls i. e. proportion of cases TUMS of controls sampled from the. A. SOLTANI population is known EBMWT proportion

attribute cohort Case. c Cross. s Classification of population Free from condition Case. control Disease+exposure+- Sample represented Non-diseased Uncertain Source ? Survivors in time Temporal sequence Prospective retrospective function Compare incidence rates compare prevalence of EX Retrospective relation revealed at a point or period Describe association of EX & disease simultaneously EBMWT A. SOLTANI TUMS

attribute cohort Case. c Cross. s Outcome comparison Incidence of disease Prevalence of EX Prevalence of disease in EX &UNEX Risk measurement RR, AR (precise) Odds ratio Prevalence ratio(inexact estimate of RR)also Odds ratio Evidence for causality strong Need more Only suggestive careful analysis bias Easy to manage Need more effort to manage EBMWT A. SOLTANI May be very difficult to manage TUMS

basis cohort Case. c Cross. s Rare condition Not practical best NA To determine precise risk best Only estimate possible Gives relative prevalence Temporality best NA NA For administrative purpose NA NA best If attrition is a serious problem (if >30% conclusion will be suspected NA Attrition is usually minimal Attrition may have occurred before the study If selective survival is a problem best NA NA If all factors are unknown best NA Less appropriate Time and money Most expensive Least expensive In between EBMWT A. SOLTANI TUMS

Self-evaluation n n For which of the following disease situations could a cross-sectional study design be used to determine prevalence? a. Disease w, a highly fatal disease in which the average length of survival after diagnosis is approximately one month. b. Disease x, a disease with a long latency period where infection is identifiable by a serological test. c. Disease y, a disease which is caused by a chromosomal malformation present at birth (or before). d. Both b and c EBMWT A. SOLTANI TUMS

answer 1. Correct answer: d. Both b and c are situations in which a cross-sectional study design could be used To determine a measure of prevalence. Prevalence can be determined if a disease has a long latency period that can be detected through a serological test. In choice c, the duration of disease is not affected by exposure. The condition is due to a chromosomal malformation, and will be present throughout life regardless of how long a person was exposed. n Prevalence is often used to describe conditions that are permanent, and present at birth, such as congenital malformations. EBMWT TUMS n A. SOLTANI

answer n EBMWT a. Incorrect. In this situation, highly fatal cases would not be represented in a measure of prevalence since many people with the disease may have died before the questionnaire was administered. Prevalence is not an accurate measure when a disease is highly fatal. A. SOLTANI TUMS

Thank you EBMWT A. SOLTANI TUMS