I would like to comment on the conclusions of the recent published meta-analysis by Thorlund et al in the BMJ and the subsequent sensationalist editorial. The conclusions are at odds with my own personal experience and that of my peer group.

Although the methodology of the study is valid I would take issue with the conclusions reached.

The meta-analysis on benefit started with nearly 1800 studies and was reduced to 10 (although in the text of the article, only 9 references are cited after this statement). The meta-analysis on harm was on a separate series starting with 2330 studies reduced to 9.

The devil is in the detail and I will address each study individually. I review papers for journals and would certainly not have allowed some of the conclusions reached in the papers, past the peer review stage.

To start with, the 9 (or 10!) studies on benefit (or not) are a mixed bunch dealing with degenerative change (3), degenerative meniscal tears (6 - 2 of which are the same study at different time intervals) or a combination of the two (1). Two papers would never pass an ethics committee in the UK and would not be considered for publication in the UK. One of the papers on meniscal injury is related to depression and anxiety and in an obscure journal. I have excluded this from my analysis as I don't generally perform surgery to improve a patients psyche and commenting on this is outside my sphere of competence.

Benefits (or lack of!)

Partial Meniscectomy

Yim et al - well conducted study but in common with other papers, diagnosis had a large reliance on MRI findings and only 2/3 of eligible patients entered the study. Patients who had any form of bony surgery were excluded. This study was performed in South Korea in a population that is very different, culturally and biomechanically from patients in the West and its findings are only directly applicable to similar populations.

Gauffin et al - This paper states that patients having surgery had significantly less pain than the non-surgery group at 12 months.

Sihvonen et al - this paper includes a sham surgery arm, which would be regarded as unethical in the UK. Of 205 eligible patients 49 were excluded. Interventions were carried out in 5 centres meaning only 30 in each centre. The authors state only meniscal surgery was performed - "no other surgical procedure was carried out". Tucked away in the supplementary appendix is the "findings at arthroscopy" - only 20% of the intervention group had no or mild degenerative change and 33% of the sham surgery group. 54% of the intervention group had degenerative change and 40% of the sham surgery group and an equal number in each group were osteoarthritic. This study only addresses intervention for one part of the knee when the pathology is likely to be multifactorial and in particular where there was no clear identification of symptoms directly attributable to a meniscal tear.

Herrlins et al's 2 papers are also interesting, being a 2 and 5 year follow up in the same cohort of patients. It is perhaps the most disturbing, in that the authors of the meta-analysis neatly discard the results and conclusions of the 5 year study "as the other outcomes are only out to 2 years". Perhaps this is because the conclusions of the 5 year outcomes are very different from those of the 2 year outcomes. Specifically - in the 2007 paper (2 year follow up) the abstract states - "arthroscopic partial medial meniscectomy alone, followed by exercise therapy was not superior to supervised exercise alone". In the 2013 (5 year follow up) paper the authors make the same statement but add "however, one third of patients from the exercise group, still had disabling symptoms but improved to the same level after surgery".

Meniscal tear and Osteoarthritis.

There is only one, widely quoted, paper in this group by Katz et al. Once again there is a large reliance on MRI diagnosis for degenerative change and meniscal tearing. Enrollment was based on one symptom, of at least 4 weeks duration, that might suggest a meniscal tear and a positive MRI. There were 7 centres involved in the study with numbers per centre ranging from119 to 17 patients. Of 1330 patients eligible for the study only 351 were enrolled. Of the 150 patients in the physical therapy group 51 had crossed to the surgery group in the first 6 months and a further 8 did so between 6-12 months. After surgery these patients improved to the level of the initially operated on patients. At 6 months 67% of surgical patients improved by 8 points on the WOMAC scale compared with 43% patients in the physical therapy group (bear in mind a third of these had crossed over to the operated group). In the UK it is very unusual to see a patient with symptoms of only 4 weeks duration and one must ask the reason why 2/3 of the patients declined to enter the study (not defined by the authors). The most likely reason in my experience is many patients will already have had physiotherapy before referral and not surprisingly, wouldn't want to be randomized to a group receiving treatment that had already failed them.

Osteoarthritis

Chang et al's paper is a small series, with 18 in one group and 14 in the lavage group ie both had an intervention. About half of each group had significant improvement at 12 months, but the authors reach the conclusion that "arthroscopic surgery may be beneficial in certain subgroups". This study is smaller than a study by Hubbard and one has to ask why Hubbards paper (see below), with a larger number of patients, wasn't included despite being almost identical in design.

Kirkley et al, again compares conservative management with conservative management + meniscal surgery. Once again 26% of those eligible did not complete the trial. The authors excluded any patient with grade 1 radiographic changes and 50% of the patients had grade 3 or 4 osteoarthritis. Generally in the UK there would be little indication for surgical intervention in more advanced OA unless there are associated mechanical symptoms.

Moseley et al's paper would again not be accepted by an English orthopaedic journal as it includes a sham surgery arm. It has already had a fairly robust rebuttal in an editorial in "Arthroscopy". Of the 324 patients eligible for the trial 144 declined to be enrolled. 25% of the patients had "severe osteoarthritis" - a group that would generally not be considered for surgery in the UK. To quote Bob Jackson "This study was seriously flawed, first by the selection of patients with no clear indication as to how severe the arthritic state was in each case. Second, patients in the Veterans Affairs system do not represent the typical population, which consists of younger people and women (97% of study patients were men). Moreover, VA patients have a vested interest in getting continued benefits for a disability. Third, the statistics used in the paper have been strongly criticized by independent statisticians, as the authors changed their direction on 3 occasions, first using an established device to show superiority of one group over another (which it did not), then using another proven device to demonstrate "equivalence" (which it did not), and then finally using an unvalidated measurement device of their own making, which did prove equivalence of the 3 groups."

Adverse Events

The authors of the meta-analysis boldly state - "Arthroscopy is associated with harms" without any qualification of this statement.

Professor Andy Carr has also recently stated in the BMJ that the death rate from arthroscopic surgery is approximately 1/1000 and deep vein thrombosis as 4 per 1000. Perhaps most worrying is this statement has been published without being referenced and despite emailing him I have not had a reply as to where these figures come from. This is, quite frankly, irresponsible of the BMJ. In my immediate group we have performed approximately 50,000 arthroscopies without a death.

The authors of the meta-analysis have been disingenuous in that they have taken a separate series of studies on complications and then applied them to the separate series of surgical results they have analysed. This cannot be logical or scientifically valid.

There are some large studies but these generally include all arthroscopic surgery, including complex ligament surgery, arthroscopic washout for septic arthritis etc, and not just meniscectomy and debridement.

Maletis et al, analyzing some 21,800 cases , report only one surgically attributable death.

Hetsroni et al quote a 2.8 /10,000 rate of pulmonary embolism but this is strongly associated with increasing age, complexity of surgery and operating time ie is not directly related to simple arthroscopic surgery.

Jameson's study included all arthroscopic surgery including ligament reconstruction, with problems once again associated with more complex surgery. Perhaps the most relevant point in this series is that a lower rate of problems was associated with high volume units. The authors state "Complications following arthroscopy of the knee are rare. It is a safe procedure."

Hame et al's paper states "post operative complications are rare", and this is in a series of patients who were all over 65 - a noted risk factor.

The authors of the meta-analysis also include 2 papers from the analysed "benefit" series in their analysis of adverse events.

Although they include the paper by Sihvonen et al the only adverse event was a deep infection, which occurred 4 months after surgery following a dental procedure and surely can't be directly attributed to the operation.

Perhaps the most interesting paper in their series is the widely quoted paper by Katz et al , which compared surgery against physical therapy followed up for 12 months. They reported 2 deaths - however one occurred in the physical therapy group! Adverse events of mild to moderate severity occurred in 15 patients post surgery and 13 in the physical therapy group. Ie physical therapy seems just as dangerous to the patient.

Discussion

I do not believe the stated conclusions of the meta-analysis can be justified by the evidence provided, and the paper does not seem to have been robustly reviewed or edited in appropriate detail. I hope I have demonstrated this in the above commentary.

In my own experience, appropriate debridement in early osteoarthritis can produce pain relief and improved function for anything up to 4 to 10 years. This has been shown in the literature in a prospective randomized, single surgeon study on consecutive patients - Hubbard MJS JBJS 1996 (for some reason not included in the meta-analysis) - which reports 50% of patients with continued pain relief at 5 years. Aichcroth et al's paper in 1991 (large number with only a small number lost to follow up) showed 75% of patients had minimal discomfort and improved function at 4 years post surgery and only 14% of patients had had to progress to further surgery. Although this study is not randomized, it was prospective and for a group of patients who had significant symptoms of significant duration who had already failed conservative management.

As far as meniscal tear goes, from the papers presented, this would seem to be a valid intervention even in series where I would take issue with some of the diagnostic criteria. It has been clearly demonstrated that diagnosis of clinically relevant meniscal tears is more sensitive and specific in the hands of an experienced knee clinician than MRI. - O'Shea K.J., et al Am J Sports Med 1996, and Gelb et al Am J Sports Med 1996.

To make the diagnosis of meniscal tear based on joint line pain and a positive MRI is clinical laziness at best. Posteromedial tenderness and fullness (often associated with a small parameniscal cyst) which mimics the patients pain, decreased flexion, pain on crouching and pain on sleeping with the knees together, almost always results in immediate pain relief flowing arthroscopic resection of the tear.

Even despite the poor clinical discrimination, 30% of patients failed conservative management with "continuing disabling symptoms" and were provided with pain relief on surgical intervention. An intervention which, in skilled hands is quick, relatively painless, and has a very, very low morbidity.

The authors state "these finding do not support the practice of arthroscopic surgery for middle aged or older patients with knee pain with or without signs of osteoarthritis" without defining the cause of the pain. This statement is not supported by the evidence provided. A summation of erroneous conclusions does not equate to a new "evidence based" conclusion. To baldly state "arthroscopy is associated with harms" with having looked at the complication rate specifically for simple uncomplicated arthroscopic surgery is unacceptable and should not have been allowed through by the editorial team.

The worry is that symptomatic patients will be denied appropriate treatment based on the findings of this "scientific study" which are in reality not scientifically valid. As is so often the case, it is the surgeon performing the procedure and not the procedure itself that is critical in getting a good outcome.

Steve Bollen Consultant Orthopaedic Surgeon.

Conflict of Interest:

None declared