Correspondence  |   August 2016
Missing Randomization …
Author Notes
  • Charing Cross Hospital and Imperial College, London, United Kingdom (J.P.).
  • (Accepted for publication March 28, 2016.)
    (Accepted for publication March 28, 2016.)×
Article Information
Correspondence   |   August 2016
Missing Randomization …
Anesthesiology 8 2016, Vol.125, 417-418. doi:10.1097/ALN.0000000000001174
Anesthesiology 8 2016, Vol.125, 417-418. doi:10.1097/ALN.0000000000001174
To the Editor:
Wigmore et al.1  report “an association between volatile inhalational anesthesia (INHA) and a reduction in the long-term survival of cancer patients” and the hypothesis that “volatile inhalational agent in anesthesia may augment cancer cell growth.”
We have three criticisms of the link between the hypothesis and the presented data.
  • Omission of important confounding: In the United Kingdom, the choice of total intravenous anesthesia (TIVA) or INHA is at the discretion of the individual anesthesiologist, and each surgeon habitually works with only one or two anesthesiologists. As each surgeon’s patients will have distinct survival characteristics, their outcomes will also be strongly correlated with the anesthesiologist,2  and hence also TIVA/INHA use. The differences reported by Wigmore et al. may, therefore, stem from differences in surgery, rather than the choice of anesthetic, that is, the type of anesthesia may be proxying the identity of the surgeon. (Including the surgeons’ identity as a simple covariate or using a shared frailty Cox model would have reinforced their analysis.)

  • Timing: The article’s survival curves show that the differences between the groups of patients emerged in the first 9 months or so after surgery. Thereafter the curves appear almost parallel. Inclusion of CIs on the Kaplan–Meier plots would have made this clearer. The hypothesis implies that differences in mortality should appear later. So the authors’ own data suggest that some other mechanism was responsible for the observed differences.

  • Choice of survival model: We believe that the choice of a simple multivariable Cox regression model should have been examined more closely. Use of the Cox model makes a strong assumption that the hazard ratio remains constant with time, but this is inconsistent with both the hypothesis and the Kaplan–Meier plots. The authors should have tested the proportional hazard assumption, as the parameter estimates may be invalid. It might have been better to fit a piecewise Cox model with different hazard ratios before and after 9 months or use a fully parametric technique such as accelerated failure time regression.

Wigmore et al. have touched on questions of immense clinical significance. But we are not yet convinced that their data and analysis have implications for clinicians outside their own hospital. We agree with them that prospective research, with randomization of TIVA versus INHA, is needed.
Competing Interests
The authors declare no competing interests.
John Picard, B.A., M.A., D.E.A., B.M., B.Ch., F.R.C.A., Jason Wilson, B.A., M.B., B.S., M.Sc., F.R.C.A. Grad.Stat. Charing Cross Hospital and Imperial College, London, United Kingdom (J.P.).
Wigmore, TJ, Mohammed, K, Jhanji, S Long-term survival for patients undergoing volatile versus IV anesthesia for cancer surgery: A retrospective analysis.. Anesthesiology. (2015). 124 69–79 [Article]
Glance, LG, Kellermann, AL, Hannan, EL, Fleisher, LA, Eaton, MP, Dutton, RP, Lustik, SJ, Li, Y, Dick, AW The impact of anesthesiologists on coronary artery bypass graft surgery outcomes.. Anesth Analg. (2015). 120 526–33 [Article] [PubMed]