Odds & Odds Ratios - Biostatistics

1. In a case-control study, 40% of lung cancer cases were exposed to radon, compared to 15% of controls. If there are 250 cases and 500 controls in the study, what is the odds ratio for lung cancer associated with radon exposure?
   Answer: 3.78 When you encounter case-control studies with exposure data, you're calculating an odds ratio to measure the strength of association between exposure and disease. The odds ratio compares the odds of exposure among cases to the odds of exposure among controls. First, set up your 2×2 table. With 250 cases and 40% exposed to radon, you have 100 exposed cases and 150 unexposed cases. With 500 controls and 15% exposed, you have 75 exposed controls and 425 unexposed controls. Calculate the odds ratio using the formula: $OR = \frac{a \times d}{b \times c}$ where a = exposed cases (100), b = exposed controls (75), c = unexposed cases (150), and d = unexposed controls (425). $OR = \frac{100 \times 425}{75 \times 150} = \frac{42,500}{11,250} = 3.78$ This confirms answer B is correct. Answer A (2.67) likely results from incorrectly calculating the ratio of proportions (40%/15% = 2.67) rather than the odds ratio. This is a common error - proportions and odds are different measures. Answer C (4.25) might come from switching numerator and denominator values in the calculation, while answer D (2.27) could result from various calculation errors or confusion about which cells to multiply. Remember: odds ratios require calculating odds (probability of event/probability of no event) for each group, then taking their ratio. Don't confuse this with simple proportion ratios, which will give you incorrect results in case-control studies.
2. A case-control study examines the relationship between cell phone use and brain tumors. The study includes 200 cases and 400 controls. If 70% of cases and 45% of controls report heavy cell phone use, what is the odds of heavy cell phone use among cases?
   Answer: 2.33 to 1 When you encounter case-control studies asking about odds, you're being tested on your ability to calculate and interpret odds ratios, which measure the strength of association between exposure and outcome. To find the odds of heavy cell phone use among cases, you need to calculate the ratio of exposed to unexposed cases. With 200 cases total and 70% reporting heavy use, that's 140 cases with heavy use and 60 cases without heavy use. The odds are therefore $\frac{140}{60} = 2.33$, expressed as 2.33 to 1. Answer A (2.33 to 1) correctly represents this calculation. This means cases are 2.33 times more likely to have heavy cell phone use than not. Answer B (1 to 0.43) represents the reciprocal relationship but uses incorrect formatting for odds notation. While mathematically related, odds are conventionally expressed with the larger number first. Answer C (0.43 to 1) would be the odds of not having heavy cell phone use among cases ($\frac{60}{140} = 0.43$), which answers the opposite of what's being asked. Answer D (1.56 to 1) appears to confuse this calculation with an odds ratio comparison between cases and controls, or represents a computational error in the basic odds calculation. Study tip: Remember that odds within a single group equals exposed/unexposed, while odds ratios compare odds between two groups. Always double-check whether the question asks for odds, odds ratios, or relative risk—these are different measures that students frequently confuse in case-control study questions.
3. A meta-analysis combines odds ratios from three studies of the same exposure-outcome relationship. Study A (n=500) reports OR=1.8, Study B (n=1200) reports OR=2.2, and Study C (n=800) reports OR=1.5. If the studies used similar methodologies and populations, what is the most reasonable interpretation when the overall random-effects odds ratio is 1.9?
   Answer: The odds ratio reflects appropriate weighting by sample size with minimal between-study heterogeneity A weighted average considering sample sizes: Study A weight ∝ 500, Study B weight ∝ 1200, Study C weight ∝ 800. The larger studies (B and C) have ORs of 2.2 and 1.5, averaging around 1.85, which is close to the meta-analysis result of 1.9. This suggests appropriate weighting with reasonable consistency. Choice A incorrectly focuses on the largest study alone. Choice C confuses publication bias with natural variation. Choice D incorrectly emphasizes the median over proper weighting.
4. In a cross-sectional study, the odds of hypertension among obese individuals is 3:2, and the odds of hypertension among non-obese individuals is 1:4. What is the odds ratio for hypertension comparing obese to non-obese individuals?
   Answer: 6.0 When you encounter odds ratios in cross-sectional studies, you're measuring the strength of association between an exposure (obesity) and an outcome (hypertension). The odds ratio compares the odds of the outcome in exposed versus unexposed groups. First, convert the given odds to decimal form. For obese individuals, odds of hypertension = 3:2 = 3/2 = 1.5. For non-obese individuals, odds of hypertension = 1:4 = 1/4 = 0.25. The odds ratio is calculated as: $\text{OR} = \frac{\text{Odds in exposed}}{\text{Odds in unexposed}} = \frac{1.5}{0.25} = 6.0$ This means obese individuals have 6 times the odds of having hypertension compared to non-obese individuals. Looking at the wrong answers: B) 1.5 represents only the odds for obese individuals, not the ratio comparing groups. C) 0.17 would result from incorrectly calculating the inverse (0.25/1.5), which would suggest obesity is protective against hypertension—clearly contradicting the given data. D) 2.4 might result from calculation errors, such as incorrectly adding or subtracting the odds rather than taking their ratio. The correct answer is A) 6.0. Study tip: Always remember that odds ratios compare odds between groups, not individual group odds. When given odds as ratios (like 3:2), convert to decimals first, then divide exposed odds by unexposed odds. An OR > 1 indicates increased risk in the exposed group, while OR < 1 suggests protection.
5. A pharmaceutical company conducts a clinical trial comparing a new antidepressant to placebo. Among 300 patients receiving the drug, 45 experience weight gain. Among 300 patients receiving placebo, 18 experience weight gain. The company wants to report the odds ratio for weight gain. What value should they report?
   Answer: 2.89 When you encounter odds ratio questions, you're comparing the odds of an event occurring in two different groups. The odds ratio tells you how much more (or less) likely an event is in one group compared to another. To calculate the odds ratio, you need to set up a 2×2 table and work with odds, not probabilities. First, organize your data: Drug group has 45 with weight gain and 255 without (300-45). Placebo group has 18 with weight gain and 282 without (300-18). The odds of weight gain in the drug group = 45/255 = 0.176. The odds of weight gain in the placebo group = 18/282 = 0.064. The odds ratio = 0.176/0.064 = 2.89, confirming answer B is correct. Looking at the wrong answers: A) 2.50 likely comes from incorrectly calculating 45/18 = 2.50, which gives you a risk ratio (comparing proportions directly) rather than an odds ratio. C) 0.40 represents the reciprocal of the correct answer (1/2.89), which you'd get if you accidentally flipped your comparison groups. D) 1.25 suggests a calculation error, possibly from using incorrect denominators or mixing up the formula. Remember this key distinction: risk ratios compare proportions directly (45/300 ÷ 18/300), while odds ratios compare odds (successes/failures in each group). Always double-check that you're calculating what the question asks for, and remember that odds ratios are typically larger than risk ratios when dealing with the same data.
6. In a study examining the relationship between coffee consumption and insomnia, the odds ratio is calculated as 2.4. If the odds of insomnia among non-coffee drinkers is 1:9, what are the odds of insomnia among coffee drinkers?
   Answer: 0.27 to 1 When you encounter odds ratio problems, you're dealing with the relationship between exposure and outcome across two groups. The odds ratio tells you how much the odds of an outcome change when comparing exposed to unexposed groups. Here, you need to use the fundamental relationship: $\text{Odds Ratio} = \frac{\text{Odds in exposed group}}{\text{Odds in unexposed group}}$ Given that the odds ratio is 2.4 and the odds of insomnia among non-coffee drinkers is 1:9, you can set up the equation: $2.4 = \frac{\text{Odds in coffee drinkers}}{1:9}$ Converting 1:9 to decimal form gives you 1/9 ≈ 0.111. Solving for the odds in coffee drinkers: $\text{Odds in coffee drinkers} = 2.4 × \frac{1}{9} = \frac{2.4}{9} = 0.267$ Converting back to ratio form: 0.267:1, which rounds to 0.27:1. This matches answer D. Answer A (2.4 to 1) incorrectly assumes the odds ratio itself represents the odds in the exposed group. Answer B (1 to 3.75) inverts the correct calculation and expresses it backwards. Answer C (3.75 to 1) represents the reciprocal of answer B but still reflects a fundamental misunderstanding of how to apply the odds ratio formula. Remember this key pattern: when the baseline odds are less than 1:1 (like 1:9 here), even a substantial odds ratio will often result in final odds that remain below 1:1. Don't let the magnitude of the odds ratio fool you into picking unreasonably high final odds.
7. Two case-control studies of the same exposure-disease relationship are conducted in different populations. Study 1 reports OR=2.5 with 95% CI (1.8, 3.5), while Study 2 reports OR=2.4 with 95% CI (1.2, 4.8). If both studies have the same number of cases and controls, what can be concluded about the exposure prevalence in the two control populations?
   Answer: Study 2 likely had lower exposure prevalence in controls, resulting in wider confidence intervals due to fewer exposed controls When you encounter case-control studies with identical sample sizes but different confidence interval widths, you're seeing the effects of exposure prevalence on statistical precision. The width of confidence intervals around odds ratios depends heavily on how many exposed individuals are in each group, particularly the control group. Both studies have nearly identical point estimates (OR = 2.5 vs 2.4), but Study 1 has a much narrower confidence interval (1.8, 3.5) compared to Study 2 (1.2, 4.8). Since both studies have the same number of cases and controls, this difference in precision must stem from the distribution of exposed versus unexposed individuals within those groups. Study 2's wider confidence interval indicates fewer exposed controls, which reduces statistical power and precision. When exposure prevalence is low in the control group, you have fewer exposed controls to compare against, leading to less stable estimates and wider confidence intervals. Study 1 likely had higher exposure prevalence in controls, providing more exposed individuals for comparison and thus greater precision. Option A incorrectly suggests Study 1 had higher exposure prevalence, but gets the causal direction wrong. Option B wrongly assumes similar point estimates mean similar exposure prevalence—precision depends on the actual numbers in each exposure category, not just the final ratio. Option D is incorrect because confidence interval width does provide information about the underlying data structure when sample sizes are known to be equal. Study tip: In case-control studies, remember that precision improves with balanced exposure groups. Rare exposures in controls always lead to wider confidence intervals, even with large sample sizes.
8. A study reports that the odds of obesity among individuals with diabetes is 4:1, while the odds of obesity among individuals without diabetes is 1:3. A student calculates the odds ratio as 12.0. What is the correct odds ratio?
   Answer: 12.0 When you encounter odds ratio problems, you're calculating how much the odds of an outcome differ between two groups. The odds ratio formula is: $\frac{\text{odds in exposed group}}{\text{odds in unexposed group}}$. Here, you need to identify which group is "exposed." Since we're studying the association between diabetes and obesity, diabetes status is the exposure. The odds of obesity among those with diabetes is 4:1 (or 4.0 when expressed as a ratio). The odds of obesity among those without diabetes is 1:3 (or 0.33 when expressed as a ratio). The odds ratio is therefore: $\frac{4.0}{0.33} = 12.0$. This means individuals with diabetes have 12 times the odds of being obese compared to those without diabetes. Looking at the wrong answers: Choice B (3.0) might result from incorrectly subtracting the odds rather than dividing them. Choice C (4.0) represents only the odds in the diabetes group, not the ratio between groups. Choice D (0.33) is just the odds in the non-diabetes group, or potentially the inverse of the correct odds ratio if you mistakenly flipped the exposure groups. The student's calculation of 12.0 is actually correct, making choice A right. Remember: odds ratios compare odds between groups, not individual odds values. Always identify your exposure group first, then divide exposed odds by unexposed odds. Values greater than 1 indicate increased odds in the exposed group, while values less than 1 indicate decreased odds.
9. A study reports an odds ratio of 2.4 for the association between smoking and coronary heart disease. If the odds of coronary heart disease among non-smokers is 1:15, what are the odds of coronary heart disease among smokers?
   Answer: 1:6.25 The odds ratio formula is OR = (odds in exposed)/(odds in unexposed). Given OR = 2.4 and odds in non-smokers = 1/15 = 0.0667, we have: 2.4 = (odds in smokers)/0.0667. Therefore, odds in smokers = 2.4 × 0.0667 = 0.16 = 1:6.25. Choice B incorrectly presents the calculation components. Choice C uses an incorrect calculation. Choice D confuses the OR value with actual odds.
10. In a case-control study of pancreatic cancer, the crude odds ratio for smoking is 3.2. After stratifying by age (≤50 vs >50 years), the age-specific odds ratios are 3.1 and 3.3 respectively. What does this suggest about age as a confounder?
    Answer: Age is not a confounder because the stratified ORs are similar to the crude OR When evaluating confounding in stratified analyses, you need to compare how much the crude odds ratio changes when you account for the potential confounder. A confounder creates a meaningful difference between crude and adjusted measures of association. Here, the crude OR is 3.2, while the age-stratified ORs are 3.1 and 3.3. These values are remarkably similar – the crude OR falls almost exactly between the two stratified estimates. This minimal variation indicates that age is not meaningfully distorting the smoking-pancreatic cancer relationship. Answer B correctly identifies that age is not a confounder because the stratified ORs closely match the crude OR. When a variable isn't confounding, controlling for it produces little change in the effect estimate. Answer A incorrectly assumes any difference indicates strong confounding. The differences here (3.2 vs 3.1 and 3.2 vs 3.3) are trivial – well within expected random variation. Strong confounding would show substantial differences, like a crude OR of 3.2 changing to stratified ORs of 1.8 and 2.0. Answer C misinterprets what to compare. The small difference between stratified ORs (3.1 vs 3.3) doesn't indicate moderate confounding. You compare crude to stratified values, not stratified values to each other. Answer D is wrong because you can assess confounding by comparing effect estimates across strata. While knowing age distributions would provide additional context, the OR comparison alone sufficiently demonstrates the absence of meaningful confounding. Remember: look for substantial changes (typically >10-15%) between crude and stratified estimates to identify meaningful confounding.
11. A cohort study follows 1,000 smokers and 1,000 non-smokers for 10 years. Among smokers, 80 develop lung cancer and 920 do not. Among non-smokers, 20 develop lung cancer and 980 do not. What are the odds of developing lung cancer among smokers?
    Answer: 0.087 to 1 When you encounter cohort study data like this, you're typically being asked to calculate measures of association. The key distinction here is between odds and risk - odds compare those who develop the outcome to those who don't, while risk is simply the proportion who develop the outcome. To find the odds of developing lung cancer among smokers, you need to set up the ratio of smokers who developed cancer to smokers who didn't develop cancer. Among the 1,000 smokers, 80 developed lung cancer and 920 did not. Therefore, the odds are 80:920, which simplifies to 80/920 = 0.087. This means the odds are 0.087 to 1. Looking at the incorrect answers: B (1 to 11.5) gives you the reciprocal relationship - this would be the odds of NOT developing cancer (920:80 = 11.5:1, or 1:11.5 when flipped). C (11.5 to 1) represents those same odds of not developing cancer. D (0.080 to 1) appears to be calculating risk instead of odds - this would be 80/1000 = 0.08, which is the proportion (risk) of smokers who developed cancer, not the odds. The correct answer is A because it properly expresses the odds as the ratio of those with the outcome to those without the outcome among the exposed group. Study tip: Remember that odds = events/non-events, while risk = events/total population. When you see "odds" in a question, always think about the ratio comparing those with and without the outcome, not just the proportion with the outcome.
12. A researcher reports an odds ratio of 0.6 (95% CI: 0.3-1.2) for the association between regular exercise and diabetes. What is the most accurate interpretation of this finding?
    Answer: The association is not statistically significant at α = 0.05 When interpreting odds ratios and confidence intervals, you need to examine both the point estimate and whether the confidence interval crosses 1.0 to determine statistical significance. The odds ratio of 0.6 suggests that exercisers have 0.6 times the odds of diabetes compared to non-exercisers, which appears protective. However, the 95% confidence interval (0.3-1.2) crosses 1.0, meaning we cannot rule out the possibility that exercise has no effect (OR = 1.0) or even increases diabetes risk. Since the interval includes 1.0, this association is not statistically significant at α = 0.05, making C correct. A is wrong because you cannot claim a significant reduction when the confidence interval includes 1.0. The word "significantly" implies statistical significance, which this result lacks. B makes two errors: first, it incorrectly calculates the effect size (an OR of 0.6 represents a 40% reduction in odds, not 60%), and second, it ignores that the result isn't statistically significant. D is incorrect because while the confidence interval does include values above 1.0 (suggesting possible increased risk), the point estimate of 0.6 actually suggests decreased risk. More importantly, the non-significant result means we cannot conclude exercise either increases or decreases diabetes risk. Study tip: Always check if confidence intervals for odds ratios cross 1.0. If they do, the association is not statistically significant, regardless of what the point estimate suggests. Never interpret effect sizes when results aren't statistically significant.
13. Cases: 90% exposed; controls: 75% exposed. OR for non-exposure?
    Answer: 0.33 Set 100 cases and 100 controls. Non-exposure odds among cases are 10/90 = 0.111, and among controls are 25/75 = 0.333. Divide 0.111 by 0.333 to get 0.33. The tempting wrong answer is 3.00, which is the odds ratio for exposure; asking about non-exposure means you invert it.
14. A clinical trial randomizes 200 patients to receive either a new drug or placebo. In the treatment group (100 patients), 15 experience side effects and 85 do not. In the placebo group (100 patients), 8 experience side effects and 92 do not. What is the odds ratio for experiencing side effects in the treatment group compared to the placebo group?
    Answer: 2.04 When you encounter a clinical trial comparing two groups for a binary outcome like side effects, you're typically dealing with an odds ratio calculation. The odds ratio compares the odds of an event occurring in one group versus another. To calculate the odds ratio, first find the odds for each group. Odds equals the number with the outcome divided by the number without the outcome. For the treatment group: 15 patients with side effects ÷ 85 without side effects = 0.176. For the placebo group: 8 with side effects ÷ 92 without = 0.087. The odds ratio is the treatment odds divided by placebo odds: 0.176 ÷ 0.087 = 2.04. You can also use the cross-multiplication shortcut: $\text{OR} = \frac{15 \times 92}{85 \times 8} = \frac{1380}{680} = 2.04$. This confirms answer B is correct. Looking at the wrong answers: A) 1.87 likely results from a calculation error or rounding mistake during the computation. C) 0.53 represents the inverse of the correct odds ratio - this happens when students accidentally calculate placebo odds divided by treatment odds instead of treatment divided by placebo. D) 1.76 suggests another computational error, possibly from incorrectly setting up the 2×2 table. Remember the formula structure: odds ratio always equals (a×d)/(b×c) where a and d are diagonal cells in your 2×2 contingency table. Double-check which group you're comparing to which - the reference group goes in the denominator.
15. In a case-control study of lung cancer, researchers find that among 200 cases, 160 were exposed to asbestos, while among 300 controls, 90 were exposed to asbestos. If the prevalence of asbestos exposure in the general population is 25%, what is the ratio of the odds of exposure in cases to the odds of exposure in the general population?
    Answer: 5.33 First, calculate odds of exposure in cases: 160/40 = 4.0. Then, calculate odds of exposure in general population: 0.25/0.75 = 0.333. The ratio is 4.0/0.333 = 5.33. Choice B uses the wrong denominator for cases (160/200 instead of 160/40). Choice C incorrectly uses the control group odds. Choice D uses prevalence instead of odds for the general population.
16. A meta-analysis combines results from multiple studies examining the association between vitamin D deficiency and depression. Study A reports OR = 2.1 (95% CI: 1.3-3.4), Study B reports OR = 1.8 (95% CI: 1.1-2.9), and Study C reports OR = 2.5 (95% CI: 1.4-4.5). What can be concluded about the consistency of these findings?
    Answer: The studies are consistent because all confidence intervals exclude 1.0 and overlap substantially When evaluating consistency across studies in meta-analysis, you need to assess whether the effect estimates are reasonably similar and whether the confidence intervals show meaningful overlap. Consistency doesn't require identical point estimates—some variation is expected due to different populations, methodologies, and sample sizes. The correct answer is B because all three studies show statistically significant positive associations (all confidence intervals exclude 1.0) and demonstrate substantial overlap in their confidence intervals. Study A (1.3-3.4), Study B (1.1-2.9), and Study C (1.4-4.5) all overlap considerably, suggesting the true effect likely falls within a similar range across studies. The point estimates (2.1, 1.8, 2.5) are also reasonably close, differing by less than one unit. A is incorrect because there's no established rule that differences greater than 0.5 units indicate inconsistency—this arbitrary threshold ignores confidence intervals and statistical significance. C misunderstands that wider confidence intervals simply reflect larger uncertainty (often due to smaller sample sizes) but don't indicate inconsistency if they overlap substantially with other studies. D incorrectly suggests Studies A and C are inconsistent when their confidence intervals actually overlap from 1.4 to 3.4, showing reasonable agreement. Study tip: In meta-analysis questions, focus on confidence interval overlap and whether all studies point in the same direction (all above or below the null value of 1.0 for odds ratios). Don't get distracted by minor differences in point estimates—look for the bigger picture of effect direction and magnitude consistency.
17. A systematic review identifies 5 studies examining the association between air pollution exposure and asthma. The individual odds ratios are: 1.8, 2.1, 1.6, 2.4, and 1.9. If a simple average is calculated, what would be the mean odds ratio, and why might this approach be problematic for meta-analysis?
    Answer: Mean OR = 1.96; problematic because it doesn't weight studies by sample size or precision When you encounter meta-analysis questions, remember that combining effect estimates requires careful consideration of how studies should be weighted and combined mathematically. Let's calculate the simple average: $(1.8 + 2.1 + 1.6 + 2.4 + 1.9) ÷ 5 = 9.8 ÷ 5 = 1.96$. So the mean odds ratio is 1.96, which immediately eliminates options C and D. The key issue with this simple averaging approach is that it treats all studies equally, regardless of their sample sizes or precision. In proper meta-analysis, studies with larger sample sizes or narrower confidence intervals (higher precision) should receive more weight because they provide more reliable estimates. A study with 10,000 participants should influence the pooled estimate more than a study with 100 participants, but simple averaging ignores this critical distinction. Option A mentions study quality, which is important for inclusion decisions but isn't the primary mathematical problem with simple averaging. Option C incorrectly calculates the mean as 2.0 and suggests log-scale averaging, but while log transformation is sometimes used in meta-analysis, the main issue here is weighting. Option D also miscalculates the mean and focuses on exposure definition consistency, which is a study selection issue rather than a statistical combination problem. Study tip: In meta-analysis questions, always consider how studies should be weighted. Simple averages treat all studies equally, but proper meta-analysis uses inverse-variance weighting or other methods that give more influence to studies with greater precision or larger sample sizes.
18. Case-control OR for exposure=0.40. Interpretation?
    Answer: Case exposure odds 60% lower An OR of 0.40 means cases have 0.40 times the exposure odds of controls, so the reduction is 1 - 0.40 = 0.60, or 60% lower. You compare case odds to control odds, not to 1 directly. The tempting mistake is choosing 40% lower, which treats the OR itself as the percent reduction instead of the remaining fraction.
19. Exposed: 40 of 100 diseased; unexposed: 24 of 200 diseased. OR?
    Answer: 4.89 Set up the 2x2 table: exposed diseased 40, exposed non-diseased 60; unexposed diseased 24, unexposed non-diseased 176. The odds ratio is (40 x 176) / (60 x 24) = 7040 / 1440 = 4.89. The tempting wrong answer 5.56 comes from using the total unexposed count 200 instead of the unexposed non-diseased count 176.
20. In a case-control study of bladder cancer, the crude odds ratio for smoking is 3.2. After stratifying by age group, the odds ratios are 2.8 for ages <50, 3.0 for ages 50-65, and 3.1 for ages >65. The Mantel-Haenszel adjusted odds ratio is 2.95. What is the most appropriate interpretation of these findings?
    Answer: Age shows minimal confounding effect, and the association is relatively consistent across age groups When evaluating confounding and effect modification in stratified analyses, you need to assess two key relationships: how much the crude odds ratio differs from the adjusted estimate (confounding), and whether the stratum-specific odds ratios vary meaningfully across strata (effect modification). Here, the crude OR is 3.2, while the Mantel-Haenszel adjusted OR is 2.95 - a difference of only 0.25 or about 8%. This small difference indicates minimal confounding by age. Additionally, the stratum-specific ORs (2.8, 3.0, 3.1) are remarkably consistent, varying by only 0.3 units around the adjusted estimate of 2.95. This consistency suggests the smoking-bladder cancer association is similar across age groups, making D correct. Option A is wrong because a "substantial" difference would typically be at least 10-15% change in the OR, not the modest 8% seen here. Option B incorrectly suggests upward bias persists after adjustment - but the Mantel-Haenszel OR accounts for age stratification, so any age-related confounding has been addressed. Option C misinterprets the small variation in stratum-specific ORs (2.8-3.1) as effect modification, but this 0.3-unit range represents normal sampling variation, not meaningful heterogeneity that would indicate true effect modification. Study tip: For confounding, look for >10% difference between crude and adjusted estimates. For effect modification, stratum-specific ORs should differ by more than just sampling variation - typically showing a clear pattern or substantially different magnitudes across strata.

Start this quiz