The benefits of steroids versus steroids plus antivirals for treatment of Bell’s palsy: a meta-analysisBMJ 2009; 339 doi: https://doi.org/10.1136/bmj.b3354 (Published 07 September 2009) Cite this as: BMJ 2009;339:b3354
All rapid responses
Re: The benefits of steroids versus steroids plus antivirals for treatment of Bell’s palsy: a meta-analysis
So at last, one and a half year after my second rapid response, a correction was made to the manuscript: “The benefits of steroids versus steroids plus antiviral medication for treatment of Bell’s palsy: a meta-analysis.”
The authors now agree that a small miscalculation caused the pooled odds ratio to rise from a mere positive to significant, with a lower limit of the confidence interval above 1. The authors state ‘just’ above 1. Why ‘just’? A positive pooled odds ratio for combination therapy including antiviral medication in a disease with a probable viral cause is what we might have expected beforehand, isn’t it? So why not just state: combination therapy significantly promotes recovery.
When one reads the further the correction of the manuscript it becomes clear that the authors want to hang on to their previous, albeit incorrect, conclusion.
Once the corrections were made, it became difficult to stick with the previous conclusion, which prompted the authors to come up with the ‘intention-to-treat analysis.’ This ‘intention-to-treat analysis’ was not mentioned in their first draft. From where I stand, it appears to be a bit of ‘changing the rules after the match’. Moreover, because of the already high spontaneous recovery rate the intention-to-treat analysis more than likely dilutes the effect of treatment. (Notice that the effect of treatment is already diluted by including patients of all severity!)
Furthermore, what really strikes me is that the authors of the original paper do not respond to any of the other points that were raised in the two rapid responses.
So, let me try to explain the ‘case’ once again. And, I sincerely hope that the authors would be willing to respond at this time. And perhaps, they would even explain exactly why it is that they disagree. That might shed some light on how they actually reached their latest conclusion, because I simply cannot understand how they managed that at all.
a) The facial nerve passes through a bony canal
b) Bell’s palsy, though idiopathic, is probably caused by a viral infection
c) The probable pathogenesis: an inflammatory reaction, a combat response that causes swelling of the facial nerve, which subsequently becomes entrapped within the bony canal
d) 71 -85% of patients will recover quite well without therapy
e) Corticosteroids enhance recovery
f) Elderly patients and patients with an initial severe paresis might have an elevated risk of failure to recover
g) Research should preferably focus on those groups that are at an increased risk of failure to recover.
h) A retrospective study by Hato et al., in 2003, showed significantly increased improvement in the combination therapy group in the group at risk
i) Although the results of retrospective studies, including the Hato study, might be influenced by a lot of methodological flaws, there is a good chance that what we see still is a genuine effect due to the pathogenesis. A priori, antivirals are likely to be beneficial for symptoms caused by viral inflammation.
j) The result of the pooled odds ratio in the Quant analysis concurs completely with the Hato study
k) Only one outlier exists. That is the Sullivan study. That study included patients with a low-grade paresis; it included 7% patients with an initial House Brackmann grade 1 – who were already fully recovered at the inclusion-. In this study a dosage of antivirals was used that was too low to combat a possible Zoster infection. It is the only study in which combination therapy seems to worsen the situation. The authors suggest a Jarisch-Herxheimer reaction to account for this strange outcome. An outcome that I have never seen with this type of combination therapy and that has not been mentioned in any other study on this subject. It might also indicate that the authors themselves were surprised by this outcome.
l) All further studies show a positive outcome!
m) Leaving out the most negative and the most positive result gives a very nice result in the heterogeneity index. (As does simply leaving out Sullivan.) This clearly indicates that indeed there is only one outlier, the rest of the results are very homogeneous.
n) The studies by Ryu (2012) and by Lee (2013) clearly show a trend that in severe cases combination therapy is favourable.
o) Failure to recover beyond grade V as shown by Lee and not beyond grade IV as shown by Minnerop, only occurs in those patients that receive monotherapy corticosteroids, not in the group receiving combination therapy.
p) Synkinesis as shown by Axelsson (2013) occurs far more frequently in the group receiving monotherapy as compared to those receiving combination therapy.
Therefore, I think that except for one large trial with a rather dubious outcome, all studies suggest a positive effect of combination therapy. Quant et al. cannot but conclude that a positive pooled odds ratio, given the a priori chance and all circumstantial evidence, is indeed indicative for a positive effect; QED.
(A small mistake - as seen in the miscalculation that made this correction necessary - is easily made. We are human beings after all and even Tayside Pharmaceuticals, the company that is responsible for the placebo and acyclovir blinded boxes, is not infallible.
Let us assume, for arguments sake, that in the Sullivan study the placebo and antiviral medication were accidentally switched. In that case - if I am not mistaken – their treatment strategy:
1) Prednisolone would exist of a group receiving Prednisolone/Placebo
and of a group receiving Prednisolone /Antiviral
2) No Prednisolone would exist of a group receiving Antiviral / Placebo
and of a group receiving Placebo / Placebo
3) Acyclovir would exist of a group receiving Prednisolone / Placebo
and of a group receiving Placebo / Placebo
4) No Acyclovir would exist of a group receiving Prednisolone / Antiviral
and of a group receiving Antiviral / Antiviral
Since, we know that corticosteroids have a larger effect than antivirals alone, and antivirals are probably better than placebo, the a priori suggestion is that the best chances of recovery would be found, successively, in groups 1,4,3 and 2.
And what was the outcome of the Sullivan study? Indeed: group 1: 94.4%, group 4: 90.8%, group 3: 85.4%, group 2: 81.6%.
This would also explain why patients suffering a severe case of Bells palsy fare better with placebo prednisone than with combination therapy in their study. So, no Jarish-Herxheimer explanation. Simplex veri sigillum! A perfect fit I would say. Not to mention what this assumption would mean for heterogeneity indexes etc. )
Competing interests: No competing interests
Bell's palsy, idiopathic paralysis of the facial nerve, may be caused
by infection with the herpes simplex virus, and in some cases by zoster
sine herpete. The inflammatory reaction causes swelling of the nerve,
which then becomes trapped in the bony canal. The main priority here is to
reduce the swelling. Corticosteroids seem to be the ideal treatment. Can
antiviral drugs add anything to the recovery? Given the fact that on
average more than 70% make a full recovery and 85% make a good recovery
without therapy, and in view of the fact that these percentages are
significantly higher when prednisone is prescribed, it will be difficult
to prove any added value, as there will be a ceiling effect.
However, there are also two other points in the analysis of Quant et
al. that we would like to focus on. The most useful to consider is the
forest plot in figure 2 of the meta-analysis. In our opinion, the figures
from Engstrom's trial have been reversed in this analysis. The percentage
of good recoveries in the corticosteroids group should be 0.86 (160/186)
and in the combined therapy group it should be 0.91 (164/180).
If we were to create a forest plot using the correct figures, the trial
with the highest quality according to Jadad would become positive after
all. The OR of 0.93 mentioned by Quant then becomes 1.66. The OR of the
entire meta-analysis then becomes 1.72 [1.02-2.88].
Although this effect is marginally significant, the magnitude of the
estimated effect (OR = 1.72) suggests that there might be a strong
positive effect. The pooled proportion of patients with recovery should be
87.2 (506/580) among those who received steroids alone, compared with 92.2
(521/565) in those who received steroids and antivirals. This is in
contrast with the 88.2% and 91.2% as mentioned. A few percent more
adequate recoveries can substantially prevent morbidity in some patients.
Therefore, this 5% could be an effect that is clinically very relevant.
The above-mentioned point also causes the index for heterogeneity of the
studies to fall. The I2 as mentioned by Quant drops from 47.1 to 32.4%.
This means, in our opinion, that a fixed-effects model can also be used
instead of a random-effects model. This reduces the reliability interval,
which also points to there indeed being an effect, OR 1.69 [1.12-2.53].
Moreover, this meta-analysis indicates that there could be
publication bias, and therefore include trials in their funnel plot by
means of a trim and fill algorithm.
Although this method can be valuable, it relies heavily on the assumptions
that there are extreme negative studies, but that they are missing, and
that the distribution of the ORs can be estimated by the selected studies.
In fact, it is not a likely assumption that there 'must' be extreme
negative studies in case of antivirals used to treat a viral cause.
In view of our results, the probable pathogenesis, and in particular
the severe morbidity of permanent sequelae, antivirals continue to be part
of the arsenal against Bell's palsy. They should be prescribed in combined
therapy in the event of severe deficit and for elderly patients, as long
as there are no major contra-indications, particularly in view of the low
risk of usually temporary, moderate side-effects with this medication.
Competing interests: No competing interests
Large numbers of patients suffering from a Bell’s palsy recover
spontaneously. What’s more, almost everyone is convinced that
corticosteroids can promote this recovery. Only a very small group of
patients, especially elderly patients and those suffering from a grave
paralysis, still face a chance of poor recovery. It is this group that
urgently needs additional treatment. 1 Various authors have suggested that
antiviral medication could be of importance in these cases.
Recently, an interesting meta-analysis discussing the value of
antiviral medication was published.2 The authors clearly describe their
method of searching and evaluating literature, and the clinical problems
in this disease. I agree with the clinical implication of their
conclusion: do not routinely prescribe antiviral drugs, except for
patients with a severe palsy perhaps.
This conclusion, however, despite all the hard work, does not add
much that is new.
A point of criticism that is often overlooked in EBM, is that even though
proper search strategies are followed and the methodological quality is
explicitly looked into, there appears to be a definite lack of attention
for the clinical relevance, background knowledge and practical experience.
Too often it seems as if methodology takes precedence over knowledge. In
other words, pooling irrelevant data in a systematic review, only results
in ‘pooled irrelevant’ data.
The authors should, in our opinion, have more adequately anticipated
the clinical relevance and should have examined the value of antiviral
medication for patients suffering a severe palsy, namely House Brackmann
(IV), V, and VI. In most studies included in the review, the - clinically
speaking- most important question was not answered and the result of the
meta-analysis can therefore not possible provide any answers either.
It is therefore of great importance, that aside form a ranking such
as the Jadad–score that was used by Quant et al, a clinical ranking of
articles is formulated. Preferably, this should take place prior to the
methodological evaluation, so that less attention needs to be paid to
clinically non-relevant studies. In short, the clinical validity and
relevance should be a primary focus; the methodological validity should be
of secondary importance.
For example, with regard to the topic of Bell’s Palsy, one should
also take a critical look at the person who was responsible for the
patient evaluation, as well as the grading method that was used.
Hato et al. did find relatively large differences between the group of
patients treated with a combination therapy and the group that was treated
with prednisone alone.3 A popular criticism of their study is that “Hato
was not blinded”. 1 However, notwithstanding the fact that the value of a
double blind study indeed is important in order to prevent bias, we
personally think that a not blinded Hato still offers a more reliable
clinical picture of the facial nerve function, than the evaluation by the
average doctor on a trial team. What’s more, evaluation based on imaging
is absolutely unreliable, as the most important function of the facial
nerve is dynamic.
Furthermore, the often used House Brackmann Grading Scale has been
developed based on patients with cerebellopontine angle pathology mainly;
offering a gradation into rough categories, and as such it is not suitable
for the registration of subtle differences in function.
In the Minnerop-study, which received the lowest Jadad-score, one and
the same person re-examined the facial nerve function at follow-up.
Furthermore, this study distinguished between severe and medium gravity of
paralysis. In our view, clinically speaking this is one of the most
Apart from that we should also take a closer look at those patients that
did not recover. In Minnerop’s study, for example, we find a group of
patients that does not attain recovery beyond the HB III level. These
patients, however, are all part of the group that did not receive
The striking difference in the study of Minnerop also merits
attention. The difference in the results of combination therapy or
prednisone alone, is such that you find yourself quite willing to
prescribe medication to attain the utmost in recovery for those 25% of
patients in the severely affected group as well.
On the other hand, when one takes a good look at the results of the
Sullivan study, which is considered to be one of the most important in the
systematic review by Quant et al, in first instance, one is struck by the
fact that all gradations of severity were included.
A later subgroup analysis shows yet another problem.5 Although the placebo
group with moderate paralysis does as well as the group treated with
antiviral medication, it appears that in severe paralysis antiviral
medication actually hinders proper recovery. Dr. Daly has personally
assured me that the group classification was correct.
What remains is a possibility that the pharmacist accidentally switched
the ‘blinded’ medication, or a possibly negative interaction between
prednisone and the antiviral medication occurred, or we may have found
ourselves confronted with one of those ‘once in million’ results.
This result, however, deviates so far from our current medical knowledge,
that the use of the research results of this particular study in a meta-
analysis is cause for serious doubt.
In any case, the result is not consistent with our own practical
experience – which is comparable with the Hato-studies – where more than
90% of the severely affected patients, when treated timely with
combination therapy, show satisfactory recovery.
Quant et al correctly stated that the real question is: “which
endpoint to use for a disease in which a very high proportion of patients
recovers with standard therapy (steroids alone).” They also rightly
suggested that only famcicloir and valaciclovir should be used in any
subsequent studies. Furthermore, we think that new studies should focus on
the more severely affected category of patients, as well as on the use of
a more reliable grading system with regard to the functioning of N VII.
Our department would not easily be tempted to take up such a study. The
current policy – namely, prednisone in case of moderate severity and
combination therapy in case of grave paralysis – results in such high
percentages of patients that enjoy a complete recovery, that we are of the
opinion that not prescribing antiviral medication, even within the context
of a study, would amount to a less than optimal treatment.
J. Alexander de Ru and Erwin L. van der Veen
Department of Otorhinolaryngology, Centraal Militair Hospitaal and
University Medical Center, Utrecht, The Netherlands
1) Gilden DH. Bell’s Palsy- Is glucocorticoid treatment enough? N Engl J
2) Quant EC, Jeste SS, Muni RH, Cape AV, Bhussar MK, Peleg AY. The
benefits of steroids versus steroids plus antivirals for treatment of
Bell’s palsy: a meta-analysis. BMJ 2009;339:b3354
3) Hato N, Yamada H, Kohno H, Matsumoto S, Honda N Gyo K, et al.
Valacyclovir and prednisolone treatment for Bell’s palsy: a multicenter,
randomized, placebo-controlled study. Otol Neurotol 2007;28:408-13
4) Minnerop M, Herbst M, Fimmers R, Matz B, Klockgether T, Wüllner U.
Bell’s palsy; combined treatment of famciclovir and prednisone is superior
to prednisone alone. J Neurol 2008, online September 2008
5) Sullivan F, Swan I, Daly F. The authors reply. N Engl J Med
Competing interests: No competing interests