A Tale of Two Ratios: Calculating and Interpreting Odds Ratios vs. Relative Risks in Epidemiology

Relative Risk (RR) and the Odds Ratio (OR) are the two primary measures used in epidemiology to quantify the strength of an association between an exposure and a health outcome. The choice between them is dictated by the study design. Relative Risk compares the incidence of disease in an exposed group to an unexposed group, making it an intuitive measure of risk. It is the appropriate measure for prospective studies like cohort studies and randomized controlled trials, where researchers can follow individuals over time to directly calculate risk. In contrast, the Odds Ratio is used in case-control studies, where subjects are selected based on their disease status. The OR compares the odds of a prior exposure among those with the disease (cases) to the odds of exposure among those without the disease (controls). While less intuitive, the OR is a powerful tool. A critical concept is the 'rare disease assumption,' which states that when an outcome is rare (e.g., incidence <10%), the OR provides a close mathematical approximation of the RR. However, for common diseases, the OR will systematically overestimate the strength of the association relative to the RR, so it is crucial to interpret these measures correctly and in the context of the study from which they were derived.

<h3>Introduction: Quantifying Association in Epidemiology</h3> <p>In epidemiology, our goal is often to understand if a particular exposure (like smoking, a new medication, or a dietary habit) is associated with a specific outcome (like lung cancer, recovery from illness, or heart disease). To move beyond simply saying 'there is an association,' we need to quantify its strength and direction. Two of the most fundamental measures of association used for this purpose are the <strong>Relative Risk (RR)</strong> and the <strong>Odds Ratio (OR)</strong>. While they may seem similar, they are calculated differently, interpreted differently, and are appropriate for different study designs. Understanding their distinctions is crucial for accurately interpreting epidemiological research.</p>

<h3>The Foundation: The 2x2 Contingency Table</h3> <p>Both RR and OR are calculated from a simple 2x2 table, which cross-tabulates the exposure status against the disease status for a group of individuals.</p>

<table border="1" style="width:100%; border-collapse: collapse;"> <tr> <th style="padding: 8px; text-align: left;"></th> <th style="padding: 8px; text-align: left;">Disease Present (Cases)</th> <th style="padding: 8px; text-align: left;">Disease Absent (Controls)</th> <th style="padding: 8px; text-align: left;">Total</th> </tr> <tr> <td style="padding: 8px;"><strong>Exposed</strong></td> <td style="padding: 8px; text-align: center;">A</td> <td style="padding: 8px; text-align: center;">B</td> <td style="padding: 8px; text-align: center;">A + B</td> </tr> <tr> <td style="padding: 8px;"><strong>Unexposed</strong></td> <td style="padding: 8px; text-align: center;">C</td> <td style="padding: 8px; text-align: center;">D</td> <td style="padding: 8px; text-align: center;">C + D</td> </tr> </table>

<ul> <li><strong>A:</strong> Number of exposed individuals who have the disease.</li> <li><strong>B:</strong> Number of exposed individuals who do not have the disease.</li> <li><strong>C:</strong> Number of unexposed individuals who have the disease.</li> <li><strong>D:</strong> Number of unexposed individuals who do not have the disease.</li> </ul>

<hr/>

<h3>Section 1: Relative Risk (RR) - The Risk Ratio</h3>

<p><strong>Definition:</strong> Relative Risk, also known as the Risk Ratio, is the ratio of the probability (or risk) of an outcome occurring in an exposed group to the probability of the outcome occurring in an unexposed group. It directly compares the incidence of disease in the two groups.</p>

<p><strong>Appropriate Study Designs:</strong> RR can only be calculated in studies where we can measure incidence. This means we must follow groups over time to see who develops the disease. Therefore, RR is the measure of choice for <strong>cohort studies</strong> and <strong>randomized controlled trials (RCTs)</strong>.</p>

<p><strong>Calculation:</strong></p> <ul> <li>Risk in the exposed group = Incidence in exposed = A / (A + B)</li> <li>Risk in the unexposed group = Incidence in unexposed = C / (C + D)</li> <li><strong>Formula: RR = [A / (A + B)] / [C / (C + D)]</strong></li> </ul>

<p><strong>Interpretation:</strong></p> <ul> <li><strong>RR = 1:</strong> The risk of the outcome is identical in the exposed and unexposed groups. There is no association.</li> <li><strong>RR > 1:</strong> The risk of the outcome is higher in the exposed group. The exposure is a <strong>risk factor</strong>. An RR of 2.5 means the exposed group has 2.5 times the risk of developing the disease compared to the unexposed group.</li> <li><strong>RR < 1:</strong> The risk of the outcome is lower in the exposed group. The exposure is <strong>protective</strong>. An RR of 0.6 means the exposed group has 0.6 times the risk (or a 40% reduction in risk) of developing the disease compared to the unexposed group.</li> </ul>

<p><strong>Example:</strong> A 10-year cohort study follows 2,000 people to investigate the link between heavy coffee drinking (exposure) and heart disease (outcome). At the end of the study: <ul> <li>Of 500 heavy coffee drinkers, 50 developed heart disease (A=50, B=450).</li> <li>Of 1,500 non-coffee drinkers, 75 developed heart disease (C=75, D=1425).</li> </ul> Risk in exposed = 50 / (50 + 450) = 50 / 500 = 0.10 (or 10%) <br/> Risk in unexposed = 75 / (75 + 1425) = 75 / 1500 = 0.05 (or 5%) <br/> RR = 0.10 / 0.05 = 2.0 <br/> <strong>Interpretation:</strong> In this study, heavy coffee drinkers had <strong>twice the risk</strong> of developing heart disease compared to non-coffee drinkers over the 10-year period.</p>

<hr/>

<h3>Section 2: The Odds Ratio (OR)</h3>

<p><strong>Definition:</strong> The Odds Ratio is the ratio of the odds of an exposure among those with the disease (cases) to the odds of the exposure among those without the disease (controls). An 'odd' is the probability of an event occurring divided by the probability of it not occurring (P / (1-P)).</p>

<p><strong>Appropriate Study Designs:</strong> The OR is the primary measure of association for <strong>case-control studies</strong>. In this design, we start with a group of people who already have the disease (cases) and a group who do not (controls) and then look backward to determine their prior exposure status. Because we pre-select cases and controls, we cannot calculate the incidence or risk of disease in the population.</p>

<p><strong>Calculation:</strong></p> <ul> <li>Odds of exposure in cases = A / C</li> <li>Odds of exposure in controls = B / D</li> <li><strong>Formula: OR = (A / C) / (B / D) = (A D) / (B C)</strong></li> <li>This is often called the 'cross-product ratio' because you multiply the diagonal cells of the 2x2 table.</li> </ul>

<p><strong>Interpretation:</strong></p> <ul> <li><strong>OR = 1:</strong> The odds of exposure are the same for cases and controls. There is no association.</li> <li><strong>OR > 1:</strong> The odds of exposure are higher among cases than controls. The exposure is associated with the disease. An OR of 3.2 means the odds of having been exposed are 3.2 times higher for people with the disease than for people without the disease.</li> <li><strong>OR < 1:</strong> The odds of exposure are lower among cases than controls. The exposure is associated with protection from the disease.</li> </ul>

<p><strong>Example:</strong> A case-control study investigates the link between a specific pesticide exposure and a rare form of cancer. Researchers identify 100 people with the cancer (cases) and 200 people without it (controls). <ul> <li>Of the 100 cases, 60 had been exposed to the pesticide (A=60, C=40).</li> <li>Of the 200 controls, 40 had been exposed to the pesticide (B=40, D=160).</li> </ul> OR = (A D) / (B C) = (60 160) / (40 40) = 9600 / 1600 = 6.0 <br/> <strong>Interpretation:</strong> The <strong>odds</strong> of having been exposed to the pesticide were <strong>6 times higher</strong> among those with this cancer compared to those without the cancer.</p>

<hr/>

<h3>Section 3: The Crucial Relationship: When OR Approximates RR</h3>

<p>The interpretation of RR is more intuitive than that of an OR. We naturally think in terms of 'risk', not 'odds'. So, when can we interpret an OR as if it were an RR? This is possible under the <strong>rare disease assumption</strong>.</p>

<p>When the outcome (disease) is rare in the population (generally considered to have an incidence of <10%), the OR provides a good approximation of the RR. </p>

<p><strong>Why?</strong> In the RR formula, RR = [A / (A + B)] / [C / (C + D)]. If the disease is rare, then A is much smaller than B, and C is much smaller than D. Therefore, the denominator (A+B) is approximately equal to B, and (C+D) is approximately equal to D. The formula for RR then becomes approximately (A/B) / (C/D), which simplifies to (AD)/(BC) – the exact formula for the OR.</p>

<p>However, when a disease is common, the OR will always overestimate the strength of the association compared to the RR. That is, for a risk factor (RR > 1), the OR will be greater than the RR. For a protective factor (RR < 1), the OR will be smaller than the RR. This is why it is technically incorrect to interpret an OR from a study of a common disease (e.g., the common cold) as a 'risk ratio'.</p>

Browse public study notes