Application to screening programs

Core Metrics - The Diagnostic Foundation

• Sensitivity (Sn): The ability of a test to correctly identify individuals with the disease.

  • Formula: $Sn = TP / (TP + FN)$
  • Use: High sensitivity is crucial for screening tests to avoid missing cases (low false negatives).
  • 📌 SNOUT: A highly SeNsitive test, when Negative, helps to rule OUT disease.

• Specificity (Sp): The ability of a test to correctly identify individuals without the disease.

  • Formula: $Sp = TN / (TN + FP)$
  • Use: High specificity is vital for confirmatory tests to avoid false alarms (low false positives).
  • 📌 SPIN: A highly SPecific test, when Positive, helps to rule IN disease.

⭐ Sensitivity and Specificity are intrinsic to the test and are not affected by disease prevalence.

Predictive Values - Prevalence's Impact

• Predictive values depend heavily on disease prevalence in the target population.
• Sensitivity and Specificity are intrinsic test properties and do not change with prevalence.

• Positive Predictive Value (PPV):

  • Probability of having the disease with a positive test.
  • Directly proportional to prevalence: As prevalence ↑, PPV ↑.

• Negative Predictive Value (NPV):

  • Probability of being disease-free with a negative test.
  • Inversely proportional to prevalence: As prevalence ↑, NPV ↓.

⭐ In low-prevalence settings, even tests with high specificity have a low PPV. A positive result in a general population screen for a rare disease is more likely to be a false positive.

ROC Curves - Finding the Sweet Spot

• Function: Plots test performance across all cut-off points. Compares $Sensitivity (TPR)$ vs. $1 - Specificity (FPR)$.
• Axes:
  • Y-axis: Sensitivity (True Positive Rate)
  • X-axis: 1 - Specificity (False Positive Rate)
• Area Under the Curve (AUC):
  • Represents overall accuracy.
  • AUC = 1.0: Perfect test.
  • AUC = 0.5: Useless test (the diagonal "line of chance").
• Optimal Cut-off Point:
  • Typically the point on the curve nearest the top-left corner (0,1).
  • This point maximizes the Youden Index: $(Sensitivity + Specificity) - 1$.

curve showing multiple cut-off points and the area under the curve)

⭐ An AUC >0.8 suggests a good test; 0.7-0.8 is acceptable. An AUC <0.7 has poor diagnostic ability.

Screening Biases - The Hidden Traps

• Lead-time bias: Early detection artificially inflates survival time, even if mortality is unchanged. The "survival clock" simply starts earlier.
• Length-time bias: Screening tends to detect more slow-growing, indolent cases with a better prognosis, while missing rapidly progressive, aggressive diseases.
• Selection (Volunteer) bias: Health-conscious individuals are more likely to participate in screening, leading to better outcomes unrelated to the screening itself.

⭐ A randomized controlled trial is the gold standard to mitigate these biases and prove a screening test reduces mortality.

High‑Yield Points - ⚡ Biggest Takeaways

• Screening tests require high sensitivity to rule out disease (S-N-OUT) and avoid false negatives.
• Confirmatory tests demand high specificity to rule in disease (S-P-IN) and minimize false positives.
• Sensitivity and specificity are intrinsic to the test and unaffected by prevalence.
• Predictive values (PPV & NPV), however, are heavily dependent on disease prevalence.
• PPV is the most important measure for a positive screening test result's clinical utility.
• Adjusting a test's cut-off point creates an inverse relationship between sensitivity and specificity.