Arthroscopic partial meniscectomy for a degenerative tear

Arthroscopic partial meniscectomy is a short operation: the surgeon trims the torn part of a meniscus back to a stable rim and leaves the rest. The reasoning is intuitive. A flap of torn tissue is catching in the joint, so remove the flap.

That reasoning has been tested more rigorously than almost anything else in orthopaedic surgery, including in the way surgical procedures are hardest to test — against a sham.

The sham-controlled trial

FIDELITY randomised 146 adults aged 35 to 65 with a degenerative medial meniscus tear, confirmed on both MRI and arthroscopy, and no knee osteoarthritis. Everyone went to theatre and had a diagnostic arthroscopy. Then, in the same operation, half had the tear resected and half had a simulated procedure matching the sounds, the sensations and the time in the operating room. Nobody — patient, carer or assessor — knew which.

At two years WOMET had improved 27.3 points in the resected group and 31.6 in the sham group — a between-group difference of −4.3 (95% CI −11.3 to 2.6). Lysholm improved 23.1 and 26.3. Knee pain after exercise improved 3.5 and 3.9. Every between-group interval spanned zero, and every point estimate that was not zero pointed away from surgery.

The mechanical-symptom question, asked directly

Catching and locking are the reasons most often given for operating, so the trial prespecified them as a subgroup rather than going looking afterwards. About half of each arm reported them before surgery: 32 patients (46%) in the resection group and 37 (49%) in the sham group.

Within that subgroup, no primary or secondary outcome separated the two arms. The same held for the other prespecified subgroup, people whose tear was unstable at arthroscopy — the other patient the operation is supposed to be for. The trialists' own summary is the flat one: no evidence could be found to support the prevailing idea that patients with mechanical symptoms, or with particular tear characteristics, or who had already failed conservative treatment, are more likely to benefit.

The interesting reading is not that the operation failed these patients. It is that resecting the tear did not relieve the symptom the tear is assumed to be causing, which puts the assumption itself in question.

Five years, and what the structure did

Both arms of FIDELITY were still doing well at five years, and still doing equally well: WOMET, Lysholm and pain after exercise all sat within a couple of points of each other.

The radiographs moved. Forty-eight of 67 resected knees (72%) and 44 of 74 sham knees (60%) had progressed at least one Kellgren-Lawrence grade — an adjusted absolute risk difference of 13%, with a 95% confidence interval from -2% to 28%. The OARSI sum score moved in the same direction by 0.7 points, and that interval did not include zero. Mechanical symptoms at five years were reported by 20 of 68 patients (29%) after resection and 9 of 74 (12%) after the sham, a risk difference of 18%.

Read that carefully, because the trialists themselves flag what it is measuring. Both instruments — Kellgren-Lawrence and the OARSI atlas — are, in their words, osteophyte-driven grading systems, chosen partly because that makes them more resilient to the measurement noise in joint space width. So the structural signal here is real and it is randomised, and it is largely a signal about bone spurs rather than a direct measurement of cartilage loss. It is the strongest structural evidence this literature has, and it is weaker than "the operation wears out the cartilage".

The trials against exercise and physical therapy

Three separate teams asked a different question — surgery or conservative care — and arrived in the same place.

MeTeOR randomised 351 patients aged 45 and over with a meniscal tear and mild-to-moderate osteoarthritis. WOMAC physical function improved 20.9 points with surgery and 18.5 with physical therapy. Thirty percent of the physical-therapy arm crossed over to surgery within six months, and once they did, their twelve-month function matched the surgical arm's.

OMEX randomised 140 people, mean age 49.5, nearly all without definitive radiographic osteoarthritis, to twelve weeks of supervised exercise or to surgery alone. The two-year difference in KOOS4 was 0.9 points (95% CI -4.3 to 6.1). Thigh muscle strength improved in the exercise arm at three months, which is the one place the trial separated cleanly — and it separated in favour of not operating.

ESCAPE randomised 321 patients aged 45 to 70 in nine Dutch hospitals, as a noninferiority trial. At five years IKDC had improved 29.6 points after surgery and 25.1 after sixteen sessions of exercise-based physical therapy — a difference of 3.5 points against a noninferiority threshold of 11. Radiographic osteoarthritis progressed at comparable rates.

The pooled evidence, and the harms

Nine randomised trials of arthroscopic surgery for the degenerative knee have been pooled. Combining each trial's own primary endpoint between three and 24 months gives a pain effect size of 0.14 (95% CI 0.03 to 0.26). It is statistically significant and it corresponds to 2.4 mm on a 100 mm pain scale — a quantity below what patients can reliably detect. Split by time, the benefit is 3 to 5 mm at three and six months and gone from then to two years. Physical function does not separate at all (effect size 0.09, 95% CI -0.05 to 0.24).

The same review counted what the operation costs, which the trials individually are too small to see. Symptomatic deep vein thrombosis occurs at 4.13 events per 1,000 procedures (95% CI 1.78 to 9.60), alongside pulmonary embolism, infection and death. Those figures come from cohorts and registries rather than from randomised arms, which is a real limitation and also the only way a complication that rare gets counted at all.

Placed against the rest of orthopaedics, this is not an isolated verdict. An umbrella review of level 1 evidence across ten of the commonest elective orthopaedic operations found trial evidence of benefit over non-operative care for two of them — carpal tunnel decompression and total knee replacement. Partial meniscectomy sits in the group of six where the trials show no benefit.

Where the line falls

The conventional defence of this operation is that the trials enrolled the wrong patients: degenerative tears in middle-aged knees, rather than traumatic tears in young ones.

That population has been randomised once, and the trial tested timing rather than necessity. A hundred patients aged 18 to 45, mean age 35, a third of them competitive or elite athletes, with a recent traumatic isolated tear and no osteoarthritis, were assigned to early surgery or to physical therapy with the option of delayed surgery. At two years both groups scored 78 on IKDC. Forty-one percent of the physical-therapy arm took the delayed operation, and both arms improved by a clinically meaningful margin.

So the young traumatic tear has one trial, and what it establishes is that waiting costs nothing measurable at two years — not that the operation is never needed. Locked knees have never been randomised at all, and root tears are a different operation with a different literature, covered in the repair entry.

Reading a scan against the base rate

One number is worth holding alongside all of this. In an unselected Framingham sample aged 50 to 90, 61% of the people found to have a meniscal tear on MRI had reported no knee pain, aching or stiffness in the previous month. Among people who did have radiographic osteoarthritis, a tear was present in 63% of those with pain and 60% of those without.

A tear on the scan and pain in the knee are both common in the same people, and the scan cannot tell you which one is doing the talking.

What would change this

The variable the surgeon actually controls — how much meniscus comes out — has never been randomised, while the reviewed cadaver relationship between resection volume and contact stress is roughly linear and the observational literature consistently ranks it among the strongest predictors it has. The osteophyte-versus-cartilage split at five years needs a second trial to confirm it. And nobody has enrolled a truly locked knee, which is the presentation every guideline still treats as surgical and which no randomised evidence covers. Any of the three would tell a reader something the current evidence cannot.