Regarding the sources of bias in randomized controlled trials, all of the following statements are true EXCEPT:

Explanation

## Sources of Bias in RCTs **High-Yield:** Randomization and allocation concealment are two distinct mechanisms that work together to prevent bias. Randomization alone does NOT prevent selection bias; allocation concealment is essential. ### The Critical Distinction: Randomization vs. Allocation Concealment **Key Point:** These are often confused but serve different purposes: - **Randomization** = the method of assigning participants to groups (e.g., computer-generated sequence). It balances known and unknown confounders *in theory*. - **Allocation Concealment** = the mechanism that prevents investigators from knowing or influencing the assignment *before* enrollment. It prevents selection bias in practice. **Clinical Pearl:** A trial can be randomized but not have allocation concealment (e.g., if the randomization sequence is visible in advance). In such cases, investigators can selectively enroll participants based on predicted treatment, introducing selection bias despite randomization. ### Why Allocation Concealment Cannot Be Skipped Without allocation concealment: 1. Investigators may preferentially enroll sicker patients into the treatment group (or vice versa) 2. The randomization sequence becomes predictable 3. Selection bias occurs, even though randomization was used 4. The trial loses its internal validity **Warning:** Do not confuse "randomized" with "properly concealed." Many older RCTs were randomized but lacked allocation concealment and are now considered to have high bias risk. ### Other Biases in RCTs (All Correct Statements) | Bias Type | Definition | Prevention | | --- | --- | --- | | **Attrition Bias** | Differential dropout between groups | Complete follow-up, intention-to-treat analysis | | **Detection Bias** | Differential outcome assessment based on treatment knowledge | Blinding of outcome assessors | | **Performance Bias** | Behavior change due to knowing treatment | Blinding of participants and providers | **Mnemonic:** **SEPAD** — Selection bias, Exclusion bias, Performance bias, Attrition bias, Detection bias. Randomization + allocation concealment prevent selection; blinding prevents performance and detection; complete follow-up prevents attrition.