Calculating Mean and SD of Change from Baseline Using Correlation

Change from Baseline Calculator

Estimate mean change and SD of the change using correlation (r).

Baseline Mean:
Baseline SD:
Follow-up Mean:
Follow-up SD:
Correlation (r):

🎯 Purpose

In clinical trials and observational studies, it's often necessary to compare change scores (e.g., improvement in blood pressure or lab values from baseline to follow-up). However, most studies report only mean and SD at baseline and follow-up, not the SD of the change.

This method estimates the mean change and its standard deviation (SD) using known formulas that incorporate an assumed correlation (r) between baseline and follow-up measurements.

🧩 Required Inputs

For each group (e.g., treatment or control), you need:

${\overline{X}}_{baseline}$ : Mean at baseline
${\overline{X}}_{followup}$ : Mean at follow-up
${SD}_{baseline}$ : Standard deviation at baseline
${SD}_{followup}$ : Standard deviation at follow-up
$r$ : Correlation between baseline and follow-up (usually estimated or assumed)
$n$ : Sample size

🧮 Step-by-Step Calculations

✅ 1. Mean Change

${\overline{X}}_{change} = {\overline{X}}_{followup} - {\overline{X}}_{baseline}$

This gives the average improvement or deterioration over time.

✅ 2. Standard Deviation of Change

${SD}_{change} = \sqrt{{SD}_{baseline}^{2} + {SD}_{followup}^{2} - 2 \cdot r \cdot {SD}_{baseline} \cdot {SD}_{followup}}$

Where:

r: Pearson correlation between baseline and follow-up measurements

If r is unknown, use a conservative estimate (e.g., 0.5) or perform sensitivity analyses using $r = 0.25$ , $r = 0.5$ , or $r = 0.75$ .

🔢 Example Calculation

Inputs:

Baseline mean = 140, SD = 10
Follow-up mean = 125, SD = 12
Correlation rrr = 0.6

Step 1: Mean Change

${\overline{X}}_{change} = 125 - 140 = - 15$

Step 2: Standard Deviation of Change

${SD}_{change} = \sqrt{10^{2} + 12^{2} - 2 \times 0.6 \times 10 \times 12} = \sqrt{100 + 144 - 144} = \sqrt{100} = 10$

So:

Mean change = −15
SD of change = 10

🧠 Interpretation Notes

Negative mean change typically indicates improvement (e.g., lower BP).
The higher the correlation, the lower the SD of the change.
Sensitivity analyses using multiple r-values are advisable when r is unknown.

🧭 Clinical Application

Essential for calculating effect size (Cohen’s d) for change-from-baseline designs.
Allows reconstruction of missing change data for meta-analyses.
Accepted in Cochrane reviews and RevMan5 when direct SD of change is unavailable.