Walking with knee OA

"Wear and tear" is the wrong model. Cartilage is load-adapted tissue, not brake pads: walking delivers the cyclic load that maintains it — in the one trial to vary intensity in osteoarthritic knees, twelve weeks of supervised walking or cycling lowered cartilage breakdown markers at every intensity tested, and most at the highest — while disuse demonstrably thins it. The fear that every step spends down a fixed cartilage budget is the single most damaging misconception in this section, and the cohort data run directly against it.

Walking with OA that already exists

The Osteoarthritis Initiative followed 1,212 adults over 50 with knee OA for four years. Those who walked for exercise had less new frequent knee pain (OR 0.6, 95% CI 0.4–0.8) and less medial joint-space-narrowing progression (OR 0.8, 95% CI 0.6–1.0) than non-walkers. Two other outcomes the same analysis measured — worsening of the overall radiographic grade, and improvement in pain people already had — showed no difference either way.

One stratification in that analysis deserves more attention than it usually gets. Split by knee alignment, the benefit was not uniform. Walkers with varus knees, 48% of the cohort, did better on every progression measure. Walkers with neutral alignment had more medial joint-space worsening than non-walkers (17% against 11%), and walkers with valgus alignment had more radiographic worsening (20% against 15%) and less pain improvement (35% against 48%). The authors call these strata exploratory and ask for replication. It is the one place in this literature where the shape of a knee, rather than the dose, might change the answer.

In a separate cohort of 1,788 people with or at risk of knee OA wearing accelerometers, each additional 1,000 steps per day predicted a 16–18% lower risk of functional limitation two years later; walking at least 6,000 steps a day best separated the people who stayed functional.

The running stress test

If walking wore out knees, running should wear them out faster. It doesn't appear to. A meta-analysis spanning 125,810 people found hip or knee OA in 3.5% of recreational runners versus 10.2% of sedentary controls; only high-volume competitive running trended higher (OR 1.34, with a confidence interval crossing 1). A second meta-analysis (n=114,829) found no overall excess knee OA among runners — and lower odds of knee surgery for OA (OR 0.46). An 18-year prospective cohort found long-distance runners developed no more radiographic knee OA than controls (20% vs 32%, p=0.25), with severe OA numerically rarer in runners.

The limits of the cohort evidence

Every study above is observational. People who feel better walk more — reverse causation — and runners self-select for robust knees; no cohort fully escapes this, which is why the tier stays promising despite six-figure sample sizes. The structural finding in the OAI analysis was borderline, with a confidence interval touching 1.0, and it did not survive one of the two missing-data assumptions the authors tested — the pain finding survived both. Walking there was also recalled at a visit four years after the outcome window closed. The reassurance also has a dose edge: competitive and elite running shows higher OA prevalence (13.3% vs 10.2%), so the all-clear applies to moderate doses. And none of these cohorts included people with large cartilage defects after repair — post-op walking progression follows surgical protocols, not this entry.

Practical dosing

The functional-benefit threshold in the OA cohorts sits around 6,000 steps per day, with additional benefit per extra 1,000 steps. For painful knees, build volume gradually under the pain-monitoring rules in the load-management entry — pain up to about 5/10 that settles by the next morning — and manage flares by adjusting dose rather than stopping.

Open questions

A pragmatic RCT of structured walking in knee OA with structural endpoints has never been run. Step thresholds for post-cartilage-repair patients are unknown, and whether the competitive-running signal reflects dose or selection remains unresolved.