Scandinavian Journal of Gastroenterology, 2010; 45: 637–639

LETTER TO THE EDITOR

Comment and reply on: Effect of fat emulsion (Fabuless) on orocecaltransit time in healthy men

The recent article by Haenni et al. [1] concludes thata unique fat emulsion (Fabuless�) affects orocecaltransit time (OCTT) in healthy men. OCTT wasdetermined by measuring plasma levels of sulfa-pyridine (SP), a colonic metabolite of salazopyrine.However, their computation of OCTT is unclearand possibly flawed.The recommended methodology for estimating

OCTT is to: (1) determine time of SP appearancein plasma for each individual separately, and then (2)calculate the mean times for each treatment [2–7].Thus, OCTT should be measured per person, andtime becomes the dependent variable. InsteadHaenni et al. [1] appear to use group mean plasmaSP values as a dependent measure, and from thisestimate times, a procedure that could greatly exag-gerate the contribution of one or more subjects to themean value [4].Haenni et al. [1] refer for their methodology to

several papers [2,5,7], which all base OCTT onindividual appearance values as recommended above.These and other papers [3,8,9] all determine OCTTfrom individual values, based on the first detectableappearance of a relevant marker in plasma (or inter-polation from SP levels back to estimated individualarrival times of salazopyrine in the colon [5]).An apparent effect of using mean SP values seems

to be reflected in high reported standard deviations at4 and 5 hours by Haenni et al. [1], for their “active”versus control treatment. Gramatte and Terhaag [4]illustrate how variability of individual SP values canbe very high relative to the appropriately calculatedOCTT.There are other issues which also may affect the

conclusions drawn by Haenni et al. [1]. The 60 minintervals for plasma sampling contrasts with intervals of10-30 min by others [2–4,7]. Their extrapolation ofmean SP values to time is hard to interpret and can alsogive inaccurate estimations (versus curve fitting). Theauthors also split their total measurement period intophases of 1–4 and 5–11 hours: the justification for this,and whether it was a predetermined or post-hoc divi-sion, are not given. Lastly, the inclusion of a meal in themiddle of the critical measurement period may furthercomplicate interpretation.

The authors also refer to ileal brake effects onOCTT,and the link to satiety. Although it has been shown thatdeliveryof fat to the ileumincreasesOCTT,astrong linkbetween OCTT and satiety has not been established.Linking OCTT to satiety when satiety was not mea-sured (or at least not shown) is not clearly warranted,and indeed they also cite evidence that appetiteeffects claimed for Fabuless� have not been replicable(e.g. [10]).Overall, there is at least a need for clarification of

the methods used in the paper by Haenni et al. [1]. Ifindeed group mean values rather than individualtimes have been used as the basis for calculatingOCTT, then an analysis using the correct procedureshould be carried out and presented by the authors.

Harry Peters1, Christoph Beglinger2,Dave Mela1 & Ewoud Schuring1

1Unilever R&D Vlaardingen,Vlaardingen, the Netherlands

2Department of Gastroenterology and Hepatology,Division of Gastroenterology,

University Hospital,Petersgraben, Basel, Switzerland

Address for Correspondence:Harry PF Peters, PhD,

Department of Nutrition and health,Unilever R&D Vlaardingen,PO Box 114, Vlaardingen,3130AC, the Netherlands.

E-mail: harry.peters@unilever.com

References

[1] Haenni A, Sundberg B, Yazdanpandah N, Viberg A,Olsson J. Effect of fat emulsion (Fabuless) on orocecaltransit time in healthy men. Scand J Gastroenterol 2009;44:1186–9.

[2] Kellow JE, Borody TJ, Phillips SF, Haddad AC,Brown ML. Sulfapyridine appearance in plasma after salicy-lazosulfapyridine–another simple measure of intestinal tran-sit. Gastroenterol 1986;91:396–400.

[3] Staniforth DH. Comparison of orocecal transit timesassessed by the lactulose breath hydrogen and the

(Letter received 2 November 2009; accepted 7 December 2009)

ISSN 0036-5521 print/ISSN 1502-7708 online � 2010 Informa UK Ltd.DOI: 10.3109/00365520903536523

Scan

d J

Gas

troe

nter

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y M

ichi

gan

Uni

vers

ity o

n 10

/29/

14Fo

r pe

rson

al u

se o

nly.

sulfasalazine-sulfapyridine methods. Gut 1989;30:978–82.

[4] Gramatte T, Terhaag B. The variability of orocecal transit-time evaluated by the salicylazosulfapyridine sulfapyridinemethod. Int J Clin Pharmacol Ther 1991;29:147–50.

[5] Sunesen VH, Vedelsdal R, Kristensen HG, Christrup L,Mullertz A. Effect of liquid volume and food intake on theabsolute bioavailability of danazol, a poorly soluble drug.Eur J Pharm Sci 2005;24:297–303.

[6] Peh KK, Yuen KH. Indirect gastrointestinal transitmonitoring and absorption of theophylline. Int J Pharm1996;139:95–103.

[7] Sogni P, Chaussade S, Akue-Gohe K, Nepveux P,Homerin M, Couturier D, Guerre J. Comparative effects

of ricinoleic acid and senna on orocecal and oroanal transit-time in healthy-subjects – application of the salazopyrinmethod. Gastroenterol Clin Biol 1992;16:21–4.

[8] Lin HC, Prather C, Fisher RS, Meyer JH, Summers RW,Pimental M, et al. Measurement of gastrointestinal transit.Dig Dis Sci 2005;50:989–1004.

[9] Papasouliotis K, Gruffyddjones TJ, Sparkes AH, Cripps PJ.A Comparison of orocecal transit times assessed by thebreath hydrogen test and the sulphasalazine/sulphapyridinemethod in healthy beagle dogs. Res Vet Sci 1995;58:263–7.

[10] Logan CM, McCaffrey TA, Wallace JMW, Robson PJ,Welch RW, Dunne A, Livingstone MB. Investigation of themedium-termeffects ofOlibra� fat emulsionon food intake innon-obese subjects. Eur J Clin Nutr 2006;60:1081–91.

AUTHOR’S REPLY

We thank Peters et al. for their interest in, and for theircritical assessment of our work [1]. We will try toclarify the computation of oro-cecal transit time(OCTT), which they consider as unclear, and showthe method we are using is statistically sound, anddoes not results in flawed conclusions.There are different methods available to measure

OCTT, and we have chosen this noninvasivesalazopyrine (SP) assay, a method with minor influ-ence on the transit time as such. Although thestatistical methodology proposed by Peters et al.has been used in other studies, we are not aware ofa commonly recommended way to apply thismethod and the suggested method has some signifi-cant shortcomings.Determination of SP appearance in plasma for each

individual separately involves blood sampling at ahigh frequency, preferably at least once per 10 minif the proposed method should be applied. Suchfrequent blood sampling means that a considerableamount of blood is required per subject for an 11-hperiod. The interval between the cross-over was oneweek, hence within two weeks almost one litre bloodshould be collected from each subject. Withdrawal ofsuch volumes would neither be ethical nor approvedby ethics committee, and more importantly couldinfluence the subjects and accordingly the studyresults. The second point, definition of a detectablelevel, is not easy. In one subject, we measured a smallamount of SP in plasma, already after 1 h (in bothtests). Does this mean that food in this person passesstomach and small intestine extremely fast, or does itmean that, for whatever reason, some SP is fastabsorbed from stomach or small intestine, or that alittle amount of digesta containing SP passes stomachand small intestine extremely rapidly? In this specific

subject, SP was shown in plasma after 1 h, but thelevel remained very low until about 4–5 h. Thus,rather than using an ideal method, two problemsare being introduced.After determination of the time for the first detect-

able appearance Peters et al. propose to calculatemeans of these times as mean times per treatment.Typically, such appearance times are intrinsically notnormally distributed, and show a very large variabilitybetween subjects, hence, the method proposed byPeters et al. has its own limitations.The method utilized in this study has following

advantages. For each subject, at each time point asample was taken and analyzed for its SP content. Pertime point, the data were checked for the standardstatistical characteristics, variance, and possible out-liers. We showed that at each time point, none of thedata had to be considered as an outlier according toDixon’s and Grubs tests [2], and that the data werenormally distributed. There were no observationswhich might have influenced the outcome strongly,as indicated by Peters et al.; such information was alsoprovided in the published paper. We also confirmedequal variance (homoscedasticity) of the data over alltime points. Because the procedure includes a repe-tition of measurements within subjects, which wouldprobably be correlated, as might have been a concern,we used the MIXED procedure of SAS for ourcalculations, which allows for such correlation inthe model. We acknowledge that this has not beendescribed in detail in the paper, and this might haveclarified our procedures. On the other hand, this hasto be considered as basal checks and proceduresconducted during statistical calculations in any exper-iment as part of repeated measures. Thus individualdata points per time-point were used in our

638 Letter to the Editor

Scan

d J

Gas

troe

nter

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y M

ichi

gan

Uni

vers

ity o

n 10

/29/

14Fo

r pe

rson

al u

se o

nly.

calculations, contrary to the suggestion of Peters et al.that we have used group means.Our measurements of plasma SP values could very

well be fitted by a quadratic model. This curveremained very low during the first 4 h after applicationof SP, as might be expected by a normal passage ratethrough stomach and small intestine of about 4 h, aninterval which appears also in our data, given that> 75% of the SP values at 4 h is still “ zero”. Then, theSP level increased strongly, indicating that bowelcontents entered the colon. Interestingly, the calcu-lated curves were shown to be parallel over the timeframe 4–11 h, indicating that a time lag appeared inpassage of SP to absorption (from the colon) of thetest product vs. control, but no subsequent change ofpassage characteristics. Our calculations are, thus,based on curves fitted to the data, which we agreewith Peters et al. to be the best method.That the followed procedure results in a relatively

high standard deviation of SP levels after 4–5 h, is notthe result of using SP means as suggested by Peterset al. but large variations between humans in the firstappearance of SP. Such variation has also beenobserved by Gramatté and Terhaag [3]. This variationis not accounted for in the results as presented inTable II of our paper [1]. However, in the statisticalmixed model used, the factor “subject” is accountedfor, meaning that between – subject variation is notpart of the error term.If we would analyze our data by transforming the

SP levels to time points, individual or as a groupmean prior to statistical evaluation, we would intro-duce some significant statistical problems. SP data,as analyzed by us, are distributed normally, haveequal variance, and are symmetric. Transformationof these data to time would result in an inflated andunequal variance at the different time points (hetero-scedasticity) and skewness, as can easily be under-stood from the quadratic response observed.Proposed transformation would, thus, result in aworse distribution, and its application would bestatistically unsound.At the end of the statistical evaluation of our data,

we transformed the SP values to time. This proce-dure, which is widely applied in statistics, does notaffect significance or the confidence interval. It is, inthis case, not intended to improve the data forstatistical calculations, but to make it easier for the

reader to understand and interpret what the differencein SP-values means time-wise.As indicated above, the procedure followed, pro-

cessing of individual SP data and statistical evaluationof these data prior to transformation of the results tothe dimension “time”, is accurate, and is statistically asound procedure.When planning the meal schedule, avoidance of

any meal during daytime might not be physio-logical which was the reason why this lunch mealwas included. The scheduled meals were identicalduring both test situations and for all participants.Thus, we think that the meal schedule per se mightnot have any influence on the main results.Our investigation clearly showed that the passage rate

of digesta through the stomach and small intestine isreduced after FabulessTM consumption. It was not theaim of the current investigation to test the relationshipwith satiety, such investigations have been done inprevious studies and the design of this study wasoptimized for evaluation of transit time. However, itseems apparent that the reduction of passage rate, assuggested in the discussion in this paper, very likely isdue to activation of the feed-back system called ilealbrake. Such a feed back system has been shown toaffect satiety.We hope that the explanation and additional

description have clarified that the applied methodol-ogy in this study is appropriate and statisticallycorrect to use in this type of experiment.

Johan OlssonGood Food Practice,

Dag Hammarskjölds väg 10B,Uppsala, 75183, Sweden

E-mail: Johan.Olsson@good-food-practice.com

References

[1] Haenni A, Sundberg B, Yazdanpanah N, Viberg A, Olsson J.Effect of fat emulsion (Fabuless) on orocecal transittime in healthy men. Scand J Gasteroenterol 2009;44:1186–90.

[2] Snedecor GW, Cochran WG. Statistical methods. 7th ed.Ames, IA, USA: The Iowa State University Press; 1980.

[3] Gramatté T, Terhaag B. Int J Clin Pharmacol Ther Toxicol1991;29:147–50.

Letter to the Editor 639

Scan

d J

Gas

troe

nter

ol D

ownl

oade

d fr

om in

form

ahea

lthca

re.c

om b

y M

ichi

gan

Uni

vers

ity o

n 10

/29/

14Fo

r pe

rson

al u

se o

nly.