- Email: [email protected]
Limitations of the triolein breath test
Ciinica Chimica Acta. 205 (1992) 51-64 © 1992 Elsevier .Science Publishers B.V. All riots reserved 0009-8981/92/$05.00
51
CCA05202
Limitations of the triolein breath test A n d r e w D u n c a n a, A l i s o n C a m e r o n a, Michael .I. Stewart b a n d R o b i n I. RusseU a aGastroenterology Unit and blnstitute of Biochemistry. Royal Infirmary. Glasgow, G31 2ER (UK) (Received 23 July 1991; revision received 31 October 1991; accepted 8 November 1991)
Key words: Malabsorption; Fat absorption test; Carbon dioxide output; Breath test; Dual isotope fat absorption test
Summary Patients being investigated for intestinal absorptive capacity were classified as norreals or malabsorbers on the basis of three fat absorption tests. Malabsorbers were further classified as mild, moderate, severe or gross according to severity of malabsorption. Using this classification system the triolein breath test was evaluated in 53 patients. Seventeen patients were excluded because their graph of percentage breath [14C]carbon dioxide versus time was exponential indicating that the peak [~4C]carbon dioxide may be occurring later than the six hour duration of the test. The sensitivity and specificity of the triolein breath test were found to be 100% and 96%, respectively and moderate correlations with the individual fat absorption tests were found. However, the breath test was limited in its capacity to predict the severity of malabsorption. Carbon dioxide output was also measured in order to determine the applicability of using an assumed value. The respiratory quotient and variability of results were high in nineteen patients indicating possible hyperventilation. In 32 patients with reproducible results and normal respiratory quotients the average carbon dioxide output was 8.66 mmol/kg per hour with a wide range of 5-12.4mmol/kg per hour. Consequently the use of an assumed carbon dioxide output can introduce considerable errors in the triolein breath test. This study highlights drawbacks of the triolein breath test, particularly problems in using an assumed carbon dioxide output for its calculation, its inability to predict the severity of malabsorption and the nature of the dietary load used.
Correspondence to: Andrew Duncan, Gastroenterology Unit, Royal Infirmary, Glasgow, G31 2ER, UK.
52 Introduction
In 1979 Newcomer et al., studied three triglycerides - - trioctanoate, tripaimitate and trioleate m as fat probes in J4C breath tests for malabsorption and found that trioleate gave the best results [1]. Since this report there has been considerable interest in the triolein breath test, which has now been assessed several times [2-9]. These evaluations, however, show a considerable divergence of opinion as to the clinical vall~e of this test (Table I). Possible reasons for these discrepancies are the varied designs and methods of analysing the various studies. Most of the reports use predictive values for evaluating the data with results varying from 42 to 100% for sensitivity and 86 to 100% for specificity. The evaluation of data on the basis of sensitivity and specificity alone can be misleading. For example, a small change in the reference range of the 'gold standard' (Table I shows that there is a wide variation in what level of faecal fat is considered normal) can have a sizeable effect on the rate of false positives and negatives. Some studies also calculate the degree of correlation with the reference method and wide variations have again been reported (Table I). For example, no correlation was found in a study which quotes good predictive values [5] and a good correlation in a study which shows much poorer predictive values [7]. The choice of control population also varies between studies. Healthy volunteers, non-gastroenterology patients and gastroenterology patients with normal faecal fat have variously been used. Another study recruited patients with irritable bowel disease as controls on the assumption that their absorptive capacity is normal, a premise which has since been shown to be false [10]. Finally, a variety of protocols has been employed with large quantitative and qualitative differences in both the fat load given and the duration of the test. Perhaps the greatest difficulty in evaluating the triolein breath test is the choice of an appropriate gold standard method. Unfortunately there is no test which can be considered definitive. In the absence of a more suitable test, the measurement of faecal fat has been mainly used t~r comparisons, despite its well known pitfalls [11]. In ~his study we have evaluated the triolein breath test using a reference system which classifies each patient as a normal or malabsorber on the basis of three fat absorption tests, quantitative faecal fat measurement and two dual isotope fat absorption (DIFA) tests [12,13]. The triolein breath test is usually ca|culated using an assumed carbon dioxide (CO2) value. In this study CO2 outputs were measured to determine the extent of any e~or introduced by making this assumption. Methods and Patients Methods For 2 days prior to the test and during the 3 days of the stool collection, the patient was given a high fat diet containing 70-100 g/day of fat. After an overnight fast the patient swallowed gelatin capsules containing 400 kBq of [tdC]triolein (18.5-30 GBq/mmol) and 2000 kBq of [3H]triether (970 GBq/mmol, both from
53
0
m
ZZ
Z
,,.. ~
,,.. ,,.. ,,,.
,,,.
Z
Z
o.~
,.¢
Q.I
.g g
,=
.,o 0,
I=
..,
,.¢ m
=I ¢,)
54 Amersham International, England). A 20 g fat load in the form of double cream and 1000 kBq of [StCr]chromium chloride (0.3-2.4 GBq/mol, Amersham) dissolved in 50 ml of water were taken at the same time. At 0, 1, 2, 3, 4, 5 and 6 h end-expiratory samples of breath were collected into methylbenzethonium hydroxide (Sigma Chemical Co, England) [14]. Factors known to affect CO2 production, such as smoking and exercise [15], were minimised. The percentage t4CO2 excretion in the hourly breath samples was calculated using an assumed CO2 output of 9 mmol/kg per hour [14]. The triolein breath test result was taken as the peak of 14CO2 excretion [1]. In patients who found difficulty in collecting end-expired breath, a mixture of endexpired and normally expired breath was usually collected. To determine whether these collections were valid, 24 samples each of end-expired breath and normal breath samples were collected from seven patients who could collect end-expired breath in the proper way. During the breath test, three or four 2-min samples of breath were collected into Douglas bags via a low dead-space Hans-Rudolph valve and the CO 2 and oxygen were measured. Hyperventilation was minimised by encouraging patients to watch television or read to reduce self-awareness of their breathing rate. The CO2 output and respiratory quotient were measured [16]. No patients had respiratory complaints s~,,ch as emphysema or kyphoscoliosis which could cause abnormal respiratory quotients. 'A 3 day stool collection was homogenised and 25 ml of 154 mmol/I saline and 37.5 ml of chloroform:methanol (2:1 by volume) was added to 3-6 g of homogenate. After refluxing for I h the refluxate was cooled and centrifuged (1500 x g for 5 min). Volumes of 0.5, 1.0 and 1.5 ml of the chloroform layer were transferred to Ocounting glass vials, evaporated off, and the dried extract exposed to light until decolourised, t4C and 3H were measured and quench corrections made using chemically quenched standards (Canberra Packard, Pangbourne, England). 5tCr was measured in 5 g portions of homogenate. The DIFA tests using [3H]triether or [5tCr]chromium chloride as non-absorbable marker (DIFA-3H, DIFA-stCr, respectively) were calculated as the percentage 14C absorbed [12,13]. Faecal fat excretion was measured by the van de Kamer method [17]. On the first twenty patient tests performed, "y-radioactivity was measured in the chloroform layer to determine if StCr was a potential interferent of liquid scintillation counting. Known amounts of [3Hltriether and [ t4C]triolein were added to eleven stools in order to measure their recovery. The extraction procedure was performed in duplicate on the first twenty-five patient tests to enable imprecision to be measured. Fifteen stool samples were analysed by this extraction procedure and by a biological oxidation method. Patients
The triolein breath test and the three reference tests of fat absorption were performed on 61 consecutive patients. Thirty-seven patients who had no evidence of fat malabsorption on the basis of DIFA-3H, DIFA-stCr and faecal fat results were
55
used as a control group. The reference ranges for these three fat absorption tests were greater than 95% for DIFA tests and less than 25 mmol/day (equivalent to about 7 g/day) for faecal fat. By selecting those patients with abnormal DIFA-3H, DIFA-S~Cr and faecal fat results, a corresponding group of sixteen patients with malabsorption was established. In the remaining eight patients DIFA results and faecal fat results did not agree. Malabsorbers were categorised into four groups according to the severity of malabsorption: (a) mild malabsorption-faecal fat of 25-35 retool/day and DIFA of 90-95%; Co) moderate malabsorption-faecai fat of 35-60 retool/day and DIFA of 75-90%; (c) severe malabsorption-faecal fat of 60-100 mmoFday and DIFA of 50-75%; and (d) gross malabsorption-faecal fat over 100 retool/day and DIFA less than 50%. Nine malabsorbers could be classified using this system and in four others the/'aecai fat and DIFA results were in adjacent groups. The latter results were not excluded but were placed in intermediate groups. Results Method assessment
No StCr was measurable in chloroform extracts and so it was concluded that 3H and t4C could be confidently measured in the chloroform layer without contamination from StCr which was found predominantly in the solid interphase. The recoveries of [3Hltriether and [~4Cltriolein were 104.7% (S.E.M. = 3.7) and 98.4% (S.E.M. = 3.1), respectively. The imprecision of the DIFA-3H test varied with different degrees of malabsorption (Table 11) because of the nature of the equation used for its calculation. Thus when absorption approached 100% large differences in the 31-I/14Cratio produ:,ed relatively small changes in the percentage absorption. When the percentage absorption is low the converse is the case. In general the imprecision of the test was acceptable. Fortunately the poor imprecision found at very low levels of absorption has no clinical implications. There was a good correlation (r = 0.99) between DIFA-3H results obtained by using the extractioa procedure described and biological oxidation (Spearman rank order correlation test). No statistical difference (P = 0.99) in ~4C was found between end-expired and normal-expired breath collections OVilcoxon matched-pairs signed-rank test).
TABLE !i Precision of DIFA-3H at differing degrees of malabsorption DIFA-3H range
N
Mean
S.D.
C.V.
90-100% 75-900/0 < 20%
8 12 5
97.8% 83.3% 9.6%
0.78 3.3 2.2
0.800/0 3.9% 23%
56
Hence, it is not imperative that the breath samples collected are end-expiratory. Nevertheless for the purposes of this study, end-expiratory breath samples were collected whenever possible although results from patients who had difficulty in providing such samples were not excluded. Triolein breath test
The initial results of the triolein breath test in controls and malabsorbers showed a considerable overlap between the control and malabsorption groups (Fig. I). Such a degree of overlap would make the triolein breath test of little value but before vilifying it we w¢re keen to ensure that such unimpressive results could not have been caused by any technical errors. In particular we were concerned that the 6-h duration of the test might have been insufficient to 'catch" all of the peaks. When the times of peak t4CO2 excretion were studied it was found that these were reached during the fourth hour in 2 (3%) and during the fifth hour in 12 (20%), but that in most patients (77%) the peak in t4CO2 excretion came at the sixth, and last, hour of collection. It was possible therefore that in some patients a peak might have been reached during the seventh or eighth hour had these samples been collected. The possibility of having missed late peaks prompted us to analyse the pattern of t4CO2 excretion in more detail and so the graph of t4CO2 with time was plotted for each test. In general, three graph types were produced showing either a peak in 14CO2 excretion (Fig. 2A), plateauing of the curve at about 6 h (Fig. 2B), or a late exponential rise in breath t4CO2 (Fig. 2C). It seemed likely that in those patients whose breath t4CO2 excretion was still apparently increasing at 6 h the peak might have been missed. In addition, with this exponential peak type the earlier t4CO2
000 0
6
8
Q
".¢ 0o
°.p E
0 qP 41, 0
Controls
Malabsod)ers
Fig. I. Peak 14CO2 triolein breath test results in control patients and malabsorbers: (+) t4C02 peaks excreted during the six hours of the test; (0) probable late t4CO2 peaks.
57
results at the second and third hours were significantly lower, P < 0.01 and P < 0.001, respectively (Mann-Whitney U-test), than with the other two peak forms also suggesting that there was a delay in 14CO2 excretion. With the resultant suspicion that some results were underestimates, all patient data were reviewed in order to select those t4CO2/time curves which corresponded
0
1
2
3
4
5
6
Time (hours) A
W
,-2 J=
t
m 0 1
0
2
3
4
5
6
Time (hours)
0
1
2
3
4
5
Time (hours)
6
Fig. 2. Typical graphs of breath t4CO~ excretion versus time found in the triolein breath test: (A) obvious peak of breath t4CO2 during the 6-h time course of the test: (B) plateau of breath t4co2 at 6 h; (C) late exponential rise of breath t4CO2.
58
to Fig. 2C. Of the 53 patients investigated 17 were found to have exponential graphs of this type. In Fig. 1 these patients are represented by open circles. In the control group, 14 patients gave an exponential type of t4CO2 excretion suggestive of a late t4CO2 peak. Of the 10 control results which overlapped with the malabsorption group eight could be explained on the basis of a probable late 14CO2 peak. In the malabsorption group, three '4CO2/time graphs were increasing at 6 h suggesting a late peak. In only one of these was the result high enough (2.2% t4CO2 dose/hour) that it might have been misclassified as a false negative had a later breath sample been collected. The control and malabsorption groups were sub-divided into two categories in which (a) the t4CO2 peak was late and had probably been missed and (b) those whose peaks were valid since they fell within the six hours of the test. The means of these data and statistical comparisons of groups are shown in Table III. In the case of the control group the late peaks are significantly lower (Mann-Whitney Utest). This provided further indirect evidence that patients with exponential graphs of "SCO2excretion versus time produced late t4co 2 peaks resulting in their triolein breath test results being underestimated. The results of triolein breath tests in controls and malabsorbers were replotted omitting all results in which the peak t4CO2 excretion had probably been missed. When this was done there was an improved resolution between controls and malabsorbers (Fig. 3). The lower limit of normal was set at 2.6 (i.e. mean minus two S.D.s). Figure 3 also shows the results of the triolein breath test in patients with differing severities of malabsorption. A poor degree of correlation (r = -0.34) was found between the various severities of malabsorption and peaks of t4CO2 (Spearman rank order correlation test). The triolein breath test was correlated with the DIFA-3H and DIFA-StCr tests and with the quantitative excretion of faecal fat. A moderate correlation was produced for each: r = 0.64 for triolein breath test versus DIFA-3H; r = 0.51 for triolein breath test versus DIFA-stCr; and r = -0.60 for triolein breath test versus faecal fat excretion (Spearman rank order correlation test).
TABLE I!I Comparisons of triolein breath test results characterised by late or valid peaks in controls and malabsorbers Group
Menn (% dose/hour)
S.D.
Controls (valid peaks) Controls (late peaks)
5.18% ~.03%
1.29 ) 0,9~
<0.0001
Malabsorl~rs (valid peaks) Malabsorbers (late peaks)
0.84% !.05%
0.93 1 1.07
NS
P
J
59
04
0 Malabsorbers -¢
4
JD
o. 2
¢' •
Controls
_'o
-® 0.~
•o o
a.
~ :)
m cO
41. g'
mm (.~
Fig. 3. Peak t4CO2 triolein breath test results (excluding probable late t4CO2 peaks) in control patients and malabsorbers. Lower limit of normal = 2.6%.
Results in patients with metabolic or respiratory disorders Unlike many previous studies we did not exclude patients because of metabolic or respiratory illnesses despite the theoretical possibility of invalid results [18,19]. The triolein breath test was performed in seven patients with known diabetes, two of whom were poorly controlled, five patients with varying degrees of alcohol;,c liver disease and one patient each with chronic obstructive airways disease and abnormal ~ respiratory function due to cystic fibrosis. Of the seven diabetics five (716~o) produced late n4CO2 peaks as did three of the four (75%) patients with liver disease and the patient with cystic fibrosis. The incidence of late peaks in the other patients studied and who as far as was known did not have any metabolic abnormalities, was 17%. When these data were analysed statistically there was a significantly increased incidence oflate peaks in patients with diabetes (P < 0.01) and liver disease (P < 0.05) (Fischer's exact probability test). Carbon dioxide output Carbon dioxide output was measured in triplicate or quadruplicate in each of 55 patients aged from 12 to 88 years (mean = 53). Nineteen (35%) were excluded because the respiratory quotient was high indicating possible hyperventilation. In this group the reproducibility of results was relatively poor (S.D. = 1.94, C.V. = 14.7%). Results were also discounted in a further four (8%) in whom the respiratory quotient was low. In the 32 patients with normal repiratory quotient the precision of CO2 measurement was satisfactory (S.D. = 1.05, C.V. = 6.7%).
60 TABLE IV Absorption test results where faecal fat and DIFA test conflict Patient number
DIFA-3H
DIFA-sJcr
Faecal fat (retool/day)
Breath test (% dose/hour)
! 2 3 4 5 6 7 8
! 00% ! 00% 100% 990/0 100% 76%o 73% 990/0
! 00% ! 00% ! 00% 990/o i 00%o 890/0 87% 99.50/0
32 36 3i 39 78 19 24 29
4.4% 6.3% 3.2%o 4. 70/0 4.8% O.5% 2.50/0 2.8%
The average CO2 outputs in the 32 patients with normal respiratory quotients was 8.66 mmol/kg per hour (S.D. - 1.93) with a range of 5-12.4 mmol/kg per hour. There was no significant difference in CO a output between the eight malabsorbers (mean = 8.65, S.D. = 2.22) and 24 controls (mean = 8.67, S.D. = 1.87) who made up this population (Mann-Whitney U-test).
Results in patients with conflicting faecal fat and DIFA tests In the preliminary part of this study patients were categorised into control or malabsorption groups according to results of their DIFA tests and faecal fat excretion. On eight occasions conflicting results made it impossible to group the patients and so these results were excluded. The analyses were repeated to exclude the possibility of technical error but on each occasion similar results were produced. Table IV shows that slightly elevated faecal fat results were found in four patients with normal DIFA tests and normal triolein breath test results (numbers 1, 2, 3 and 4). In one case (number 5) the DIFA results were normal as was the triolein breath test (at 4.8) but the faecal fat was very appreciably elevated, at 78 mmol/day. When (his patient's case-notes were reviewed there seemed no obvious reason for such a gross disparity in results. However, this patient was a diagnostic enigma. He had been admitted for investigation of leg and arm oedema and was found to have hypoalbuminaemia and a slight protein losing enteropathy. He had no gastrointestinal symptoms, his protein status and oedema spontaneously settled while on the ward and after a follow-up of one year he remains well. Two patients (numbers 6 and 7) had normal faecal fat excretions while the other absorption tests were abnormal. Finally in one patient the DIFA results were apparently falsely normal (number 8). Discussion In previous evaluations of the triolein breath test an inherent problem clouding their interpretation has been the absence of a definitive reference method. By and large, the method chosen in previous assessments has been the meas,,~ement of faecal
61
fat, despite its well-known drawbacks [11]. in the present report we attempted to solve this problem by classifying patients on the basis of three fat absorption techniques employing two different principles. The two dual isotope methods used are direct, independent of complete stool collection and have previously been evaluated with favourable results [12,13]. In terms of simplicity and aesthetics, the triolein breath test was preferable to the faecal fat and DIFA tests. Other advantages are that it requires a minimum of analytical effort, provides results the following day, is well tolerated by patients, and is easily performed on out-patients. On the basis of the reviewed data in this study (i.e. when probable late peaks of 14CO2 were excluded) the triolein breath test produced a sensitivity of 100% and specificity of 96%. A moderate degree of correlation was also found with each of the three comparative fat absorption tests. In these respects our findings were comparable with other reports [1,2,4]. However, a limitation of assessing the triolein breath test solely in these terms is that no information is obtained on the ability of the test to identify patients with only mild degrees of malabsorption. Using our classification system we were able to evaluate the ability of the triolein breath test to differentiate between different severities of malabsorption. Unfortunately only one patient with a mild degree of malabsorption was found and in this instance the triolein breath test result was appropriate. One other study has attempted to gauge the ability of the triolein breath test to detect mild malabsorption and found it to be inadequate [8]. In this study we found that the triolein breath test was not capable of predicting the severity of malabsorption. For example, similar triolein breath test results of around 1.6% gere found in three patients whose severities of malabsorption were mild, moderate/severe and gross (Fig. 3). This study has illustrated that when triolein breath test results are expressed as peaks of 14CO2 excretion, then it must be known that the peak is excreted during the time course of the test. We found evidence to suggest that in some patients the peak breath t4CO2 occurred af~er the 6-h duration of the test. Studying the pattern of the 14CO2/time curve appears to be a useful means of monitoring this although definitive evidence of late peaks can only be gained by extending the duration of the test. Like us, several other authors have followed the protocol of Newcomer et al. and collected breath samples for 6 h. It is possible that in such evaluations, especially those in which the peak 14CO2 results were found predominantly in the sixth hour [I,3], that the peak 14CO2 excretion may also unwittingly have been missed. Further evidence of this possibility can be found in the report by Benino et al. who collected breath samples for 8 h and found that in only 60% was a peak 14CO2reached by the sixth hour [7]. The average peak m4CO2in their control group was 5.6% which is higher than in other studies, 4.23% [l] and 4.3% [3], and this may also be explained by late peaks being missed in the latter studies. By excluding late peaks the average 14CO2 peak in our control group increased from 4.4% to 5.2%. Differences in the size and time of peak t4CO2 peaks may also be related to variations in the fat load administered but this was similar in the above studies (20-30 g of fat) and so comparisons are probably valid. The time at which the 14CO2 peak occurs is dependent on the rate at which the [14C]triolein is digested, absorbed and metabolised. This in turn is controlled by
factors such as gastric emptying, metabolic rate and the availability of alternative energy sources. Since these factors are ultimately dependent on the dietary load it is possible that peaks of 14CO2 may be expedited by altering the dietary load. The small loading meal given at the start of the triolein breath test serves to provide a realistic challenge to absorptive capacity. Unfortunately little research has been done to establish the optimum size and content of the meal and consequently large variations are found from study to study. Fat is metabolised more slowly by the body than ~arbohydrate or protein and so has a lesser effect on tb ~ metabolism of [t4C]triolein. Consequently it has been considered the best choice as test load. However, fat delays gastric emptying [20] and so its use probably delays t4CO2 peaks [I ]. This factor probably contributed to the suspected late peaks found in this study. Increasing the calorific value of the dietary load, especially with carbohydrates which are known to inhibit fat oxidation [21,22], results in further delay in the excretion of t4CO2. For example Newcomer et al. showed that the peak of 14CO2 was substantially lower and occurred up to three hours later when a 50 g fat meal was given instead of the usual 20 g [1]. In another study when a 60 g fat meal as butterer~ toast was given the t4CO2 excretion was delayed with only 42% of the subjects excreting a peak by the ninth hour [2]. In this case the cumulative 14CO, excretion gave a better discrimination than the peak t4CO2. However, when cumulative results are calculated breath must be collected for extended time periods of eight or nine hours thus limiting the practicability of this approach. The dietary load used by Turner et al. [6] apparently expedited 14CO2 excretion, with peaks being achieved within 6 h in almost all (97%) controls. For reasons of palatability their diet comprised 10 g of carbohydl'ate, 19.3 g of fat and 1.4 g of protein as a lemon mousse. This dietary load thus appears to be more suitable than the predominant fat diets employed by other studies includirg the present one. In order to calculate triolein breath test results the CO2 output must be known [14]. However, the option of measuring the CO, output on each patient is not entirely practicable: as well as making the test more complex, the technology for measuring CO 2 output is not always available. Consequently a value of 9mmol/kg per hour based on results in normal fasting subjects [23] is usually assumed, in 32 patients with satisfactory breath collections we found the mean CO, output (8.66mmol/kg per hour) to be in accordance with previous findings of 8.7 [23] and 8. I [241. We also demonstrated a wide range of results, from 5-12.4 mmol/kg per hour. Consequently, the use of an assumed CO, output introduces a potentially sizeable error in the triolein breath test calculation. For example, by use of an assumed value the triolein breath test results of the patients at the extreme end of our range of CO2 outputs (i.e. 5 and 12.4 mmol/kg per hour) were overestimated by 80% and underestimated by 27%, respectively. There is clearly a potential for misclassification tn occur and at our lower limit of normal of 2.6% there is an effective grey area of from 1.45 to 3.6%. The triolein breath test is considered to be an inappropriate test in patients with metabolic disorders [18,19]. In support of this we found that patients with liver disease and diabetes were more likely to produce late 14CO2 peaks probably due to slower metabolism of [ t4C]triolein. Of the patients on whom fat absorption studies
63 were requested a sizable proportion (23%) had known respiratory or metabolic complaints. Pederson also found that such patients constituted a significant proportion (46%) of those investigated [8]. Despite the apparently favourable findings of previous evaluations, we feel that a complete appraisal of the triolein breath test is not possible with the design of such studies. Consequently some of the enthusiasm for the triolein breath test may not be justified. The present study also contained drawbacks: (a) most of the patients recruited had substantial degrees o f malabsorption and so the ability of the triolein breath test to detect milder degrees of malabsorption could not be assessed; (b) the evidence for late peaks, although strong, was not definitive and so the conclusions based on their exclusion can only be considered as persuasive rather than conclusive. We found the predictive value and correlations of triolein breath test with the other fat absorption methods to be encouraging. By adjusting the dietary load, perhaps using the meal designed by T u r n e r et al. the suspected problems in missing late t4CO2 peaks may be minimised. A more troublesome finding was the wide individual variations found in CO2 outputs and the practical difficulties in its compensation. Failure to correct for individual CO2 outputs can introduce considerable errors into the triolein breath test. Coupled with an inability of the triolein breath test to predict the severity of malabsorption and the relatively high frequency of disorders where its use is inappropriate, we now have serious misgivings as to its cole in the investigation of malabsorption.
References i NewcomerAD, Hofmann AF, DiMagno El>,Thomas PJ, Carlson GL. Triolein breath test. A sensitive and specific test for fat malabsorption. Gastroenterology 1979;76:6-13. 2 WestPS, Levin GE, Griffin GE, Maxwell JD. Comparison of simple screeningtests for fat malabsorption. Br Med J 1981;282:1501-1504. 3 EinarssonK, Bjorkhem i, Eklof R, Blomstrand R. t4C-Triolein breath test as a rapid and convenient screening test for fat malabsorption. Scand J Gastroenterol 1983;18:9-12. 4 AkessonB, Floren C-H. Use of the triolein breath test for the demonstration of fat malabsorption in coeliac disease. Scand J Gastroenterol 1984;19:307-314. 5 ButlerRN, lawson MJ, Gehling NJ, Grant AK. Clinical evaluation of the t4C-trioleinbreath test: a critical analysis. Aust N Z J Med 1984;14:ii I-! 13. 6 Turner JM, lawrence S, Fellows IW, Johnson !, Hill PG, Holmes GKT. t4C-Trioleinabsorption: a useful test in the diagnosis of malabsorption. Gut 1987;28:694-700. 7 lknino L, Scuro LA, Menini E, et al., is the t4C-trioleintest useful in the assessmentof malabsorption in clinical practice? Digestion 1984:29:91-97. 8 PedersenNT, Andersen BN, Marqversen J. Estimation of [t4Cltriolein assimilationas a test of lipid assimilation. Breath test or measurement of serum radioactivity. Scand J Gastroenterol 1982,17:309-316. 9 Luey K, Pattinson NR, Hinton D, Cook HB, Campbell CB. Comparison of three methods to estimate steatorrhoea. N Z Med J 1981;93:36-38. 10 MylvaganumK, Hudson PR, Ross A, Williams CP. 14C triolein breath test: a routine test in the gastroenterology clinic? Gut 1986;27:1347-!352. II RosenbergIH, Sitrin MD. Screening for fat malabsorption. Ann Intern Med 1981;95:776-777. 12 Nelson LM, McKenzie.IF, Russell RI. Measurement of fat absorption using [ 3H]glycerol tricther and [taCiglycerol trioleate in man. Clin Chim Acta 1980;103:325-334 13 PedersenNT. Estimatiov.of lipid assimilation from faecal samples using [14C]trioleinas tracer and [51Crlchromium chloride as non-absorbable marker. Scand J Clin Lab Invest 1983;43:323-328.
64 14 Fromm H, Hofmann AF. Breath tests for altered bile-acid metabolism. Lancet 1971;ii:621-625. 15 Kachroo RL, Rehani MM, Kachroo N, Khuroo MS, Zarger SA. Optimising the value of C-14 breath test. Indian J Mud Res 1987;86:! 19-123. 16 Cotes JE. Lung Function. Assessment and application in medicine. 4th Edn. Oxford: Blackwell Scientific Publications, 1979. 17 Van de Kamer JH, Huinink HTB, Weijers HA. Rapid method for the determination of fat in faeces. J Bioi Chum 1949;177:347-355. 18 Caspary WF. Breath tests. Clin Gastroenterology 1978;7:351-374. 19 Tomkin GH, Bell TK, Hadden DR. Evaluation of malabsorption test using t4C-triglyceride, lr J Mud Sci 1971;140:449-454. 20 McCailum RW. Motor function of the stomach in health and disease. In: MH Sleisenger and JS Fordtran eds. Gastrointestinal disease m pathophysiology, diagnosis, management. 4th edn. Philadelphia: W$ Saunders Co. 1989;675-713. 21 Lossow WJ, Chaikoff I L. Carbohydrate sparing of fatty acid oxidation. ! The relation of fatty acid chain length to the degree of sparing. 11 The mechanism by which carbohydrate spares. Arch Biochem 1955;57:23-38. 22 Bragdon JH. 14CO2excretion after the intravenous administration of labeled chylomicrons in the rat. Arch Biochem 1958:75:528-539. 23 Winchell HS, Stahelin H, Kusubov N e t al. Kinetics of CO2-HCO3 in normal adnlt males. J Nucl Med 1970;I 1:711-715. 24 King CE, Toskes PP. Alteration of carbon dioxide production during nonfasting isotopic CO 2 breath tests: concise communication. J Nucl Mud 1981;22:955-958.
51
CCA05202
Limitations of the triolein breath test A n d r e w D u n c a n a, A l i s o n C a m e r o n a, Michael .I. Stewart b a n d R o b i n I. RusseU a aGastroenterology Unit and blnstitute of Biochemistry. Royal Infirmary. Glasgow, G31 2ER (UK) (Received 23 July 1991; revision received 31 October 1991; accepted 8 November 1991)
Key words: Malabsorption; Fat absorption test; Carbon dioxide output; Breath test; Dual isotope fat absorption test
Summary Patients being investigated for intestinal absorptive capacity were classified as norreals or malabsorbers on the basis of three fat absorption tests. Malabsorbers were further classified as mild, moderate, severe or gross according to severity of malabsorption. Using this classification system the triolein breath test was evaluated in 53 patients. Seventeen patients were excluded because their graph of percentage breath [14C]carbon dioxide versus time was exponential indicating that the peak [~4C]carbon dioxide may be occurring later than the six hour duration of the test. The sensitivity and specificity of the triolein breath test were found to be 100% and 96%, respectively and moderate correlations with the individual fat absorption tests were found. However, the breath test was limited in its capacity to predict the severity of malabsorption. Carbon dioxide output was also measured in order to determine the applicability of using an assumed value. The respiratory quotient and variability of results were high in nineteen patients indicating possible hyperventilation. In 32 patients with reproducible results and normal respiratory quotients the average carbon dioxide output was 8.66 mmol/kg per hour with a wide range of 5-12.4mmol/kg per hour. Consequently the use of an assumed carbon dioxide output can introduce considerable errors in the triolein breath test. This study highlights drawbacks of the triolein breath test, particularly problems in using an assumed carbon dioxide output for its calculation, its inability to predict the severity of malabsorption and the nature of the dietary load used.
Correspondence to: Andrew Duncan, Gastroenterology Unit, Royal Infirmary, Glasgow, G31 2ER, UK.
52 Introduction
In 1979 Newcomer et al., studied three triglycerides - - trioctanoate, tripaimitate and trioleate m as fat probes in J4C breath tests for malabsorption and found that trioleate gave the best results [1]. Since this report there has been considerable interest in the triolein breath test, which has now been assessed several times [2-9]. These evaluations, however, show a considerable divergence of opinion as to the clinical vall~e of this test (Table I). Possible reasons for these discrepancies are the varied designs and methods of analysing the various studies. Most of the reports use predictive values for evaluating the data with results varying from 42 to 100% for sensitivity and 86 to 100% for specificity. The evaluation of data on the basis of sensitivity and specificity alone can be misleading. For example, a small change in the reference range of the 'gold standard' (Table I shows that there is a wide variation in what level of faecal fat is considered normal) can have a sizeable effect on the rate of false positives and negatives. Some studies also calculate the degree of correlation with the reference method and wide variations have again been reported (Table I). For example, no correlation was found in a study which quotes good predictive values [5] and a good correlation in a study which shows much poorer predictive values [7]. The choice of control population also varies between studies. Healthy volunteers, non-gastroenterology patients and gastroenterology patients with normal faecal fat have variously been used. Another study recruited patients with irritable bowel disease as controls on the assumption that their absorptive capacity is normal, a premise which has since been shown to be false [10]. Finally, a variety of protocols has been employed with large quantitative and qualitative differences in both the fat load given and the duration of the test. Perhaps the greatest difficulty in evaluating the triolein breath test is the choice of an appropriate gold standard method. Unfortunately there is no test which can be considered definitive. In the absence of a more suitable test, the measurement of faecal fat has been mainly used t~r comparisons, despite its well known pitfalls [11]. In ~his study we have evaluated the triolein breath test using a reference system which classifies each patient as a normal or malabsorber on the basis of three fat absorption tests, quantitative faecal fat measurement and two dual isotope fat absorption (DIFA) tests [12,13]. The triolein breath test is usually ca|culated using an assumed carbon dioxide (CO2) value. In this study CO2 outputs were measured to determine the extent of any e~or introduced by making this assumption. Methods and Patients Methods For 2 days prior to the test and during the 3 days of the stool collection, the patient was given a high fat diet containing 70-100 g/day of fat. After an overnight fast the patient swallowed gelatin capsules containing 400 kBq of [tdC]triolein (18.5-30 GBq/mmol) and 2000 kBq of [3H]triether (970 GBq/mmol, both from
53
0
m
ZZ
Z
,,.. ~
,,.. ,,.. ,,,.
,,,.
Z
Z
o.~
,.¢
Q.I
.g g
,=
.,o 0,
I=
..,
,.¢ m
=I ¢,)
54 Amersham International, England). A 20 g fat load in the form of double cream and 1000 kBq of [StCr]chromium chloride (0.3-2.4 GBq/mol, Amersham) dissolved in 50 ml of water were taken at the same time. At 0, 1, 2, 3, 4, 5 and 6 h end-expiratory samples of breath were collected into methylbenzethonium hydroxide (Sigma Chemical Co, England) [14]. Factors known to affect CO2 production, such as smoking and exercise [15], were minimised. The percentage t4CO2 excretion in the hourly breath samples was calculated using an assumed CO2 output of 9 mmol/kg per hour [14]. The triolein breath test result was taken as the peak of 14CO2 excretion [1]. In patients who found difficulty in collecting end-expired breath, a mixture of endexpired and normally expired breath was usually collected. To determine whether these collections were valid, 24 samples each of end-expired breath and normal breath samples were collected from seven patients who could collect end-expired breath in the proper way. During the breath test, three or four 2-min samples of breath were collected into Douglas bags via a low dead-space Hans-Rudolph valve and the CO 2 and oxygen were measured. Hyperventilation was minimised by encouraging patients to watch television or read to reduce self-awareness of their breathing rate. The CO2 output and respiratory quotient were measured [16]. No patients had respiratory complaints s~,,ch as emphysema or kyphoscoliosis which could cause abnormal respiratory quotients. 'A 3 day stool collection was homogenised and 25 ml of 154 mmol/I saline and 37.5 ml of chloroform:methanol (2:1 by volume) was added to 3-6 g of homogenate. After refluxing for I h the refluxate was cooled and centrifuged (1500 x g for 5 min). Volumes of 0.5, 1.0 and 1.5 ml of the chloroform layer were transferred to Ocounting glass vials, evaporated off, and the dried extract exposed to light until decolourised, t4C and 3H were measured and quench corrections made using chemically quenched standards (Canberra Packard, Pangbourne, England). 5tCr was measured in 5 g portions of homogenate. The DIFA tests using [3H]triether or [5tCr]chromium chloride as non-absorbable marker (DIFA-3H, DIFA-stCr, respectively) were calculated as the percentage 14C absorbed [12,13]. Faecal fat excretion was measured by the van de Kamer method [17]. On the first twenty patient tests performed, "y-radioactivity was measured in the chloroform layer to determine if StCr was a potential interferent of liquid scintillation counting. Known amounts of [3Hltriether and [ t4C]triolein were added to eleven stools in order to measure their recovery. The extraction procedure was performed in duplicate on the first twenty-five patient tests to enable imprecision to be measured. Fifteen stool samples were analysed by this extraction procedure and by a biological oxidation method. Patients
The triolein breath test and the three reference tests of fat absorption were performed on 61 consecutive patients. Thirty-seven patients who had no evidence of fat malabsorption on the basis of DIFA-3H, DIFA-stCr and faecal fat results were
55
used as a control group. The reference ranges for these three fat absorption tests were greater than 95% for DIFA tests and less than 25 mmol/day (equivalent to about 7 g/day) for faecal fat. By selecting those patients with abnormal DIFA-3H, DIFA-S~Cr and faecal fat results, a corresponding group of sixteen patients with malabsorption was established. In the remaining eight patients DIFA results and faecal fat results did not agree. Malabsorbers were categorised into four groups according to the severity of malabsorption: (a) mild malabsorption-faecal fat of 25-35 retool/day and DIFA of 90-95%; Co) moderate malabsorption-faecai fat of 35-60 retool/day and DIFA of 75-90%; (c) severe malabsorption-faecal fat of 60-100 mmoFday and DIFA of 50-75%; and (d) gross malabsorption-faecal fat over 100 retool/day and DIFA less than 50%. Nine malabsorbers could be classified using this system and in four others the/'aecai fat and DIFA results were in adjacent groups. The latter results were not excluded but were placed in intermediate groups. Results Method assessment
No StCr was measurable in chloroform extracts and so it was concluded that 3H and t4C could be confidently measured in the chloroform layer without contamination from StCr which was found predominantly in the solid interphase. The recoveries of [3Hltriether and [~4Cltriolein were 104.7% (S.E.M. = 3.7) and 98.4% (S.E.M. = 3.1), respectively. The imprecision of the DIFA-3H test varied with different degrees of malabsorption (Table 11) because of the nature of the equation used for its calculation. Thus when absorption approached 100% large differences in the 31-I/14Cratio produ:,ed relatively small changes in the percentage absorption. When the percentage absorption is low the converse is the case. In general the imprecision of the test was acceptable. Fortunately the poor imprecision found at very low levels of absorption has no clinical implications. There was a good correlation (r = 0.99) between DIFA-3H results obtained by using the extractioa procedure described and biological oxidation (Spearman rank order correlation test). No statistical difference (P = 0.99) in ~4C was found between end-expired and normal-expired breath collections OVilcoxon matched-pairs signed-rank test).
TABLE !i Precision of DIFA-3H at differing degrees of malabsorption DIFA-3H range
N
Mean
S.D.
C.V.
90-100% 75-900/0 < 20%
8 12 5
97.8% 83.3% 9.6%
0.78 3.3 2.2
0.800/0 3.9% 23%
56
Hence, it is not imperative that the breath samples collected are end-expiratory. Nevertheless for the purposes of this study, end-expiratory breath samples were collected whenever possible although results from patients who had difficulty in providing such samples were not excluded. Triolein breath test
The initial results of the triolein breath test in controls and malabsorbers showed a considerable overlap between the control and malabsorption groups (Fig. I). Such a degree of overlap would make the triolein breath test of little value but before vilifying it we w¢re keen to ensure that such unimpressive results could not have been caused by any technical errors. In particular we were concerned that the 6-h duration of the test might have been insufficient to 'catch" all of the peaks. When the times of peak t4CO2 excretion were studied it was found that these were reached during the fourth hour in 2 (3%) and during the fifth hour in 12 (20%), but that in most patients (77%) the peak in t4CO2 excretion came at the sixth, and last, hour of collection. It was possible therefore that in some patients a peak might have been reached during the seventh or eighth hour had these samples been collected. The possibility of having missed late peaks prompted us to analyse the pattern of t4CO2 excretion in more detail and so the graph of t4CO2 with time was plotted for each test. In general, three graph types were produced showing either a peak in 14CO2 excretion (Fig. 2A), plateauing of the curve at about 6 h (Fig. 2B), or a late exponential rise in breath t4CO2 (Fig. 2C). It seemed likely that in those patients whose breath t4CO2 excretion was still apparently increasing at 6 h the peak might have been missed. In addition, with this exponential peak type the earlier t4CO2
000 0
6
8
Q
".¢ 0o
°.p E
0 qP 41, 0
Controls
Malabsod)ers
Fig. I. Peak 14CO2 triolein breath test results in control patients and malabsorbers: (+) t4C02 peaks excreted during the six hours of the test; (0) probable late t4CO2 peaks.
57
results at the second and third hours were significantly lower, P < 0.01 and P < 0.001, respectively (Mann-Whitney U-test), than with the other two peak forms also suggesting that there was a delay in 14CO2 excretion. With the resultant suspicion that some results were underestimates, all patient data were reviewed in order to select those t4CO2/time curves which corresponded
0
1
2
3
4
5
6
Time (hours) A
W
,-2 J=
t
m 0 1
0
2
3
4
5
6
Time (hours)
0
1
2
3
4
5
Time (hours)
6
Fig. 2. Typical graphs of breath t4CO~ excretion versus time found in the triolein breath test: (A) obvious peak of breath t4CO2 during the 6-h time course of the test: (B) plateau of breath t4co2 at 6 h; (C) late exponential rise of breath t4CO2.
58
to Fig. 2C. Of the 53 patients investigated 17 were found to have exponential graphs of this type. In Fig. 1 these patients are represented by open circles. In the control group, 14 patients gave an exponential type of t4CO2 excretion suggestive of a late t4CO2 peak. Of the 10 control results which overlapped with the malabsorption group eight could be explained on the basis of a probable late 14CO2 peak. In the malabsorption group, three '4CO2/time graphs were increasing at 6 h suggesting a late peak. In only one of these was the result high enough (2.2% t4CO2 dose/hour) that it might have been misclassified as a false negative had a later breath sample been collected. The control and malabsorption groups were sub-divided into two categories in which (a) the t4CO2 peak was late and had probably been missed and (b) those whose peaks were valid since they fell within the six hours of the test. The means of these data and statistical comparisons of groups are shown in Table III. In the case of the control group the late peaks are significantly lower (Mann-Whitney Utest). This provided further indirect evidence that patients with exponential graphs of "SCO2excretion versus time produced late t4co 2 peaks resulting in their triolein breath test results being underestimated. The results of triolein breath tests in controls and malabsorbers were replotted omitting all results in which the peak t4CO2 excretion had probably been missed. When this was done there was an improved resolution between controls and malabsorbers (Fig. 3). The lower limit of normal was set at 2.6 (i.e. mean minus two S.D.s). Figure 3 also shows the results of the triolein breath test in patients with differing severities of malabsorption. A poor degree of correlation (r = -0.34) was found between the various severities of malabsorption and peaks of t4CO2 (Spearman rank order correlation test). The triolein breath test was correlated with the DIFA-3H and DIFA-StCr tests and with the quantitative excretion of faecal fat. A moderate correlation was produced for each: r = 0.64 for triolein breath test versus DIFA-3H; r = 0.51 for triolein breath test versus DIFA-stCr; and r = -0.60 for triolein breath test versus faecal fat excretion (Spearman rank order correlation test).
TABLE I!I Comparisons of triolein breath test results characterised by late or valid peaks in controls and malabsorbers Group
Menn (% dose/hour)
S.D.
Controls (valid peaks) Controls (late peaks)
5.18% ~.03%
1.29 ) 0,9~
<0.0001
Malabsorl~rs (valid peaks) Malabsorbers (late peaks)
0.84% !.05%
0.93 1 1.07
NS
P
J
59
04
0 Malabsorbers -¢
4
JD
o. 2
¢' •
Controls
_'o
-® 0.~
•o o
a.
~ :)
m cO
41. g'
mm (.~
Fig. 3. Peak t4CO2 triolein breath test results (excluding probable late t4CO2 peaks) in control patients and malabsorbers. Lower limit of normal = 2.6%.
Results in patients with metabolic or respiratory disorders Unlike many previous studies we did not exclude patients because of metabolic or respiratory illnesses despite the theoretical possibility of invalid results [18,19]. The triolein breath test was performed in seven patients with known diabetes, two of whom were poorly controlled, five patients with varying degrees of alcohol;,c liver disease and one patient each with chronic obstructive airways disease and abnormal ~ respiratory function due to cystic fibrosis. Of the seven diabetics five (716~o) produced late n4CO2 peaks as did three of the four (75%) patients with liver disease and the patient with cystic fibrosis. The incidence of late peaks in the other patients studied and who as far as was known did not have any metabolic abnormalities, was 17%. When these data were analysed statistically there was a significantly increased incidence oflate peaks in patients with diabetes (P < 0.01) and liver disease (P < 0.05) (Fischer's exact probability test). Carbon dioxide output Carbon dioxide output was measured in triplicate or quadruplicate in each of 55 patients aged from 12 to 88 years (mean = 53). Nineteen (35%) were excluded because the respiratory quotient was high indicating possible hyperventilation. In this group the reproducibility of results was relatively poor (S.D. = 1.94, C.V. = 14.7%). Results were also discounted in a further four (8%) in whom the respiratory quotient was low. In the 32 patients with normal repiratory quotient the precision of CO2 measurement was satisfactory (S.D. = 1.05, C.V. = 6.7%).
60 TABLE IV Absorption test results where faecal fat and DIFA test conflict Patient number
DIFA-3H
DIFA-sJcr
Faecal fat (retool/day)
Breath test (% dose/hour)
! 2 3 4 5 6 7 8
! 00% ! 00% 100% 990/0 100% 76%o 73% 990/0
! 00% ! 00% ! 00% 990/o i 00%o 890/0 87% 99.50/0
32 36 3i 39 78 19 24 29
4.4% 6.3% 3.2%o 4. 70/0 4.8% O.5% 2.50/0 2.8%
The average CO2 outputs in the 32 patients with normal respiratory quotients was 8.66 mmol/kg per hour (S.D. - 1.93) with a range of 5-12.4 mmol/kg per hour. There was no significant difference in CO a output between the eight malabsorbers (mean = 8.65, S.D. = 2.22) and 24 controls (mean = 8.67, S.D. = 1.87) who made up this population (Mann-Whitney U-test).
Results in patients with conflicting faecal fat and DIFA tests In the preliminary part of this study patients were categorised into control or malabsorption groups according to results of their DIFA tests and faecal fat excretion. On eight occasions conflicting results made it impossible to group the patients and so these results were excluded. The analyses were repeated to exclude the possibility of technical error but on each occasion similar results were produced. Table IV shows that slightly elevated faecal fat results were found in four patients with normal DIFA tests and normal triolein breath test results (numbers 1, 2, 3 and 4). In one case (number 5) the DIFA results were normal as was the triolein breath test (at 4.8) but the faecal fat was very appreciably elevated, at 78 mmol/day. When (his patient's case-notes were reviewed there seemed no obvious reason for such a gross disparity in results. However, this patient was a diagnostic enigma. He had been admitted for investigation of leg and arm oedema and was found to have hypoalbuminaemia and a slight protein losing enteropathy. He had no gastrointestinal symptoms, his protein status and oedema spontaneously settled while on the ward and after a follow-up of one year he remains well. Two patients (numbers 6 and 7) had normal faecal fat excretions while the other absorption tests were abnormal. Finally in one patient the DIFA results were apparently falsely normal (number 8). Discussion In previous evaluations of the triolein breath test an inherent problem clouding their interpretation has been the absence of a definitive reference method. By and large, the method chosen in previous assessments has been the meas,,~ement of faecal
61
fat, despite its well-known drawbacks [11]. in the present report we attempted to solve this problem by classifying patients on the basis of three fat absorption techniques employing two different principles. The two dual isotope methods used are direct, independent of complete stool collection and have previously been evaluated with favourable results [12,13]. In terms of simplicity and aesthetics, the triolein breath test was preferable to the faecal fat and DIFA tests. Other advantages are that it requires a minimum of analytical effort, provides results the following day, is well tolerated by patients, and is easily performed on out-patients. On the basis of the reviewed data in this study (i.e. when probable late peaks of 14CO2 were excluded) the triolein breath test produced a sensitivity of 100% and specificity of 96%. A moderate degree of correlation was also found with each of the three comparative fat absorption tests. In these respects our findings were comparable with other reports [1,2,4]. However, a limitation of assessing the triolein breath test solely in these terms is that no information is obtained on the ability of the test to identify patients with only mild degrees of malabsorption. Using our classification system we were able to evaluate the ability of the triolein breath test to differentiate between different severities of malabsorption. Unfortunately only one patient with a mild degree of malabsorption was found and in this instance the triolein breath test result was appropriate. One other study has attempted to gauge the ability of the triolein breath test to detect mild malabsorption and found it to be inadequate [8]. In this study we found that the triolein breath test was not capable of predicting the severity of malabsorption. For example, similar triolein breath test results of around 1.6% gere found in three patients whose severities of malabsorption were mild, moderate/severe and gross (Fig. 3). This study has illustrated that when triolein breath test results are expressed as peaks of 14CO2 excretion, then it must be known that the peak is excreted during the time course of the test. We found evidence to suggest that in some patients the peak breath t4CO2 occurred af~er the 6-h duration of the test. Studying the pattern of the 14CO2/time curve appears to be a useful means of monitoring this although definitive evidence of late peaks can only be gained by extending the duration of the test. Like us, several other authors have followed the protocol of Newcomer et al. and collected breath samples for 6 h. It is possible that in such evaluations, especially those in which the peak 14CO2 results were found predominantly in the sixth hour [I,3], that the peak 14CO2 excretion may also unwittingly have been missed. Further evidence of this possibility can be found in the report by Benino et al. who collected breath samples for 8 h and found that in only 60% was a peak 14CO2reached by the sixth hour [7]. The average peak m4CO2in their control group was 5.6% which is higher than in other studies, 4.23% [l] and 4.3% [3], and this may also be explained by late peaks being missed in the latter studies. By excluding late peaks the average 14CO2 peak in our control group increased from 4.4% to 5.2%. Differences in the size and time of peak t4CO2 peaks may also be related to variations in the fat load administered but this was similar in the above studies (20-30 g of fat) and so comparisons are probably valid. The time at which the 14CO2 peak occurs is dependent on the rate at which the [14C]triolein is digested, absorbed and metabolised. This in turn is controlled by
factors such as gastric emptying, metabolic rate and the availability of alternative energy sources. Since these factors are ultimately dependent on the dietary load it is possible that peaks of 14CO2 may be expedited by altering the dietary load. The small loading meal given at the start of the triolein breath test serves to provide a realistic challenge to absorptive capacity. Unfortunately little research has been done to establish the optimum size and content of the meal and consequently large variations are found from study to study. Fat is metabolised more slowly by the body than ~arbohydrate or protein and so has a lesser effect on tb ~ metabolism of [t4C]triolein. Consequently it has been considered the best choice as test load. However, fat delays gastric emptying [20] and so its use probably delays t4CO2 peaks [I ]. This factor probably contributed to the suspected late peaks found in this study. Increasing the calorific value of the dietary load, especially with carbohydrates which are known to inhibit fat oxidation [21,22], results in further delay in the excretion of t4CO2. For example Newcomer et al. showed that the peak of 14CO2 was substantially lower and occurred up to three hours later when a 50 g fat meal was given instead of the usual 20 g [1]. In another study when a 60 g fat meal as butterer~ toast was given the t4CO2 excretion was delayed with only 42% of the subjects excreting a peak by the ninth hour [2]. In this case the cumulative 14CO, excretion gave a better discrimination than the peak t4CO2. However, when cumulative results are calculated breath must be collected for extended time periods of eight or nine hours thus limiting the practicability of this approach. The dietary load used by Turner et al. [6] apparently expedited 14CO2 excretion, with peaks being achieved within 6 h in almost all (97%) controls. For reasons of palatability their diet comprised 10 g of carbohydl'ate, 19.3 g of fat and 1.4 g of protein as a lemon mousse. This dietary load thus appears to be more suitable than the predominant fat diets employed by other studies includirg the present one. In order to calculate triolein breath test results the CO2 output must be known [14]. However, the option of measuring the CO, output on each patient is not entirely practicable: as well as making the test more complex, the technology for measuring CO 2 output is not always available. Consequently a value of 9mmol/kg per hour based on results in normal fasting subjects [23] is usually assumed, in 32 patients with satisfactory breath collections we found the mean CO, output (8.66mmol/kg per hour) to be in accordance with previous findings of 8.7 [23] and 8. I [241. We also demonstrated a wide range of results, from 5-12.4 mmol/kg per hour. Consequently, the use of an assumed CO, output introduces a potentially sizeable error in the triolein breath test calculation. For example, by use of an assumed value the triolein breath test results of the patients at the extreme end of our range of CO2 outputs (i.e. 5 and 12.4 mmol/kg per hour) were overestimated by 80% and underestimated by 27%, respectively. There is clearly a potential for misclassification tn occur and at our lower limit of normal of 2.6% there is an effective grey area of from 1.45 to 3.6%. The triolein breath test is considered to be an inappropriate test in patients with metabolic disorders [18,19]. In support of this we found that patients with liver disease and diabetes were more likely to produce late 14CO2 peaks probably due to slower metabolism of [ t4C]triolein. Of the patients on whom fat absorption studies
63 were requested a sizable proportion (23%) had known respiratory or metabolic complaints. Pederson also found that such patients constituted a significant proportion (46%) of those investigated [8]. Despite the apparently favourable findings of previous evaluations, we feel that a complete appraisal of the triolein breath test is not possible with the design of such studies. Consequently some of the enthusiasm for the triolein breath test may not be justified. The present study also contained drawbacks: (a) most of the patients recruited had substantial degrees o f malabsorption and so the ability of the triolein breath test to detect milder degrees of malabsorption could not be assessed; (b) the evidence for late peaks, although strong, was not definitive and so the conclusions based on their exclusion can only be considered as persuasive rather than conclusive. We found the predictive value and correlations of triolein breath test with the other fat absorption methods to be encouraging. By adjusting the dietary load, perhaps using the meal designed by T u r n e r et al. the suspected problems in missing late t4CO2 peaks may be minimised. A more troublesome finding was the wide individual variations found in CO2 outputs and the practical difficulties in its compensation. Failure to correct for individual CO2 outputs can introduce considerable errors into the triolein breath test. Coupled with an inability of the triolein breath test to predict the severity of malabsorption and the relatively high frequency of disorders where its use is inappropriate, we now have serious misgivings as to its cole in the investigation of malabsorption.
References i NewcomerAD, Hofmann AF, DiMagno El>,Thomas PJ, Carlson GL. Triolein breath test. A sensitive and specific test for fat malabsorption. Gastroenterology 1979;76:6-13. 2 WestPS, Levin GE, Griffin GE, Maxwell JD. Comparison of simple screeningtests for fat malabsorption. Br Med J 1981;282:1501-1504. 3 EinarssonK, Bjorkhem i, Eklof R, Blomstrand R. t4C-Triolein breath test as a rapid and convenient screening test for fat malabsorption. Scand J Gastroenterol 1983;18:9-12. 4 AkessonB, Floren C-H. Use of the triolein breath test for the demonstration of fat malabsorption in coeliac disease. Scand J Gastroenterol 1984;19:307-314. 5 ButlerRN, lawson MJ, Gehling NJ, Grant AK. Clinical evaluation of the t4C-trioleinbreath test: a critical analysis. Aust N Z J Med 1984;14:ii I-! 13. 6 Turner JM, lawrence S, Fellows IW, Johnson !, Hill PG, Holmes GKT. t4C-Trioleinabsorption: a useful test in the diagnosis of malabsorption. Gut 1987;28:694-700. 7 lknino L, Scuro LA, Menini E, et al., is the t4C-trioleintest useful in the assessmentof malabsorption in clinical practice? Digestion 1984:29:91-97. 8 PedersenNT, Andersen BN, Marqversen J. Estimation of [t4Cltriolein assimilationas a test of lipid assimilation. Breath test or measurement of serum radioactivity. Scand J Gastroenterol 1982,17:309-316. 9 Luey K, Pattinson NR, Hinton D, Cook HB, Campbell CB. Comparison of three methods to estimate steatorrhoea. N Z Med J 1981;93:36-38. 10 MylvaganumK, Hudson PR, Ross A, Williams CP. 14C triolein breath test: a routine test in the gastroenterology clinic? Gut 1986;27:1347-!352. II RosenbergIH, Sitrin MD. Screening for fat malabsorption. Ann Intern Med 1981;95:776-777. 12 Nelson LM, McKenzie.IF, Russell RI. Measurement of fat absorption using [ 3H]glycerol tricther and [taCiglycerol trioleate in man. Clin Chim Acta 1980;103:325-334 13 PedersenNT. Estimatiov.of lipid assimilation from faecal samples using [14C]trioleinas tracer and [51Crlchromium chloride as non-absorbable marker. Scand J Clin Lab Invest 1983;43:323-328.
64 14 Fromm H, Hofmann AF. Breath tests for altered bile-acid metabolism. Lancet 1971;ii:621-625. 15 Kachroo RL, Rehani MM, Kachroo N, Khuroo MS, Zarger SA. Optimising the value of C-14 breath test. Indian J Mud Res 1987;86:! 19-123. 16 Cotes JE. Lung Function. Assessment and application in medicine. 4th Edn. Oxford: Blackwell Scientific Publications, 1979. 17 Van de Kamer JH, Huinink HTB, Weijers HA. Rapid method for the determination of fat in faeces. J Bioi Chum 1949;177:347-355. 18 Caspary WF. Breath tests. Clin Gastroenterology 1978;7:351-374. 19 Tomkin GH, Bell TK, Hadden DR. Evaluation of malabsorption test using t4C-triglyceride, lr J Mud Sci 1971;140:449-454. 20 McCailum RW. Motor function of the stomach in health and disease. In: MH Sleisenger and JS Fordtran eds. Gastrointestinal disease m pathophysiology, diagnosis, management. 4th edn. Philadelphia: W$ Saunders Co. 1989;675-713. 21 Lossow WJ, Chaikoff I L. Carbohydrate sparing of fatty acid oxidation. ! The relation of fatty acid chain length to the degree of sparing. 11 The mechanism by which carbohydrate spares. Arch Biochem 1955;57:23-38. 22 Bragdon JH. 14CO2excretion after the intravenous administration of labeled chylomicrons in the rat. Arch Biochem 1958:75:528-539. 23 Winchell HS, Stahelin H, Kusubov N e t al. Kinetics of CO2-HCO3 in normal adnlt males. J Nucl Med 1970;I 1:711-715. 24 King CE, Toskes PP. Alteration of carbon dioxide production during nonfasting isotopic CO 2 breath tests: concise communication. J Nucl Mud 1981;22:955-958.