How cartilage responds to load

Articular cartilage is avascular — no arteries deliver nutrients to it. It is fed by fluid exchange: cyclic compression squeezes fluid through the matrix, and re-expansion draws nutrients back in. Quantitative MRI in living humans shows patellar cartilage giving up 3–6% of its volume during activity, and MRI synchronised to loading shows the deformation is not simple squashing: under half body weight the tibiofemoral surfaces slide against each other and the dominant strains inside the tissue are in shear, approaching 12%. That pump is the reason this whole section exists: load is not the enemy of cartilage, it is the delivery system.

The pump, measured

In vivo imaging studies quantified the deformation directly, and the amount tracks how much the knee bends: patellar cartilage gave up 2.8% of its volume after walking, 4.5% after ten minutes of cycling, 5.0% after a 200 m run and 5.9% after thirty deep knee bends. Chondrocytes transduce this cyclic load into matrix synthesis; without it, the tissue atrophies like unloaded muscle.

The kneecap is not the whole knee, and the same study is a useful corrective there. In the femorotibial compartment — where osteoarthritis actually lives — neither femoral condyle changed measurably after any of those activities, and the tibia only moved, by 6.1–7.2%, after repeated jumps from a 40 cm height. Everyday loading deforms the joint surface that faces the kneecap; it barely touches the surfaces that carry body weight.

Whether that pump remodels the tissue over months is a separate question from whether it deforms it in the moment, and this study answers only the second. Professional weight lifters and bobsleigh sprinters deformed the same amount as untrained men (2.9% and 3.9% against 4.1%), which its authors read as no evidence that adult cartilage properties are amenable to training in a way this method can see.

What happens when you load it

The flagship human trial is small but pointed: 56 adults at high OA risk after meniscectomy were randomized to four months of supervised exercise or to no intervention, and 30 were scanned at both ends. Cartilage glycosaminoglycan content on dGEMRIC imaging improved in the exercisers and declined in controls (+15 vs −15 msec, p=0.036), and the GAG gain tracked how much each person's activity actually changed (r = 0.70, 95% CI 0.31–0.89).

What the same trial did not find is worth carrying with the result. No KOOS subscale separated the groups, and of five objective performance measures only the one-leg jump did (+17 vs +7 cm, p=0.009) — knee-extensor torque and aerobic capacity did not move. The trial was sized for the scan, not for symptoms, and what it demonstrates is a compositional change rather than a clinical one.

The safety side is broader. Across 9 RCTs — 14 comparisons, two in people at increased risk and twelve in people who already had knee OA — knee-loading exercise showed no evidence of cartilage harm on imaging: six comparisons null on thickness, volume or defects, two positive and two null on collagen, and on glycosaminoglycans one positive and one negative. Across 12 RCTs measuring cartilage and inflammation biomarkers, exercise lowered them in 30% of comparisons, left them unchanged in 63%, and raised them in only 7%.

What happens when you don't

Disuse is not neutral. After spinal cord injury — complete unloading — knee cartilage thinned progressively, roughly 19–25% across compartments within 24 months. Even seven weeks of partial weight-bearing on crutches thinned healthy cartilage by 2.9–6.6% and shrank the quadriceps 11%. Total rest is an active harm state for this tissue, not a holding pattern.

The dose curve

One trial has put this question to people with osteoarthritis and measured cartilage while doing it. A hundred and twenty adults with Kellgren-Lawrence grade 2 or 3 knees were randomised to walking or cycling three times a week for twelve weeks at 50–60%, 60–70% or 70–80% of heart rate reserve, or to no exercise. Serum CTX-II — a cartilage breakdown marker — fell furthest in the hardest-working group: down 28.5% at high intensity, 18.7% at moderate, 8.2% at low, and not significantly at all in the control group. COMP followed the same order. So did symptoms: WOMAC and pain improved most at high intensity, where 73% and 80% of participants crossed the threshold for a difference they would notice. MRI cartilage thickness and volume rose slightly, and most in that arm too.

Within the range anyone has tested in a human knee, then, the relationship is monotonic rather than an inverted U: more intensity, less cartilage turnover. Three limits keep that from being the last word. The trial is thirty people per arm over twelve weeks, so it describes turnover markers and short-term symptoms rather than whether a joint lasts longer. Its highest arm stopped at 80% of heart rate reserve — it establishes that the down-slope is not inside the range it tested, not that there is no down-slope. And the cost side moved in the same direction as the benefit: transient knee pain affected two, three and five participants across the three arms, muscle soreness four in the hardest, and adherence fell from 92% to 85% as intensity rose. Nothing serious happened and nobody dropped out because of the exercise.

The practical reading is narrower than either "moderate is optimal" or "harder is better": regular, progressive loading, at an intensity you will actually keep doing.

The limits of this evidence

The dGEMRIC trial is one small, single-center study with a surrogate imaging outcome, never replicated at scale; the exercise group also began higher on KOOS Symptoms (90 vs 81, p=0.047). The exercise-cartilage systematic review found mostly null effects on cartilage morphology — "not harmful" is the defensible summary, not "regenerative" — and its own authors graded that evidence low, for too few studies with too few participants. The disuse studies are observational and confounded (spinal cord injury involves more than unloading), and 7% of biomarker comparisons did move in an unfavorable direction. Open questions include whether the GAG gain is reproducible and durable, where the ceiling sits above the 80% of heart rate reserve anyone has actually tested, and whether repaired cartilage after MACI or microfracture responds to load the same way.

The practical takeaway holds regardless: motion and load are inputs cartilage needs. The question is dose — covered in the load-management entry — not whether to load at all.