# Effect of fibre, antispasmodics, and peppermint oil in the treatment of irritable bowel syndrome: systematic review and meta-analysis

BMJ 2008; 337 doi: https://doi.org/10.1136/bmj.a2313 (Published 14 November 2008) Cite this as: BMJ 2008;337:a2313## All rapid responses

*The BMJ*reserves the right to remove responses which are being wilfully misrepresented as published articles.

This is a late apology to Professor Altman about how I was so

misleading in my trek through log scales and meta-analyses. I have been

confused and confusing.

Customarily, a forest plot will have a relative risk (or odds ratio)

on a logarithmic x- axis. A log axis is resorted to because the standard

error of relative risk is logarithmic (1). A linear framework would

possess skewed, irregular confidence intervals, which are based on the

standard errors.

One is unhappily reminded of Alice's uncomfortable variations in

dimensions through that vertiginous looking glass.

With numbers needed to treat, another difficulty pops up- that the

null hypothesis of no effect is at infinity. Various weird things then

happen to the confidence intervals! (2) A forest plot becomes most

problematic. Curiouser and curiouser...

What one might fear is that forest plots are not always 'particularly

elegant and intuitively comprehensible' (3). They suffer from mathematical

complexities, revolving round the confidence intervals. There are relative

axes that are linear, that are to log 2 or log 10, that could thus be

manipulated to make them appear what they are not in the cliche of

evidence biased medicine.

Professor Altman claims: 'the NNT should never be used as the basis

for a meta- analysis' (4).

But the actual article on Irritable Bowel Syndrome (IBS) is a meta-

analysis that does resort to numbers needed to treat (5). NNT is quoted in

the conclusion ('What This Study Adds'). NNT is a mathematical draw- back

only when the confidence interval is not statistically significant. NNT

is, as Professor Altman nervously implies, only partially applicable in a

meta- analysis.

Anti- spasmodics are said to be first- line treatment by NICE and the

British Society of Gastroenterology, but have an inferior NNT to the

apparently natural remedy of peppermint oil. Peppermint oil is not

recommended at all in the official prescribing guidelines. Without this

article, and its homely reliance on NNT, we would be less aware of the

peppermint oil factor in IBS.

Graphs more than any other part of statistical presentation can be

prone to deformation and distortion. Against all the brave efforts of

people such as Professor Altman, not enough down to earth explanations of

statistical methods exist. Medicine is even more bewildered than Alice as

it confronts the Chesire cats and Jabberwockys of mathematics.

Mathematical problems can be medical problems.

Again I am sorry to Professor Altman and the BMJ for having been such

an idiot over these dreadfully complex issues. Any layman must feel

bewildered by the not always well fitting nexus between medicine and

mathematics.

REFERENCES:

(1) Statistics with Confidence. M.J.Gardner and D.G.Altman. BMJ.

1990. pg.52.

(2) Confidence Intervals and the number needed to treat. D.G.Altman.

BMJ 1998;317:1309-12/

(3) How to read and understand and use systematic reviews and meta-

analyses. S.Leucht et al. Acta Psychiatr Scand 2009:119:pg. 447.

(4) Re: Forest Plot: not seeing the wood for the logs. BMJ Rapid

response. D.G.Altman. 18 May 2009.

(5) Effect of fibre, antispasmodics and peppermint oil in the

treatment of irritable bowel syndrome: systematic review and meta-

analysis. Alexander c. Ford et al. BMJ 2008;337:a2313.

Competing interests:

None declared

**Competing interests: **
No competing interests

I thank Professor Altman for his response, since science cannot

progress without scepticism- or debate.

The lie factor is not Einstein with graphs! Take, for example, a meta

- analysis where the relative risks are all between 1 and 2. Then a spaced

out x- axis to log base two will make the figures seem move removed from

the y- axis than an x- axis to log base ten. Of course, the real data are

the same, but the appearances? Ha! It is just a mere visual thing, it is

not high powered mathematics (1). Some relative risks in forest plots are

on a linear axis, and then we would have all sorts of distortions to the

confidence intervals. The mathematical problem is that the standard error,

and therefore the confidence interval, with relative risk (as a ratio) is

logarithmic (base e).

Statistical theory would be impossible without that natural log, e.

As for numbers needed to treat, this seems a practical down to earth

sort of way for clinicians to examine data. That a meta- analysis,

according to Professor Altman, is unable to include numbers needed to

treat is a sign something is deficient about the excessively theoretical

nature of a meta- analysis.

What of absolute risk difference? Again, the way this looks visually

would alter according to what sort of x- axis lay out we had. The damned

lies label on statistics derives from spinning identical data, often

through graphs, to make it look different- quite unjustifiably.

REFERENCES:

The Cambridge Dictionary of Statistics in the Medical sciences. B. S.

Everitt. 1995. Pg. 142.

'Lie factor=apparent size of effect show in graph divided by actual

size of effect in data'

Competing interests:

None declared

**Competing interests: **
No competing interests

It would be unfortunate if Zekria Ibrahimi’s comments (1)

mislead anyone into thinking that forest plots are unreliable. It is of

course true that logarithms to base 10 or base 2 give different values.

But when a confidence interval is calculated for a quantity on a log

scale (such as log relative risk) the values obtained are then back-

transformed to the original scale (here, the relative risk). Thus the

answer will be the same whether using logs to base 2, 10, e or anything

else. So the forest plot for a meta-analysis will look identical

regardless of the base used as long as the scale shows the relative risk,

not the log relative risk, as is good practice. It is not correct to say

that "different log bases would manipulate the visual lay-out, perhaps to

misrepresent data."

In the paper by Ford et al (2) the relative risk was indeed plotted on a

log scale rather than the log relative risk being plotted. Unfortunately

the original axis labelling incorrectly showed some negative values, but a

correction was posted on 8 May.

The comment about number needed to treat (NNT) is also misleading as the

NNT should never be used as the basis for a meta-analysis.

Douglas G Altman

1 Ibrahimi Z. Forest plot: not seeing the wood for the logs.

http://www.bmj.com/cgi/eletters/337/nov13_2/a2313#213704

2 Ford AC, Talley NJ, Spiegel BMR, Foxx-Orenstein AE, Schiller L, Quigley

EMM, Moayyedi P. Effect of fibre, antispasmodics, and peppermint oil in

irritable bowel syndrome: Systematic review and meta-analysis. BMJ

2008;337:a2313.

Competing interests:

None declared

**Competing interests: **
No competing interests

**18 May 2009**

According to Wikipedia (1), a forest plot may have a natural

logarithmic scale on the x- axis so that confidence intervals can appear

symmetrical about the means, and ratios (relative risk being a ratio) seem

equivalent both above one and below one. A log scale will give the same

length in a confidence interval for a ratio of ten, whether this is ten to

one or one to one tenth.

'One' at the bottom of the y- axis is the null hypothesis of no

effect. The forest plot provides a display of the null hypothesis as the y

- axis. If the confidence interval intrudes on this y- axis, then there is

no statistical significance.

But ... the forest plot on a logarithmic scale still may be capable

of the lie factor (2)- that is, graphical misrepresentation.

How?

Natural logaritms are to base e, that is, 2.718 ... Confidence

intervals are worked out to base e in the BMJ book, Statistics with

Confidence (3).

However, most forest plots are to base 10, not base e. Traipsing

unsteadily through the Internet, I also happened upon a forest plot of

base 2 on the x- axis(4). There are linear forest plots too.

Different log bases would manipulate the visual lay- out, perhaps to

misrepresent data.

I am not a mathematician, and my own previous rapid responses to this

article, and indeed the transcription error in the article, indicate the

confusion possible over forest plots and their x- axis.

The difficulty with the x- axis in forest plots emerges most when the

effect size is numbers needed to treat- then the null hypothesis is set at

infinity (5)!

The term, forest plot, only emerged during the 1990's. The unhappy

suggestion has to be that it remains an immature statistical device. It is

not yet thoroughly confident and sound in root and branch.

As far as forest plots are concerned, one is not seeing the wood for

the logs!

REFERENCES:

(1) Forest Plot. Wikipedia.

(2) The Cambridge Dictionary of Statistics in the Medical Sciences.

B. S. Everitt. CUP. 1995. Pg. 142.

(3) Statistics with Confidence- Confidence intervals and statistical

guidelines. Martin J Gardner and Douglas G Altman. BMJ. 1990. Pgs. 51-52.

(4) Statins and cancer Risk: A Literature- Based Meta- Analysis and

Meta- Regression Analysis of 35 Randomized Controlled trials. Journal of

Clinical Oncology. Vol. 24. No. 30. Pgs. 4808-4817. (October 20). 2006. S.

Bonovas et al.

(5) Confidence intervals for the number needed to treat. Douglas G

Altman. BMJ 1998;317:1309-12.

Competing interests:

None declared

**Competing interests: **
No competing interests

There is a general problem with these sort of forest plots for

relative risk.

Skimming uneasily through medical literature, one finds that relative

risk is often formatted as in this article on irritable bowel syndrome (1)

, viz with a logarithmic x- axis.

Such representation verges on medically misleading, because of the

contrary nature of logarithms.

An increase of relative risk of, for example, one half, would occupy

a small section on the side above one, but a much larger section on the

side below one. Logarithms are not neat and linear. By definition, and

most inconveniently here, they expand exponentially.

So we have a medically skewed picture through such a logarithmic

framework for relative risk. Fractional figures below one are stretched

out; figures above one are (relatively) squeezed in. Presumably,

logarithms are deployed because the data- always guaranteed to be

irregular and nasty- would otherwise be scattered too far off the page.

Alas, one might misuse such a logarithmic basis to almost misrepresent

data on a graph. Relative risk above one would be rendered less important

in terms of the graph, of the visual set- up.

The transcription error partly happened because the logarithmic

transformation of the data was not being made explicit.

Yet again, the mathematics and the medicine are capable of being at

cross purposes. These two monsters just do not want to trot prettily in

tandem.

REFERENCES:

(1) Effect of fibre, antispasmodics and peppermint oil in the

treatment of irritable bowel syndrome: systematic review and meta-

analysis. Alexander C Ford et al. BMJ 2008; 337: a2313

Competing interests:

None declared

**Competing interests: **
No competing interests

It is not pleasant to go where statisticians fear to tread.

The article on how difficult it is to treat irritable bowel syndrome

(1) ran into a quagmire of graphs (Figs 2, 3, 4). According to one rapid

response (2), the x- axis wrongly implied there could be a negative

relative risk.

Relative risk is a ratio that can range from zero beyond 1 and then

possibly to infinity. Infinity here would be merely theoretical, as so

much unfortunately is in statistics. Relative risk compares control to

treatment, non- exposed to exposed (3).

So why were there negative values of relative risk on the graphs?

What hellish error had been made?

Ah, the spacings are not linear, but logarithmic. We have a

logarithmic x- axis. And logarithms can be negative for fractions below

one. The logarithm of one in base ten is- zero. So Figs 2, 3, and 4 would

seem valid after all if they were in the form of logarithmic graphs.

Still, they have proved confusing. They were not defined as logarithmic in

the forest plots.

And why choose logarithms? These would make the relative risk appear

less than it actually was. Perhaps there was not enough space for an

ordinary linear graph.

Maybe the intention was to give the data an unfair graphical spin.

And here is the twist in this statistical tale. By picking a log or a

linear axis, we can make the data look as we would wish it to on a graph,

for all sorts of malevolent purposes. We can squeeze it in or stretch it

out.

There are lies, damned lies, statistics- and graphs.

REFERENCES:

(1) Effects of fibre, antispasmodics, and peppermint oil in the

treatment of irritable bowel syndrome: systematic review and meta-

analysis. Alexander C Ford et al. BMJ 2008;337:a2313

(2) Error in figs 2,3, 4. BMJ Rapid Response. Mitchell Levine and

benjamin Asa. 6 May. 2009.

(3) Statistical Methods in Medical Research. P. Armitage, G.Berry,

J.N.S.Matthews. Blackwell. Fourth edition. 2002. Psgs. 671- 673.

Competing interests:

None declared

**Competing interests: **
No competing interests

We were reading this article only recently and noticed that there is

a printing error in the labels for the x-axis in figures 2,3 and 4. The

centre point should not be "zero", but should be labelled as "one". In

addition, the numbers to the left of the centre point should not have

negative values.

When presenting relative risks in a figure it is not possible to plot

zero anywhere on the axis, nor can a relative risk have a negative value.

Since it is an on-line version that readers are accessing, I assume

that changes to the document are still possible and you might want to

consider correcting the problem.

Competing interests:

None declared

**Competing interests: **
No competing interests

Editor

We are grateful for the interest that our recent article concerning

the efficacy of fibre, antispasmodic drugs, and peppermint oil in

irritable bowel syndrome (IBS) has generated, and the important and

exciting debate it has stimulated

We agree with Dr. Dib that IBS is a chronic, relapsing and remitting

condition. We chose a minimum duration of therapy in randomised controlled

trials (RCTs) of the agents we studied in our systematic review and meta-

analysis of 1 week for the reasons we discussed in our article. [1]

However, 34 of the 35 eligible studies randomised patients to therapy in

excess of 1 week. The remaining trial [2] did not have any significant

impact on the overall conclusions when it was excluded from the analysis

(number needed to treat (NNT) with antispasmodics = 6; 95% confidence

interval 4 to 10). The minimum duration of therapy used in the trials of

fibre in IBS that we identified was 4 weeks, and the majority of RCTs used

12 weeks or more.

With regard to Professor Whorwell’s comments, we stated clearly in

the abstract of our article that the beneficial effect of fibre was

limited to ispaghula husk, [1] which is also known as psyllium. We cannot

be held responsible for the misrepresentation of our findings by the

“popular press”.

Dr. Leeds et al. make the important point that antidepressant drugs,

particularly tricyclic antidepressants, were also of benefit in IBS, with

an NNT in our systematic review and meta-analysis of 4. [3] However, there

was also a beneficial effect of a newer class of antidepressants,

serotonin re-uptake inhibitors, in IBS in this meta-analysis. The NNT was

3.5, albeit in a smaller number of RCTs containing fewer patients.

With respect to any role for non-pharmacological therapies in the

management of IBS, we have also examined this issue. We are not aware of

any published RCTs of homeopathy in IBS. There is an existing systematic

review of the efficacy of various psychological therapies in IBS,

conducted by Lackner et al., [4] which identified 10 trials that provided

extractable dichotomous data for pooling. As Drs. Plotnikoff and Weisberg

correctly state the NNT with psychological therapies reported in this

systematic review was 2. While this may appear very appealing,

unfortunately there were only a total of 185 patients included in these

RCTs, and 9 of the 10 trials in this systematic review emanated from a

single centre in the United States, which suggests that the treatment

effect may have been overestimated. In our own systematic review and meta-

analysis of the efficacy of psychological therapies in the management of

IBS, which identified 20 RCTs containing 1278 patients, we reported an NNT

to improve or cure 1 patient’s symptoms of 4. [3] We conducted a

sensitivity analysis that excluded the 9 RCTs conducted by the same group

of investigators, which demonstrated a reduced, though still statistically

significant, treatment effect of psychological therapies on global IBS

symptoms. There were only 2 of these 20 RCTs comparing hypnotherapy to

supportive therapy or waiting list control in 40 patients.

The medical management of IBS remains unsatisfactory, and no single

intervention has been shown to alter the natural history of the condition

convincingly, but data from these recent studies, which were performed to

inform the updated American College of Gastroenterology monograph on the

management of IBS, [5] suggest there are both pharmacological and

psychological therapies that are effective in IBS, at least in the short-

term.

Authors

Alexander C Ford1, Nicholas J Talley2, Eamonn MM Quigley3, Paul

Moayyedi1.

1Gastroenterology Division, McMaster University, Health Sciences

Centre, Hamilton, Ontario, L8N 3Z5, Canada.

2Professor of Medicine, Department of Medicine, Mayo Clinic Florida,

Jacksonville, Florida, FL 32224, USA.

3Department of Medicine, Clinical Sciences Building, Cork University

Hospital, Cork, Ireland.

References

1. Ford AC, Talley NJ, Spiegel BMR, Foxx-Orenstein AE, Schiller L,

Quigley EMM, Moayyedi P. Effect of fibre, antispasmodics, and peppermint

oil in irritable bowel syndrome: Systematic review and meta-analysis. Br

Med J 2008;337:a2313.

2. Virat J, Hueber D. Colopathy pain and dicetel. Prat Med 1987;43:32-34.

3. Ford AC, Talley NJ, Schoenfeld PS, Quigley EMM, Moayyedi P. Efficacy of

antidepressants and psychological therapies in irritable bowel syndrome:

Systematic review and meta-analysis [published online first: 10 November

2008]. Gut 2008;doi:10.1136/gut.2008.163162.

4. Lackner JM, Mesmer C, Morley S, Dowzer C, Hamilton S. Psychological

treatments for irritable bowel syndrome: A systematic review and meta-

analysis. J Consult Clin Psychol 2004;72:1100-13.

5. American College of Gastroenterology IBS Task Force. An evidence-based

position statement on the management of irritable bowel syndrome. Am J

Gastroenterol 2009;104 (suppl 1):S1-S35.

Competing interests:

Alexander C Ford: none declared. Nicholas J Talley: has received consultancy fees from Procter and Gamble, Lexicon Genetics, Inc., Astellas Pharma US, Inc., Pharma Frontiers, Ltd., Callisto Pharmaceuticals, AstraZeneca, Addex Pharma, Ferring Pharma, Salix, MGI Pharma, McNeil Consumer, Microbia, Dynogen, Conexus, Novartis, and Metabolic Pharmaceuticals, and has received research support from Novartis, Takeda, GlaxoSmithkline, Dynogen, and Tioga. Eamonn MM Quigley: has received consultant’s and speaker’s bureau fees from Nycomed, Boehringer Ingelheim, Procter and Gamble, Reckitt Benckiser and Prometheus, and holds equity in Alimentary Health. Paul Moayyedi: chair at McMaster University partly funded by an unrestricted donation by AstraZeneca, and has received consultant’s and speaker’s bureau fees from AstraZeneca, AxCan Pharma, Nycomed, and Johnson and Johnson.

**Competing interests: **
No competing interests

**15 January 2009**

We read with interest the meta-analysis by Ford and colleagues (1).

This important study provides a much needed supportive evidence base for

our current practice. Jones also makes a critical point that a ‘holistic

and integrated approach’ is necessary (2). With this perspective in mind

we would like to make the suggestion that either within the ‘what is

already known on this topic’ of Ford’s paper or in the discussion, that

there should be the mention of the potential benefits of amitriptyline. At

low dose (10mg – 25mg nocte) it appears that some patients particularly

with diarrhoea predominant IBS derive symptomatic benefit. A recent meta-

analysis by the same authors has suggested a positive effect with a number

needed to treat of 4 (3). Although further work may be required in this

area – amitriptyline still remains an ‘old’ but important addition to the

clinicians’ armamentarium and is in keeping with a holistic approach when

considering therapeutic options in IBS patients (4).

1. Ford AC, Talley NJ, Spiegel BMR, Foxx-Orenstein AE, Schiller L,

Quigley EMM, Moayyedi P. Effect of fibre, antispasmodics and peppermint

oil in the treatment of irritable bowel syndrome: systematic review of the

literature and meta-analysis. BMJ 2008;337;a2313.

2. Jones R. Treatment of irritable bowel syndrome in primary care.

BMJ 2008;337:a2213.

3. Ford AC, Talley NJ, Schoenfeld PS, Quigley EM, Moayyedi P.

Efficacy of Antidepressants and Psychological Therapies in Irritable Bowel

Syndrome: Systematic Review and Meta-analysis. Gut 2008. Nov 10 [Epub

ahead of print].

4. National Institute for Health and Clinical Excellence. Irritable

bowel syndrome in adults. Diagnosis and management of irritable bowel

syndrome in primary care. London: NICE, 2008. www.nice.org.uk/CG061.

Competing interests:

None declared

**Competing interests: **
No competing interests

**18 December 2008**

## Re: Effect of fibre, antispasmodics, and peppermint oil in the treatment of irritable bowel syndrome: systematic review and meta-analysis

This is a very belated reply to Douglas Altman's comments on Forest Plots. I was mistaken in implying that a log scale could be to log 2, or log 10, or log e. Log scales are identical, whatever the base log may be. The problem with forest plots displayed in any article is the 'visual lie' impact. For example, an x- axis of 0 to 100 will make a confidence interval of 1 to 10 look small, whereas an x- axis of 1 to 10 will make it look large. I believe that articles do misuse this 'visual lie' factor to unfairly massage the appearance of results.

Another problem with forest plots is that the confidence intervals are asymmetrical on a linear scale, although they look symmetrical on a log scale- in other words, the appearance of a forest plot can vary, depending on whether a log or linear scale is used. A log scale will compress results. Again, the appearances of a forest plot can be ambiguous, ambivalent, and so potentially unsafe.

With NNT, the confidence intervals in a forest plot may be considerably distorted, since the value of the null hypothesis of no effect is infinity. The forest plots become highly distorted in this situation.

So what I am questioning is not the mathematics, but rather the appearances, of the forest plot. Appearances have always an element of danger in them...

Competing interests:No competing interests11 November 2014