Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Surgery for arthritic knees

Randomised trial
Size and bias

When medical therapy has failed to relieve pain of knee osteoarthritis, an additional treatment often used is lavage or debridement using arthroscopy. The evidence on which this has been based was a number of case series studies, showing that about half of patients so treated did reasonably well in the long term. A number of randomised trials using a variety of techniques and comparators have tested efficacy of the techniques. Most, but not all, claimed efficacy for lavage and/or debridement.

There are problems with these studies. All were small, some were of short duration, and some used surrogate end points like muscle strength. None used a true 'placebo' control of sham operation. The accumulation of evidence that these procedures were beneficial is neither large, nor compelling.

So when a new randomised trial, superbly conducted and relatively large pops up and tells us these procedures do not work [1], we need to put our thinking caps on. We need to balance what we think we know with what someone is telling us we should know.

Randomised trial

Patients recruited were under 75 years of age with American College of Rheumatology criteria for osteoarthritis of the knee, with pain of ≥40 mm on a 100 mm scale (moderate), and with no arthroscopy for at least two years. They were randomised into three groups:

A single experienced surgeon who had been physician to the 1996 US Olympic basketball team carried out all the operations. Postoperative outcome assessments were made by other personnel, who, like the patients, were unaware of the procedure. Information on pain and other outcomes was collected for up to two years after the operation.


Patients were predominantly men (over 90%), with a mean age of about 52 years, mostly taking non-prescription analgesics, but whose osteoarthritis was moderate or severe in about 70%. There were 60 per group at the start, with good follow up to two years.

At baseline the mean knee pain score was 65/100 mm in all groups, and was about 50/100 mm at one and two years in all three treatment groups. No outcome showed any difference between placebo and the two active treatments.


This was a trial with exemplary methods, showing conclusively that neither arthroscopy with lavage nor lavage plus debridement had any effect on pain or function. Yet among previous trials there was at least one [2] with very good methods, being properly randomised and with a blinded observer, and showing benefit for both intra-articular corticosteroid and lavage. It was small, though, with as few as 20 patients per group, and follow up was for only 24 weeks, by which time differences between groups were much less marked.

What were the differences between them? The American study [1] was in men, predominantly in their 50s. The French study [2] was mostly in women, and predominantly in their 60s. In both studies the initial pain intensities were about the same, and variability was great. Typically a mean pain score of 50 mm had a standard deviation of 20-30 mm. That means that the 95% confidence interval was 0-100 mm. Bandolier suspects, by analogy with other pain trials, that these data were not normally distributed, yet were reported as means. This would complicate things even more.

Size and bias

So what is the truth? Do we trust custom and practice, or high quality evidence? Is it all down to differences in patients selected and techniques, or is there something we can grasp? The answer may be size and avoidance of bias.

Trials of small size have been the norm until recent times, and a survey has shown that randomised trials with over 100 participants became the norm only very recently [3]. Small size is a real problem because random chance can dominate. And many of the older studies favouring arthroscopy had, in addition to small numbers, designs that might well have allowed significant bias in favour of it. Small studies with poor design, of short duration and with outcomes lacking relevance can fool us into thinking it works. Older trials that used non-treatment control groups showed that untreated patients improved just as much.

With this new trial what we can say is that on average there is no convincing evidence for any beneficial effect of arthroscopy with lavage and debridement. Proving conclusively the negative is a bit more difficult. But with an estimated 650,000 procedures costing $3.5 billion each year in the USA alone, it certainly makes one think. Bandolier has ordered all the controlled trials it can find, and will abstract them on the Internet site in the coming months.


  1. JB Moseley et al. A controlled trial of arthroscopic surgery for osteoarthritis of the knee. New England Journal of Medicine 2002 347: 81-88.
  2. P Ravaud et al. Effects of joint lavage and steroid injections in patients with osteoarthritis of the knee. Arthritis & Rheumatism 1999 42: 475-482.
  3. S McDonald et al. Number and size of randomized trials reported in general health care journals from 1948 to 1997. International Journal of Epidemiological Association 2002 31: 125-127.
previous or next story in this issue