Jump to comment:

• Vertebroplasty, the child and the bathwater
  Robert B.M. Landewé, Alexandre Sepriano
  Published on: 13 July 2021

• Published on: 13 July 2021

  Vertebroplasty, the child and the bathwater

  • Robert B.M. Landewé, Rheumatologist Amsterdam University Medical Center, The Netherlands
  • Other Contributors:
    • Alexandre Sepriano, Rheumatologist

  Some of the most compelling clinical questions are hardly amenable to experimentation in randomized controlled trials (RCTs). ‘Does vertebroplasty improve health-related quality of life in elderly patients with an acute osteoporotic fracture?’ is one of those questions that was nevertheless challenged in not less than four RCTs recently. The outcome of this challenge was a disappointment for believers in vertebroplasty (VP): one-to-three against VP, and the invasive intervention was discarded from guidelines, as Christian Roux and colleagues have beautifully explained in a recent opinionated review in RMDOpen. [1] Obviously, an unmet need remained and Roux et al. broke a lance for reconsidering VP as a treatment option in highly selected vertebral fracture (VF)-patients with a bad prognosis. They solicited proposals for clinical studies.
  Such studies should not necessarily have an RCT-design. Indisputably, RCTs provide the most unbiased results, but always at the expense of external validity. This is why clinical epidemiologists keep recalling that the absence of evidence (that VP works) does not imply that there is evidence for the absence of efficacy (of VP).
  Roux et al have a point when they claim that the trials may have focused on the wrong population, that the choice of the trials’ primary outcome was not ideal, and that the duration of follow up was too short to detect clinically meaningful effects beyond pain resolution alone. All these objections invol...

  Show More

  Some of the most compelling clinical questions are hardly amenable to experimentation in randomized controlled trials (RCTs). ‘Does vertebroplasty improve health-related quality of life in elderly patients with an acute osteoporotic fracture?’ is one of those questions that was nevertheless challenged in not less than four RCTs recently. The outcome of this challenge was a disappointment for believers in vertebroplasty (VP): one-to-three against VP, and the invasive intervention was discarded from guidelines, as Christian Roux and colleagues have beautifully explained in a recent opinionated review in RMDOpen. [1] Obviously, an unmet need remained and Roux et al. broke a lance for reconsidering VP as a treatment option in highly selected vertebral fracture (VF)-patients with a bad prognosis. They solicited proposals for clinical studies.
  Such studies should not necessarily have an RCT-design. Indisputably, RCTs provide the most unbiased results, but always at the expense of external validity. This is why clinical epidemiologists keep recalling that the absence of evidence (that VP works) does not imply that there is evidence for the absence of efficacy (of VP).
  Roux et al have a point when they claim that the trials may have focused on the wrong population, that the choice of the trials’ primary outcome was not ideal, and that the duration of follow up was too short to detect clinically meaningful effects beyond pain resolution alone. All these objections involve limitations of the rigid RCT-design template. If Roux et al’s hypothesis is right, and VP indeed reduces mortality and improves health related quality of life among these selected elderly at high risk to die, perhaps different study designs with longer follow up, in different patients and with different outcomes, should be pursued. Proponents of RCTs will immediately object that such studies cannot provide the required level of unbiasedness that RCTs may provide, eg. because of the absence of a proper control group, the likelihood of bias by indication and the room for obscuring co-interventions. But as said some clinical research questions are simply not ‘made for trials’, and sometimes the medical community should better appreciate second-best, in order to avoid throwing the child with the bathwater.
  Recently, we have applied principles of causal inference to two compelling research questions that were not suited well for RCTs; the first was whether intense short-term immunosuppression reduces mortality in patients with hyperinflammatory COVID-19. In a non-randomized study with untreated historic controls, and a proper mediator variable, we have construed the evidence that immunosuppressive treatment is indeed effective, [2] half a year before the RCTs and the guideline committees arrived at the same conclusion.
  Whether or not bDMARDs can inhibit syndesmophyte formation in patients with axial spondyloarthritis is a matter of endless scientific debate. RCTs are (ethically) difficult to perform because of slow progression-rates warranting long-term trials, but many non-randomized studies have been done in the past. We have tackled the issue by categorizing the available evidence in the literature according to criteria of causality proposed by Gvozdenovic et al. [3] While this exercise has not provided final evidence, it gave important direction to a question that will likely never be unequivocally answered in an RCT. [4]
  We invite Roux et al. to do a similar exercise in their carefully built database of studies to the efficacy of VP. They may bring some order of causation in the existing body of literature, as we did in the example of axial SpA. By doing so, they may propose a potential mediator-variable, or perhaps an instrumental-variable, that can be used in future clinical studies and shed more light on research questions that go beyond the level of pain-resolution by VP alone.

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top