TL/DR: The evidence base is bad; the authors then make several choices that make an already difficult evidence synthesis less reliable and interpret the resulting uncertainty too negatively.
Okay just getting through the paper on a first read. Haven't gotten the supplemental data looked at.
Poor study in some ways but limited by the input studies and author goals/biases more than anything else.
They try to find RCTs studying genicular RF of the three main targets, but then decide to include pulsed and bipolar RF.
They make some decisions in their statistical analysis that effect the signal greatly such as using SMD for pooled VAS. They used the WOMAC total score (pain/stiffness/function) as their output metric, which is fine, except for genicular RF only directly mechanistically effects one of those.
They tried to do some subgroup analyses
This is figure 2 for them which is the VAS. I want you to see that really, the spread in the studies is impressive, but at A (3m) and B(6m), it really isn't that clear there isn't an effect for RFA. It just isn't statistically significant but that's more the analysis I suspect than the reality. At 12 months, it's clear there isn't a durable effect.
Real quick, this is from the Malaithong paper which is old but shows their bipolar targeting.
It's suboptimal for a few reasons, primarily related to not being proximal enough IMO, and the spacing there is pretty wide to get a good bipolar lesion, the RF is 90 degrees at 180s which may help, but they inject 2 mLs of local+steroid per target before their burn. The senior author SC on that paper likely leans into the idea that fluid modulation allows him to get a bigger lesion, but that much fluid around your burn can create unreliable burns and doesn't fix the issue that non-internally cooled electrodes create limited width burns due to the charring/conduction effects. With that trajectory, width is the critical variable for durability so they're limited to what you can get with 18g needles, and if they're too far apart, the bipolar effect won't really kick in to create one long lesion but rather two thinner lesions further apart.
I also need to point out a few inherent biases. Now days folks some senior authors are doing meta-analyses to try to generate publications but, in some cases, they need to know when there isn't good enough data for it. They registered the study in PROSPERO to suggest they thought this all out before they did it, but unfortunately, they don't actually show that. It started out as let's do a review about knee stuff, and then they focused in on genicular RF, but decided to include pulsed RF which really weakens the question. They then failed to actually describe their stats plan a priori, which can create a real post hoc review design bias.
PROSPERO
www.crd.york.ac.uk
It's always hard to do a meta-analysis when authors are also the ones who published some of the papers included, or actively doing NIH funded trials on this for example with SC's SKOAP work which is self-cited throughout. Interestingly, they don't acknowledge that in the COI section because sure this idea is funding my career/time off from clinical duties, but it doesn't drive my thought process at all since it isn't dirty industry bias. They actually do a sub-group analysis around that 'influence of industry sponsorship' showing
"Supplementary Appendix 5 reveals a significant reduction in pain scores at three months with gRF compared to controls (SMD = -0.97; 95% CI: -1.53 to -0.42; P = 0.020) for the industry-sponsored studies. A nonsignificant difference was observed in the two non-industry-sponsored studies available for meta-analysis (SMD = -1.00; 95% CI: -8.42 to 6.43; P = 0.340). "
I just want to point out the problem here with statistics. The SMD or standardized mean difference here is identical. The p-values are significant for the industry trials but not the non-industry studies. That really just is a precision issue driven by the fact that industry likely runs trials better than your average academic.
Anyway, too much ****, but the real conclusion is that we have good evidence for 3 months of relief, 6 months and 12 months are more uncertain, but the effects appear positive and require more reproduction.
pubmed.ncbi.nlm.nih.gov
