Although RCT’s are the golden standard for evaluating techniques, retrospective studies such as Krych et al’s study , “Activity Levels Are Higher After Osteochondral Autograft Transfer Mosaicplasty Than After Microfracture for Articular Cartilage Defects of the Knee: A Retrospective Comparative Study”, can be an interesting read.
They compare microfracturing (se previous post) with mosaicplasty. In mosaicplasty you harvest bone and cartilage cylinders (osteochondral transfer) from sites where it’s not needed and transfer it to the injury. The technique is nice and fairly straightforward although getting all the cylinders aligned perfectly so that none are higher/lower is virtually impossible. It can also be a problem that you don’t have enough unused cartilage to harvest.
The reason why Krych’s study is nice is that they implemented a prospective register in 1999 and have therefore good quality follow-up data. I think it’s really praiseworthy that they started a registry that early. These are rare injuries – it took the team 13 years to get this article published!
48 patients from the mosaicplasty treatment were collected, 48 patients were then selected from the larger microfracture treatment pool. The microfracture group was matched on age, sex and location/size of the injury.
Results
They found that the Marx Activity Rating Scale was better in the mosaicplasty group at two-year (p = 0.001), three-year (p = 0.03), and five-year (p = 0.02) . This is interesting since this group have had previously failed microfracture surgery in 1/3 of the cases, and yet they managed to get a higher rating than the more virgin microfracture group!
The other tested scores SF-36 (quality of life), Knee Outcome Survey and IKDC showed no significant difference.
The lack of improvement in the other qualities shows two different interesting things:
- The activity is not related strongly to knee scores or quality of life
- The risk of overestimating due to patient bias may be less if only one quality is significant. In plain English, if a patient is happy to receive a fancier treatment they should score high in all the scores and not just one. Orthopaedic surgery is hard to properly evaluate since so many patients get used to their current activity level. For the patient this is a good strategy but for the study this may be very frustrating.
Some study issues
This is not a double blinded RCT but there are some issues that might have been corrected for:
- Possible selection bias may influence as the ones interested in mosaicplasty had previously had surgery and might be more motivated this time. If you add a covariate in a regression model for those with previous surgery it might give some insight.
- Why did they choose a 1:1 ratio when matching? Even if the power calculations were strong, there is little harm in including more patients in the microfracture group.
- T-test for these data might not be optimal as it is usually not normally distributed. A better approach would be to use bootstrapping or something similar.
- A Bonferroni/Sidak correction could add some statistical weight and limit the issue with multiple testing. It might be hard for the three and five year results but the two year result withstands this correction.
- Time bias – is it possible that calendar time could be an issue as treatments improve/deteriorate since 1999? A regression model with calendar time as a spline would be an interesting addition.