414 - Correspondence
8. Harrois A, Huet O, Duranteau J. Alterations of mitochon-
drial function in sepsis and critical illness. Curr Opin
Anaesthesiol 2009; 22: 143e9
© 2017 British Journal of Anaesthesia. Published by Elsevier Ltd. All rights reserved.
Clinically significant reductions in morphine consumption need to
take account of baseline risk: presentation of a novel meta-analysis
methodology
B. Doleman*, J. N. Lund and J. P. Williams
University of Nottingham, Nottingham, UK
*Corresponding author. E-mail: dr.doleman@gmail.com.
EditordIn the pages of the pain special section in the British
Journal of Anaesthesia October 2017 edition, we read with in-
terest an article presenting a meta-analysis of pregabalin for
acute pain.1 Such an analysis complements a wealth of similar
reviews published over the years on the subject. Whilst this
review had advantages over those published previously, by the
use of both low risk of bias trials and trial sequential analysis
reducing type I and II errors, the conclusion that pregabalin
‘may have minimal opioid sparing effect’ may be questioned.
The reason for this is illustrated by recently developed sys-
tematic review methodology from our group that such con-
clusions need to be taken in context with baseline risk (mean
control group morphine consumption).2
The issue of baseline risk is pervasive within clinical
medicine and underpins guidelines on a range of medical
conditions and their associated treatments including anti-
coagulation for atrial fibrillation and treatment for osteopo-
rosis. This issue has also been demonstrated in a pain
population, with patients with more pain experiencing greater
reductions in pain scores after analgesic treatment.3,4 This
dependency on baseline risk causes issues for the interpreta-
tion of meta-analyses of analgesics for postoperative pain, as
reviews that include trials with high baseline risk (high opioid
consumption) will show greater benefits, largely independent
of the analgesic effect of the agent under study. These differ-
ences for the same agent (gabapentin) can be as different as 24
h morphine reductions of 1.8 mg vs 36.8 mg for the lowest and
highest baseline risk trials, respectively.5 Furthermore, these
issues with baseline risk can also invalidate subgroup ana-
lyses, as each subgroup’s morphine reductions will also be
dependent on baseline risk (for more detail see our previous
discussion).5 In view of these issues, our method was devel-
oped which utilises meta-regression estimates (both fre-
quentist and Bayesian) to allow a particular morphine
reduction to be estimated from any given baseline risk for a
population of patients from a consumer’s surgical list (which
could be estimated from local audits). Moreover, further
covariates can be added to these models (including risk of bias)
with baseline risk held constant to reduce confounding in
subgroup analyses.
9. Belay T, Sonnenfeld G. Differential effects of catechol-
amines on in vitro growth of pathogenic bacteria. Life Sci
2002; 71: 447e56
doi: 10.1016/j.bja.2017.12.011
Advance Access Publication Date: 8 January 2018
To illustrate the problem this causes in the interpretation
of the meta-analysis of Fabritius and colleagues1 we re-
analysed their data from low risk of bias trials for 24 h
morphine consumption on which their main conclusions are
based. Their original estimate was a 5.81 mg morphine
reduction in 24 h, which is indeed of questionable clinical
significance. However, when performing meta-regression of
baseline risk (control group consumption in milligrams), 86%
of the between-study heterogeneity is explained by baseline
risk (P ¼ 0.004). The analysis shows that for surgeries where
the average consumption is around 35 mg day1 then indeed,
clinically significant reductions of 10 mg can be achieved. This
issue also becomes important when comparing different an-
algesics/subgroups (see Table 3)1 as, again, morphine re-
ductions from any agent will depend on the baseline risk of the
included trials for each analgesic rather than the efficacy of
the agent under study.
In view of the above, it was also interesting to read the
editorial in the same issue from the PROSPECT group regarding
procedure-specific evidence.5 We agree that there is certainly
a role that type of surgery plays in the response to analgesics
(and more specific to local anaesthetic techniques). However,
when we analysed 344 randomised controlled trials of multi-
modal analgesics, type of surgery was not an independent
predictor of morphine reductions when added to the model
with baseline risk. Put simply, the evidence for better anal-
gesic efficacy in certain types of surgery may be a product of
differing baseline risk (some surgeries resulting in higher
postoperative morphine consumption) rather than the type of
surgery per se. Indeed, other factors that influence post-
operative morphine consumption such as local patient factors
and concurrent analgesic strategies also contribute to baseline
risk and are controlled for with our models. Whilst we agree
with many of the issues raised in the editorial, we would
argue, as the title of the article suggests, there is a need for a
re-evaluation, although this comes from a consideration of
baseline risk, rather than just the type of surgery.
We welcome the updated evidence from Fabritius and
colleagues and particularly the advantages alluded to previ-
ously. Indeed, the finding of increased serious adverse events

with pregabalin requires serious re-evaluation for its use in
multimodal analgesic protocols (which may also be subject to
baseline risk and the issues highlighted above, for example in
elderly populations). However, the statement that a 10 mg
reduction was excluded does not consider the specific clinical
contexts where these reductions may be achieved. This is of
no fault of the authors based on established meta-analysis
methodology. Despite this, we hope this letter has illustrated
the limitations of current meta-analysis methods (both sta-
tistical and clinical) and advise future reviews take account of
baseline risk to both reduce confounding from variable base-
line risk (using meta-regression) and allow consumers to
select the specific circumstances where clinically significant
reductions may be achieved.
Declaration of interest
None declared.
References
1. Fabritius ML, Strøm C, Koyuncu S, et al. Benefit and harm of
pregabalin in acute pain treatment: a systematic review
© 2017 British Journal of Anaesthesia. Published by Elsevier Ltd. All rights reserved.
Correspondence - 415
with meta-analyses and trial sequential analyses. Br J
Anaesth 2017; 119: 775e91
2. Doleman B, Sutton AJ, Sherwin M, Lund JN, Williams JP.
Baseline morphine consumption may explain between-
study heterogeneity in meta-analyses of adjuvant analge-
sics and improve precision and accuracy of effect esti-
mates. Anesth Analg 2017. http://dx.doi.org/10.1213/ANE.
0000000000002237. Advance Access published on July 4,
2017.
3. Averbuch M, Katzper M. Severity of baseline pain and de-
gree of analgesia in the third molar post-extraction dental
pain model. Anesth Analg 2003; 97: 163e7
4. Bjune K, Stubhaug A, Dodgson MS, Breivik H. Additive
analgesic effect of codeine and paracetamol can be detec-
ted in strong, but not moderate, pain after Caesarean sec-
tion. Acta Anaesthiol Scand 1996; 40: 399e407
5. Joshi GP, Kehlet H, on behalf of the PROSPECT Working
Group. Guidelines for perioperative pain management:
need for re-evaluation. Br J Anaesth 2017; 119: 703e6
doi: 10.1016/j.bja.2017.12.012
Advance Access Publication Date: 8 January 2018