Knee arthroscopy in arthritis: an evidence-practice mismatch

Osteoarthritis, where the cartilage lining a joint gets worn down, is common (Australian data, UK data, US data). Most people will get it if they live long enough, and the knee joint is commonly affected. There is little that can be done to repair or reverse this process, and a related paper that covers many osteoarthritis treatments shows that most of the things we do (analgesics, anti-inflammatory medication, injections etc.) only provide temporary relief, and many of them hardly work at all. Treatment, if severe enough, often means a knee replacement.

Knee replacement surgery is major surgery so it is only reserved for those with severe osteoarthritis. So what do surgeons do with patients who have knee pain and mild or moderate arthritis? They often do an arthroscopy: a low risk, day-only procedure that pays well and seems to work some of the time. Hundreds of thousands are done in the US every year, and in my state the rate of arthroscopy is high and is rising.

The trouble is: it doesn’t work. Most patients still have pain, some get worse, and about 20% will end up having a knee replacement within 2 years anyway. Feel free to skip to the last paragraph for the Bottom Line, or read on for the details.

There are many studies that show that some people feel better for a while after an arthroscopy, and this matches the experience and opinion of many surgeons, but that does not constitute evidence that the arthroscopic procedure (cleaning up the knee and removing debris and torn meniscus fragments) actually improves the patient’s condition. In clinical trials comparing arthroscopy with anything else, arthroscopy never wins.

An early study showed that arthroscopy was not as good as just washing the knee out with a needle, but the bombshell article from Moseley came in 2002, in the New England Journal of Medicine. The researchers compared arthroscopic debridement (‘cleaning up”) and lavage (‘washing out’) with a sham procedure. A sham procedure, in which an incision is made and the patients are blinded (unaware of which treatment they received), is a good way of controlling for the placebo effect of surgery. The researchers measuring the outcomes did not know what group the patients were in, and when they asked the patients which group they thought they were in, they had no clue. So this study was randomly allocated, had a good placebo arm, and involved effective blinding of the patients and the assessors. They measured many outcomes (pain and function) at several times points over a two year period and found that the arthroscopic (active) groups did no better than the placebo group for any outcome at any time point. Criticisms, centred around the ways they measured pain, or on the age or gender of the patients (for example) seem a little desperate.

A later trial from 2008 comparing arthroscopy combined with medical management to medical management alone (without a sham procedure) addressed some of the criticisms of the earlier trial by using validated outcome scores, by including more women and younger patients, and by excluding those with deformity. They showed no difference between the two groups for any of the outcomes, except for a brief improvement in the operative group post-operatively, which was an expected result of the placebo effect of surgery.

When faced with evidence like this, many surgeons state: “Everybody knows that the procedure doesn’t work for everybody. It works for some, and the trick is to do this operation on the subgroup of patients for which the surgery will work.” The problem with this is the reason why it works in some people. It is quite possible that some people improve because of fluctuations in the disease, or expectations, or concomitant treatments, and not because of the procedure. Some patients improved in all of these studies – that doesn’t mean that they improved because of the surgery. The only conclusion we can make on this point is that the patients who had the surgery were no more likely to improve than the patients who did not have surgery.

The subgroups usually targeted by surgeons are those with mild arthritis and those with meniscus tears (or mechanical symptoms). Both of the studies above looked at different subgroups of arthritis severity (and excluded the severe cases) and found no correlation. In Moseley’s article, 172 of the 180 patients had mechanical symptoms and most of the patients in the later article had their torn meniscus removed. And the procedure still didn’t work. The arguments about age and gender are equally invalid, as there is no difference in the results in these groups, and there is no reason to expect a difference. Every way you look at it in every study, arthroscopy doesn’t help the patients any more than NOT doing an arthroscopy, for every outcome in every study.

Yet surgeons still say arthroscopy works for meniscus tears in younger patients. If you want to make the argument that arthroscopy will work in patients aged between 45 and 64 with mild arthritis and a confirmed meniscus tear on MRI, you will need to do a clinical trial to test that hypothesis, rather than just assume that you can pick the winners. Oh, wait: that study has been done. In a 2007 study from Sweden those exact patients were randomised to either an arthroscopy or physical therapy alone. No advantage was shown for those treated with arthroscopy, for any outcome measure, at any time point.

What about pain from arthritis behind the knee-cap, maybe there is a role for arthroscopy for those patients? In this initial study, and the later follow up study, there is no advantage in doing an arthroscopy in those patients.

The Cochrane review of arthroscopy for knee osteoarthritis can be accessed here.

The bottom line

If you have pain and osteoarthritis in your knee, then regardless of the kind of symptoms you have (‘mechanical’ or not), regardless of what your X-rays look like, regardless of where the arthritis is, regardless of how bad your pain is, and regardless of whether or not the MRI scans show your meniscus to be torn, having an arthroscopy will not increase your chances of getting better. It will not arrest or reverse the degenerative changes in your knee, nor will it “create an environment in which healing may occur” (as one surgeon states in his reports in order to justify the procedure). At this point, most patients say: “But what can I do for the pain, it’s really bad?” All I can say is that the severity of your pain does not change the fact that the operation does not work. You will have to try something from the list of (much less expensive) non-operative treatments available. I will say what surgeons seem reluctant to say: “I am sorry, but for this condition, surgery is unlikely to provide any benefit over the non-operative alternatives.”

Addit 19 Oct 2013:
In a multicentre randomised trial published in the New England Journal of Medicine in 2013 (here), patients aged 45 and up with mild to moderate osteoarthritis and a proven meniscal tear were randomised to arthroscopy or physical therapy. The results at 6 months by intention to treat analysis were not statistically or clinically different.
Within 6 months, 6% of those randomised to surgery did not have surgery, and 30% of those randomised to physical therapy had surgery. However, when analysing this study in an "as-treated" manner, it is open to bias. For example, those who believed surgery was better, or who had friends who had surgery who felt better, might not have been "satisfied" with non-operative treatment and were only satisfied when they got what they wanted. This is why blinded, placebo trials are much more effective at differentiating effectiveness between treatment options.

Addit 31 August 2014:
I wrote another post on arthroscopy here, which was prompted by yet another sham surgery trial on arthroscopy for meniscus tear WITHOUT osteoarthritis. Once again, there was no difference.