- Email: [email protected]
Serum cholesterol response to changes in the diet
Metabolism Clinical and Experimental VOL. XIV, NO. 7
JULY, 1965
Serum Cholesterol Response to Changes in the Diet I. Iodine Value of Dietary Fat versus 2S-P By ANCEL KEYS, JOSEPH T. ANDERSON AND FRANCISC~ GRANDE In men in calorie equilibrium, changes in dietary fat produce responses in the serum cholesterol level that, on the average, are predictable from the percentages of total calories provided by saturated (S) and polyunsaturated (P) fatty acid glycerides in the diets concerned. S and P have opposing actions and, in general, A Cholesterol (mg./lOO ml.) = 2.7AS - I.3AP, where A refers to the difference between 2 diets. Increasing the number of double bonds beyond 2 in polyunsaturated fatty acids does not result in proportionate increases in serum cholesterol-lowering effect. The mono-enes oleic and erucic acid have little or no effect on the serum
cholesterol level when they are ex. changed in the diet for equal calories of simple carbohydrate. Changes of fats in the diet produce serum cholesterol responses that are also correlated with the iodine values, or the square-roots of those values, of the fats concerned when the iodine value happens to be highly correlated with 2.7s - 1.3P. When changes in diet fats involve substantial differences in amounts of mono-enes, or of fatty acids containing more than 2 double bonds, the serum cholesterol response has a low or negligible correlation with the iodine value, or its squareroot, of the fats.
T
HE AMOUNT AND KIND of fats in the diet influence
the concentration of cholesterol in the blood of man and, though the mechanisms remain obscure, it is important to have empirical guidance as to what may be expected from given fat changes in the diet. In this regard, attention has been directed towards several methods of characterizing the diet: (1) the iodine value of the fats in it, (2) the proportion of the total metabolism covered by specified types of fatty acids, and (3) the P/S ratio, i.e., the ratio of polyunsaturated to saturated fatty acids. The use of the P/S ratio in this connection has been properly condemned.l-s As a means of characterizing fats in diets, the P/S ratio should be abandoned promptly because, without other information, it has little significance; it is erroneous to imply that this ratio, by itself, characterizes a fat properly or is From Laboratory of Physiological Hygiene, University of Minnesota, Minneapolis. Aided by research gTUTttS from the U. S. Public Health Service (no. HE-04997 K, no. HE-04401 to F. C. and 3. T. A.) Received for publication Feb. 19, 1965.
to A.
747
METABOLISM, VOL. 14,No. 7 (JULY),1965
748 a measure cholesterol.
KEYS, ANDERSON AND GRANDE
of the characteristic
of the diet in regard
to effect
on blood
Ahrens et ah4 proposed that when the total fat in the diet is constant the response of the serum-cholesterol level in man to fat in the diet tends to be proportional to the iodine value of the fat in it, i.c., to the average net urrsatu ration of the fatty acids. On the other hand, wc proposed that the several kinds of fatty acids may exert independent effects that are not necessarily a simple direct function of the iodine value and we emphasized the importance of the proportion of total metabolism supported by dietarv fat.” The fat in human diets consists of a mixture of triglycerides of many different fatty acids but, as a first approximation, we suggested that these be grouped into 3 classes: saturated, mono-ene, and polyunsaturated fattv acids. The amounts of the glycerides of these fattv acids. expressed as percentages of total dietary calories, are designated as S, M. and P, respectivelv, and. if Y is the serum cholesterol concentration, it is useful to csamine the model: (1)
Y =
n + bS + c’\I + dP,
in which a is the sum of all nonfat influences on the serum cholesterol level and b, c, and d are unknown but discoverable coefficients. Obviously, if manv men were studied on manv diets differing in the values of S, M, and P, but with all other conditions constant, it would be possible to obtain, from leastsquares solution of the multiple regression ecluation, the vahres of b, c, and d, so as to provide a formula to predict the average effect on the serumcholesterol level of changing the diet: (3) (2a)
Y,
-
Y,
=
h(S, - S,) + c(hI, - .\I,) + tl(dP, AY = hAS + CARI t tlAP.
-
l’,).
or,
For this purpose 41 sets of controlled experiments were carried out with groups of 12 to 27 men in each experiment. 111 each experiment one group of men changed from one to another specified diet while a matched group of their fellows made the reverse dietary change, the dietary periods being 3 to 4 or more weeks each.” The mathematical analysis of the results yielded. for the coefficients in equations (1) and (2) above, 1, = 2.74. c = 0, d = - 1.31, and equation (2a) becomes, roughly. (2b)
AY = 2.7~s
-
1.3nP
or. AU =
1.35
(2aS
-
AI’).
This analysis indicated opposing actions of saturated and poly-ene fatty acids, the effect of a given amount of saturated fatty acid being about twice that of poly-ene fatty acids, while mono-ene‘, or more specifically oleic acid. may be ignored. Tests of the 2S-P Formulation The fact that the controlling variable in most natural or experimental human diets is very close to 2S-P, was later shown in four ways. First, independent critical experiments showed that, indeed, natural (cis) mono-ene in the human diet has little or no effect on serum cholesterol; this was found both with oleic acid,” and with erucic acid, the mon-ene that dominates rapeseed
SERUM
CHOLESTEROL
RESPONSE
TO CHANGES
IN WET
749
0i1.~ Second, in the diet, the highly unsaturated fatty acids in fish oils were found to be similar in effect on serum cholesterol to equal amounts of the di-ene, linoleic acid, which acccunts for almost all of P in all normal nonfish diets.*Bg Third, another series of 18 sets of controlled experiments with 12 to 22 men in each setlo gave almost identical results in a new multiple regression analysis in which, again, mono-ene had no significant effect and the characteristic of the dietary fat that affects the cholesterol level proved to be not significantly different from that previously observed, i.e., a function of 2.7S1.3P or, roughly, 1.35( 2S-P). Test with the results of controlled experiments in other laboratories is the fourth, and critical, test. Ahrens et a1.i’ agreed that their own data fitted closely to our prediction but argued that “2.74( S)-1.31 (L) . , is synonymous to iodine value.“1* Apart from the fact that their term L (for linoleic acid) is more restricted than our term P (for total poly-ene), their argument is in error. They point out, correctly, that if icdine value = V and if s, m, and p are, respectively, percentages of the fat represented by glycerides of saturated, mono-ene and di-ene, then V = 90 - 0.9( s-p). But s-p # 2s-p, let alone 2S-P, and the distinction is critical; if s-p were the controlling variable, then equal amounts of saturated and polyunsaturated fatty acids would have opposite but equal effects. An example may help clarify the distinction. Consider 2 diets, both providing 40 per cent of calories from fats, but the fat in diet A is made up of equal parts of palmitoleic and oleic acids while in diet B there are equal parts of palmitic and linoleic acids. In both diets the iodine value is identical but ecluation (2b) predicts a rise of about 27 mg. of cholesterol/100 ml. of serum if the diet is changed from A to B. The expression 2S-P = F(2s-p), where F is the proportion of total dietary calories provided by fat and s and p are percentages of fatty acid glycerides in the fat. For many mixed food fats there is a high correlation between iodine value and 2s-p but there is not necessarily any correlation between iodine value and F; when total fat is a constant percentage of total dietary calories, 2s-p can be used for comparisons; this is not possible unless F, = FZ. The experiments of Malmros and Wigand12 nicely test the validity of 2S-P though they underestimate the effect of changes in dietary fats; dietary periods of I to 2 weeks are less than the minimum length needed to approach the ultimate effect on serum cholesterol. Figure 1 summarizes the results with their data; observed cholesterol change was related to n (2S-P) with r = 0.96, though only “book” values were available for fatty acid composition of the fats fed. The solid circles are from comparisons involving rapeseed oil as the fat in one of the 2 diets compared; the circles with crosses similarly involve olive oil. These points from diets containing fats dominated by the mono-enes erucic and oleic acid (rapeseed and olive oil, respectively), correspond to the general correlation and they indicate that, as concluded earlier, natural monoenes can be neglected in estimating effects of diet fat on serum cholesterol. Turpeinen et al. I3 studied the response of patients in mental hospitals to a measured change in the diet. For 59 patients whose average serum cholesterol
750
KEYS, ANDERSON
0
10
20
30
40
50
60
70
00
AND GRANDE
5
A(2S-P)
Fig. l.-Relationship between observed effect on serum cholesterol of a change in fats in the diet and the dietary change expressed in terms of percentages of calories from saturated (S) and polyunsaturated (P) fatty acids. Average values of
groups of persons studied by Malmros and Wigand. was 268.7 mg./lOO ml., initially, the average observed value after dietary change was 243.7; using our detailed formulationlo they predicted a value of 241.4. For 58 patients with an initial average of 202.1, the subsequent observation was 187.3 and the prediction was 188.7. Data and Analysis of Gunning et al. Recently, Gunning et al., R from data on 4 patients, stated: “a high index of correlation (r = 0.9) was found to exist between the square-root of the iodine number of the dietary fat and the plasma cholesterol level,” and, further, “there is a rough ‘fit’ if one uses the formula suggested by Keys (but) it is not as good as the square-root of the iodine number. ” Both of the above statements are in error. In the first place, actual calculation shows that the correlation coefficient is not r = 0.9; it is about r = -0.82. Gunning et al. list different values for the iodine value of one fat (avocado oil) in their tables 1 and 2; with the value in table 1, r = -0.818; from table 2, r = -0.829. In the second place, it seems that Gunning et al. made no calculations with “the formula suggested by Keys.” Table 1 summarizes the data on the 10 fats used by Gunning et al.,” including the values for 2s-p, where s and p refer to percentages of the fat made up of saturated and polyunsaturated fatty acids. Calculation shows that the coefficient of correlation between 2s-p and the square-root of the iodine value is r = -0.98 or -0.99, depending on which value is taken for avocado oil. Obviously there can be no significant difference between these 2 measures of dietary fat in regard to the relationship to serum cholesterol; actual calculation shows that both the (negative) square-root of the iodine value and 2s-p are correlated with the serum cholesterol with r = 0.8. Figure 2 shows the relationships of both fat variables with the serum cholesterol concentration. It should
751
SERUM CHOLESTEXiOL RESPONSE TO CHANGES IN DIET
Table l.--Fats
and Oils Used by Gunning et aL3
Iodine Fat
Value
(I.V.)“2
2s-_p
Iodine Value
Fat
(I.V.)‘/?
2s-p
Avocado Beef Butter
90 40 27
9.5 6.4 5.2
16 114 125
Corn Cottonseed Mix UEF 178
115 106 90
10.7 10.3 9.5
-30
Chicken Coconut
81 13
9.0 3.6
41 176
Safflower Soybean
142 119
11.9 10.9
-51 -26
1 6
Table 2 1)
2)
3) 4) % Diet Calories
No.of (I.V.)‘/Z
:
5)
6)
7) Chol.. mx.%
Expt.
Men
ABMF
23
53
7.3
34
17
3
31
214 C 6.4”
ABLF
23
55
7.4
13
6
1
11
185 2
5.7
ABC
23
112
10.6
44
10
20
0
158 f
5.4
ACMF
24
54
7.3
35
17
3
31
TotalFat
S
P
(2S-P)
ACLF
24
60
7.7
12
5
1
9
ADM
23
83
9.1
41
11
9
13
213 4 6.4 180 *
5.9
188 -c 6.8
ADB
23
43
6.6
41
24
3
45
ADRO
23
96
9.8
41
7
9
5
187 2 7.1
AGMO
23
140
11.8
42
15
11
19
199 ti 7.3
AGVO
23
77
8.7
42
16
11
21
195 -c 5.5
PLF
16
59
7.7
9
4
1
I
168 -c 5.0
233 i
9.1
PO0
16
80
8.9
36
8
4
12
182 2 5.3
PSO
16
69
8.3
18
8
4
12
182 & 5.8
Summary glycerides the
I.V.
in -_
serum.
lS.E.
of of
13
experiments
saturated
“I.V.”
=
and iodine
on
physically
polyunsaturated value
of total
healthy fatty
diet
men
acids
in
a
locked
respectively.
metabolic “Chol.”
=
unit.
S
P
are
total
cholesterol
and
in
fat.
of mean.
be noted that with this material the iodine value itself is just as good (r = -0.8) as the square root of that value. The data of Gunning et al. are, in fact, grossly inadequate for a critical analysis of these questions. Their data cover only 17 patient-diet trials and for one patient the data refer to a single diet so no calculations can be made about response to diet fat. Among the other 16 trials, in 7 cases dietary cholesterol was also a variable though it was unmeasured and ignored. There remain a total of 9 diet trials on 2 patients, one of whom (“controlled hypothyroidism”) could hardly be termed metabolically “normal.” But apart from these limitations, the fats selected do not include the kinds that differentiate sharply between iodine value and S-p; for that purpose it is desirable to include fats extremely high in mono-ene, such as olive oil and rapeseed oil, and fats such as fish oils which contain fatty acids with many double bonds. Gunning et ah3 ignored differences in the total fat in the diet which, in effect, implies that high-fat and low-fat diets are equivalent so long as iodine value of the separated fat is constant. However, a more reasonable relationship may be postulated: if Y is serum cholesterol, mg./106 ml., F is percentage of total calories provided by total fats, Q is the iodine value (or its square-root), and a and b are constants,
752
KEYS, ANDERSON AND GRANDE 34c
,- 0
320
0
t 8 3oc
,.
& CI
0 0 200
I =u.A 0
0 180
1
L
12
11
IO
9
(IODINE
8 VALUE I
7
6
5
3
3
k!
Fig. Za.-Relationship between observed serum cholesterol and the square-root of the iodine value of the dietary fat. Data from Gunnjng et al.:’ 340 0
320
300 2 E 8
280
z ”
260
0 0 0
? 0
0 0 0 0
r=D.8
0 0 J
180 -40
0
40
80
120
160
2co
29P
Fig. 2b.-Same data as in figure 2a, but the dietary fat characterized by 2s-p, where s and p are percentages of the fat accounted for by glycerides of saturated (s) and polyunsaturated (p) fatty acids. Y = a + bFQ, and AY = b(F<,Q., _ - - FIQ,).
(3) (3a) Later,
we shall examine
this formulation
with more adequate
tlata.
METHODS Below we present some data subjects were physically healthy
that will be reported in greater detail elsewhere. adult male mental patients in a locked metabolic
The unit
SERUM
CHOLESTEROL
with special
kitchen
RESPONSE TO CHANGES
and dining
facilities.
eaten under supervision. The methodology was identical Each
series
of experiments
used
753
IN DIET’
These
men received
no other food besides
with that used in many studies previously a low-fat
basal
diet
of rotating
menus
that
reported.6s14J5 to which
was
added the experimental fat or oil, purchased in large quantity and stored at -20 C. so as to have unchanged materials throughout the series of experiments. All foods served to each man at each meal were measured and plate waste was estimated. Besides the analytical data on the test fats and oils and the data computed from tables of composition for the foods in composite as fed, matter in the lipid Each series of individually
the basal diet, chemical analyses were periodically made on the diet these analyses covering water, ash, protein, lipid extract, nonsaponifiable extract, and fatty acids, by gas liquid chromatography, in the glycerides. experiments was preceded by prolonged control on. a standardized and per
cent
of calories from total fats, 18 and 4 per cent being from glycerides of saturated polyunsaturated fatty acids, respectively. The caloric metabolic balance point for
measured
version
of the regular
hospital
diet
which
averaged
and each
man was estimated and his individual ration level was set accordingly; thereafter tended to gain or lose weight his diet was adjusted weekly. All calorie adjustments
if he were
made by changing the amounts of fat-free bread and jelly in the rations. came ill or changed in weight more than 3 Kg. in an B-weeks experiment from the analysis of results, Venous
40
Men who bewere dropped
blood was drawn from each man on 2 or more days at the close of each dietary
period. The separated serum was analyzed in duplicate portions our standardized version of the method of Abel1 et al.16 RESULTS
OF CRITICAL
for total
cholesterol
by
EXPERIMENTS
Table 2 summarizes the results from 5 sets of experiments with diets designed to allow critical analysis of relationships among fats in the diet and serum cholesterol. In each series, subgroups of the men, matched in serum cholesterol level at the end of a preceding period on a standard control diet, subsisted on the experimental diets in different sequences. The experimental diets consisted of a basic low-fat diet of fixed composition covering 55 to 60 per cent of calorie needs to which was added the experimental fat and/or simple carbohydrate (bread made without fat plus sugar as fruit-flavored jelly) to bring total calories to energy balance. In the AB series the iodine value of the fat was almost identical in 2 experiments but there was almost a threefold difference in 2S-P. The third diet had an iodine value double that of the 2 diets while the value of 2%P was practically zero. The serum cholesterol values proved to be parallel to the values of 2S-P but not to iodine value, square-root of iodine value or total fat in the diets. The AC series consisted primarily of a comparison of a low-fat and a moderately high-fat diet, the iodine values and fatty acid compositions in the 2 diets being similar. Serum cholesterol values mirrored 2S-P and total fat, In the AD series, fat provided 41 per cent of the calories in all 3 experiments but 2S-P covered a wide range. The experimental fat in experiment ADRO was rapeseed oil which is remarkable because the mono-ene erucic acid makes up at least half of the total fatty acid. But the results show no great difference from what would be expected from an equal amount of oleic acid, the usual mono-ene in our diets, or equal calories as carbohydrate. The AG series provided a particularly critical test because the 2 diets, both
754
PEYS,
ANDERSON
AND
GRANDE
providing 42 per cent of calories from total fats, were closely similar in 2S-P but greatly different in iodine value .g In experiment AGMO the experimental fat was menhaden oil, specially prepared and supplied to us by Technological Laboratory (Seattle) of the Bureau of Commercial Fisheries, LJ. S. Department of the Interior. The serum cholesterol values indicate, as concluded earlier, that in effect on serum cholesterol of man, the poly-enes in fish fat are similar to linoleic acid.“BlO The P series was primarily a test of the theory that oleic acid is substantially neutral in regard to serum cholesterol and that, in effect, a mixture of equal parts of saturated fatty acid and linoleic acid is not the equivalent of oleic acid. The results are in full accord; the olive oil diet (POO) provided 24 per cent of total calories from oleic acid but it produced exactly the same serum cholesterol as the diet (PSO) m . which vegetable oils low in oleic acid were mixed to give the same value of 2S-P but only one-fourth as much oleic acid (6 per cent of calories) in the diet. The data in table 2 may be used to make a new least-squares solution for equation ( 1 ), above. The result is, again, that mono-ene fails to have a coefficient significantly different from zero and the answer is: (Ia)
Y = 166.2+ 3.07s - 1.76P.
The coefficient of correlation between observed serum cholesterol and the value predicted by (la) 1s . r = 0.954. But if the coefficients 2.7 and -1.3, previously obtained, are used instead of 3.07 and -1.76, the predicted values are not significantly different. Accordingly, there is no reason to change the previous coefficients and we prefer (lb)
? x 167.8+ 2.7s -1.:3P.
Figure 3 shows the agreement between the mean observed cholesterol values and those predicted from equation ( lb ) . Th e 1owest point (circle) in figure 3 represents experiment ABC in which corn oil provided 30 per cent of total calories. As reported previously,‘“.” corn oil consistently gives slightly lower serum cholesterol values than predicted from the fatty acid composition alone, presumably because it contains much more plant sterol than the other fats generally used. When corn oil provides 30 to 40 per cent of total calories, an extra cholesterol depression of about 6-12 mg./lOO ml. is the usual finding. The cross (x) in figure 3 shows where the point for Experiment ABC should be if this allowance is made. In experiments summarized in table 2 and figure 3, corn oil was included in large amounts only in experiment ABC. The analysis resulting in equation ( lb), and the predicted values in figure 3, may not give the best possible correlation because it is implied that, on the average, the men in all 5 sets of experiments are similar, as groups, in their intrinsic cholesterol characteristic; i.e., on a given diet, the average serum values would be the same for all 5 groups of men. Actually, though many of the subjects participated in several sets of experiments and a few in all of them, it cannot be proved that the S groups of men are all random samples from a common population.
SERUM
CHOLFZSTEROL
RESPONSE
TO CHANGES
IN DIET
755
240
230 E 0 0 22Q& a
9; 167.0 + 2.75 -1.3 P
Fig. 3.-Observed serum cholesterol versus the value predicted tion Y = 167.8 + 2.7%1.3P. Data from table 2.
from the equa-
However, with no such assumption, the analysis can be made equally well in terms of differences between experiments within sets; that was the method used in previous analyses where the samples of men were not always so well matched.5J0 There are, then, from table 2, materials for 11 sets of comparisons: ABMF vs. ABLF, ABMF vs. ABC, ABLF vs. ABC, ACMF vs. ACLF, etc. When these differences, AY, are predicted from 2.7AS-MAP, the result closely agrees with the observed values; r = 0.95. The data of table 2, provide materials for critical testing of the hypothesis that the square-root of the iodine value is a reliable predictor of the serum cholesterol value. As indicated above, it is na’ive to neglect the proportion ,of total calories provided by total fat in the diet. If the effect of a dietary fat is directly related to the iodine value of that fat, the more of that fat there is in the diet, the greater will be its effect. Accordingly, equation3 above can be examined with the data from table 2, comparing the average serum cholesterol level on a given diet with the product of Q, square-root of the iodine value of the diet fat, F, the percentage of the total calories provided by that fat. Figure 4 summarizes the results. There is no significant correlation (r = O-10), and the postulate must be rejected. However, as noted above, the serum cholesterol level does tend to be correlated with the iodine value, or its square-root, when total dietary fat is constant, provided that the fats contain no significant amount of fatty acids more unsaturated than linoleic acid, and the proportion of mono-ene does not vary widely.
756
KEYS,
ANDERSON
AND
GHANDE
0
0
00
0 0
180
0
0
0 160.
15OL 0
100
200
300
400
500
E 10
FC W”2
Fig. 4.-Observed Data from table 2.
serum cholesterol
versus the square-root
of the iodine value.
DISCUSSION
The question as to the effects of mono-enes are crucial to consideration of 3S-P versus the iodine value, or its square-root, in the prediction of serum cholesterol response to change of dietary fat. Previous data, as well as those presented here, make it evident that in this respect oleic acid has little or no effect different from equal calories of carbohydrate in the diet. The case for erucic acid being similar is supported by a critical experiment reported her?, but there are also confirmatory data in the literature. We noted earlier the results of the experiment of Malmros and Wigand’” with rapeseed oil which did not appear to depart from expectations if its large content of mono-ene (erucic acid) is ignored. Linko’g has also provided rAvidence on this point from 15 patients who consumed 100 Gm. of rapeseed oil daily for 14 days following a control period. Cholesterol values varied in the serum of those patients but on the average there was a decIine of only about 5 mg. of cholesterol/100 ml. This may seem remarkable if attention is focussed on the iodine value of 104 of the rapeseed oil. But 100 Gm. of rapeseed oil containing 21 Gm. of polyunsaturated and 7 Gm. of saturated fatty acids means that at a total calorie level of 2400 the estimate of 2.7S-1.3P is about -4 mg./lOO ml. for expected change. The diet is certainly not the only factor that influences the serum cholesterol level in man and in the diet there are other variables to consider besides the percentages of total calories provided by the glycerides of the 2 broad classes of saturated and of polyunsaturated fatty acids. However, for direct relevance to free-living men, it is useful to focus on the nutrients, and their quantitative
SERUM
CHOLEXI’JlROL
RESPONSE
TO CHANGES
IN DIET
757
variations, as they occur in natural, spontaneously chosen human diets. Highly abnormal diets can be artificially concocted and fed in special experiments; the interesting results in these situations, and from studies on experimental animals, must not be confused with the situation in long-time real-life subsistence of man. Gross deficiencies of some nutrients may produce some special effects; here we are concerned only with the state of caloric balance and the absence of gross deficiencies of proteins, vitamins and minerals, as well as of special metabolic disorders, and diets in which total fats provide from about 5 to 55 per cent of total calories. In such diets, which surely must cover the range of almost all natural variation, at least in the so-called Western World, besides S and P, on which we have focussed so far, it is necessary to consider several other items. Plant sterols are ubiquitous but they are ingested in only minute amounts except in special experiments, and by the few persons who deliberately incorporate great amounts of some oils, notably corn oil, in the diet. Beveridge et al.1° estimate that around 300 mg. of sitosterol/lOOO calories will produce a statistically significant decrease in serum cholesterol and that a diet providing 15 per cent or more of calories in the form of corn oil could have this effect. Most other vegetable food oils are much lower in sitosterol content. We have reported previously that with experimental diets containing much corn oil it is necessary to allow for this special effect. I7 For each 10 per cent of calories supplies by corn oil, perhaps it is reasonable to estimate an average fall of about 2 to 3 mg. of cholesterol/100 ml. of serum independent of the effect of the fatty acids in the corn oil. Other dietary items to be considered include exogenous cholesterol and the effects of particular saturated fatty acids; these will be discussed separately in later parts of this report. Finally, there is the important fact that people differ intrinsically in their cholesterol metabolism, both as shown by their serum levels on a given diet and in their response to a given dietary change. We have examined this matter previously, I” but propose to discuss it further in Part III of this communication. ACKNOWLEDGMENTS Many persons aided the work reported here but special thanks are due to the volunteer aides from the Church of the Brethren, who literally lived with the subjects at the Hastings State Hospital and more recently at the Faribault State School and Hospital, and to the staff members of those hospitals, particularly Mrs. Helen Williams. Mrs. Mary Jane Schweich and Miss Judy Erickson. For the dietary analyses we are indebted main11 to Miss hlargaret Neibling, Dr. Paul Kupchs and Mrs. Ethel Owens Sowl. Blood analyses in recent years were mainly done by Mrs. Gayle Frohlich Olson and Mrs. S. Nishimoto, among others. Mrs. Nedra Foster did much work on the dietary designs as well as on supervision of the chemical work. Willis Parlin, Norris Schulz and h4iss Joan La Riviere greatly aided the statistical analyses.
REFERENCES 1. Kinsell, L. W.: Some thoughts regarding the P:S ratio concept. Amer. J. Clin. Nutr. 12:228, 1963. 2. Grande, F.: Prediction of serum choles-
terol changes produced by dietary fats. Amer. J. Clin. Nutr. 13:122, 1963. 3. Gunning, B., Imaichi, K.. Splitter, S. D.,
758
KEYS,
and Kinsell. L. W.: Effects of different dietary fats on plasma-lipid levels. Lancet 2:336, 1984. 4. Ahrens. E. H.. Jr,. Hirsch. J.. Insull. W.. Jr.. Tsaltas, T. T.. Blomstrand. R.. and Peterson. h4. L.: The influence of dietary fats on serum-lipid levels in man. Lancet 1:943. 1957. 5. Keys. A., Anderson. J. T.. and Grande. F.: Prediction of serum chole: terol responses of 111an to changes in fats in the diet. 1 ,ancet 2:939, 1957. 6. -,
-, and -: Efl’.sc t on serum cholxterol in man oE mono-ene fatty acid (oleic acid) in the diet. Proc. Sot. Exp. Biol. Med. 98:387. 1958. 7. Grande. F.. Mataumoto. Y.. Anderson. J. T.. and Keys. A.: Effect of dietary rapeseed oil on man’s serum lipids. Circnlalion 26:653, lS62. 8. Ahrens. E. H.. Jr.. 11~~11. W.. Hirsch. J.. Stoffel. W.. Peterson. L., Farquhar. J. W.. >4iller, T.. Thomasson. H. J.: The effects human sermn-lipids of a dietary highly unsaturated. hut poor in
Jr,. hl. and on fat. es-
sential fatty acitls. Lancet 1: 115, 1959. 9. Grande, F., Ancler:.on. J. T.. and Keys, A.: A comparison of the effect of fish oil polyunsaturated fatty acids and of linoleic acid on man’s sermll lipids. Amer. J. Clin. Nutr. 12:331. 1963. 10. Keys. -4.. Anderson. J. T.. and Grand?, F.: Serunl cholesterol in man: diet fat and intrinsic responsiveness. CirIl.
culation 19:201. 1959. Ahrens. E. H.. JI-._ Hirsch. J.. Insull. W.. Jr.. and Peterson. M.: Chemistry of Lipides as Related to Atherosclero-
ANDERSON
AND
GHANDE
sis. Ed. by I. H. Page. Springfield Ill.. Charles C Thomas. 1938. p, 222. 12. Slalmros. A.. and Wigand. G.: The effect on serum-cholesterol of diets containing different fats. Lancet 2: 1 1957. 1:3. Turpeinen. 0.. Roine. I’.. Pekkarincn. M., Karvonen. M. J.. Rarttanen. Y.. Runeberg. J.. and Alavirta. I’.: Efi’cct of serum cholesterol level of reglac~nlcnt of dietary milk fat by soybra oil. Lancet 1: 196. 1960. 1-I. Anderson. J. T., (:rande. F., Chlouverakis. C.. Proja. ?\I.. and Keys. .4.: Effect of dietary cholesterol on scr,mI cholesterol level in Nan. Fed. Proc. 21: 100. 1962. -. Mats~mioto. Y.. and Keys. A.: Glucose. sucro:x and lactose in tlrc: diet and blood lipid in man. J. Nlltr. 79: 349. 19ci:3. 16. -. ant1 Keys. A.: Cholesterol in serlun alld lipoprotein fractions: Its measurement and stability. Clin. Chem. 2:
15. -_
145, 1956. 17. (Zrande. F.. Anderson.
J,
T.. and Ke1.s.
A.: Serunl cholesterol in man ;kntl unsaponifiahle fraction of corn oil in the diet. Proc. Sot. Exp. Biol. Med. 98:436, 1958. 18. Linko. E.: Vegetable oils and sermil experiments cholesterol. Short-term with rapeseed and sunflower seed Acta med. Stand. 159:475. oils. 1957. 19. Beveridge, J. M. R.. Haust. II. Connell. W. F.: Magnitude hypocholesterolemic effect of sitosterol in man. J. Nlltr. 1964.
Ancel K~Js, Ph.D., Professor, School of Public Health, Director, Laboratory of Physiological Hygiene, Vniversity of Minnesota, Mi.nneapolis, Minn. Joseph T. Anderson, Ph.D., Professor, School of Public Health, University of Minnesota, Minneapolis, Minn. Francisco Grade, M.D., Professor, School of Public Health, University of Minnesota, Director, Jay Phillips Research Laboratory, Mount Sinai Hospital, Blinncapolis, Minn.
L.. and of the dietary 8:): 119.
JULY, 1965
Serum Cholesterol Response to Changes in the Diet I. Iodine Value of Dietary Fat versus 2S-P By ANCEL KEYS, JOSEPH T. ANDERSON AND FRANCISC~ GRANDE In men in calorie equilibrium, changes in dietary fat produce responses in the serum cholesterol level that, on the average, are predictable from the percentages of total calories provided by saturated (S) and polyunsaturated (P) fatty acid glycerides in the diets concerned. S and P have opposing actions and, in general, A Cholesterol (mg./lOO ml.) = 2.7AS - I.3AP, where A refers to the difference between 2 diets. Increasing the number of double bonds beyond 2 in polyunsaturated fatty acids does not result in proportionate increases in serum cholesterol-lowering effect. The mono-enes oleic and erucic acid have little or no effect on the serum
cholesterol level when they are ex. changed in the diet for equal calories of simple carbohydrate. Changes of fats in the diet produce serum cholesterol responses that are also correlated with the iodine values, or the square-roots of those values, of the fats concerned when the iodine value happens to be highly correlated with 2.7s - 1.3P. When changes in diet fats involve substantial differences in amounts of mono-enes, or of fatty acids containing more than 2 double bonds, the serum cholesterol response has a low or negligible correlation with the iodine value, or its squareroot, of the fats.
T
HE AMOUNT AND KIND of fats in the diet influence
the concentration of cholesterol in the blood of man and, though the mechanisms remain obscure, it is important to have empirical guidance as to what may be expected from given fat changes in the diet. In this regard, attention has been directed towards several methods of characterizing the diet: (1) the iodine value of the fats in it, (2) the proportion of the total metabolism covered by specified types of fatty acids, and (3) the P/S ratio, i.e., the ratio of polyunsaturated to saturated fatty acids. The use of the P/S ratio in this connection has been properly condemned.l-s As a means of characterizing fats in diets, the P/S ratio should be abandoned promptly because, without other information, it has little significance; it is erroneous to imply that this ratio, by itself, characterizes a fat properly or is From Laboratory of Physiological Hygiene, University of Minnesota, Minneapolis. Aided by research gTUTttS from the U. S. Public Health Service (no. HE-04997 K, no. HE-04401 to F. C. and 3. T. A.) Received for publication Feb. 19, 1965.
to A.
747
METABOLISM, VOL. 14,No. 7 (JULY),1965
748 a measure cholesterol.
KEYS, ANDERSON AND GRANDE
of the characteristic
of the diet in regard
to effect
on blood
Ahrens et ah4 proposed that when the total fat in the diet is constant the response of the serum-cholesterol level in man to fat in the diet tends to be proportional to the iodine value of the fat in it, i.c., to the average net urrsatu ration of the fatty acids. On the other hand, wc proposed that the several kinds of fatty acids may exert independent effects that are not necessarily a simple direct function of the iodine value and we emphasized the importance of the proportion of total metabolism supported by dietarv fat.” The fat in human diets consists of a mixture of triglycerides of many different fatty acids but, as a first approximation, we suggested that these be grouped into 3 classes: saturated, mono-ene, and polyunsaturated fattv acids. The amounts of the glycerides of these fattv acids. expressed as percentages of total dietary calories, are designated as S, M. and P, respectivelv, and. if Y is the serum cholesterol concentration, it is useful to csamine the model: (1)
Y =
n + bS + c’\I + dP,
in which a is the sum of all nonfat influences on the serum cholesterol level and b, c, and d are unknown but discoverable coefficients. Obviously, if manv men were studied on manv diets differing in the values of S, M, and P, but with all other conditions constant, it would be possible to obtain, from leastsquares solution of the multiple regression ecluation, the vahres of b, c, and d, so as to provide a formula to predict the average effect on the serumcholesterol level of changing the diet: (3) (2a)
Y,
-
Y,
=
h(S, - S,) + c(hI, - .\I,) + tl(dP, AY = hAS + CARI t tlAP.
-
l’,).
or,
For this purpose 41 sets of controlled experiments were carried out with groups of 12 to 27 men in each experiment. 111 each experiment one group of men changed from one to another specified diet while a matched group of their fellows made the reverse dietary change, the dietary periods being 3 to 4 or more weeks each.” The mathematical analysis of the results yielded. for the coefficients in equations (1) and (2) above, 1, = 2.74. c = 0, d = - 1.31, and equation (2a) becomes, roughly. (2b)
AY = 2.7~s
-
1.3nP
or. AU =
1.35
(2aS
-
AI’).
This analysis indicated opposing actions of saturated and poly-ene fatty acids, the effect of a given amount of saturated fatty acid being about twice that of poly-ene fatty acids, while mono-ene‘, or more specifically oleic acid. may be ignored. Tests of the 2S-P Formulation The fact that the controlling variable in most natural or experimental human diets is very close to 2S-P, was later shown in four ways. First, independent critical experiments showed that, indeed, natural (cis) mono-ene in the human diet has little or no effect on serum cholesterol; this was found both with oleic acid,” and with erucic acid, the mon-ene that dominates rapeseed
SERUM
CHOLESTEROL
RESPONSE
TO CHANGES
IN WET
749
0i1.~ Second, in the diet, the highly unsaturated fatty acids in fish oils were found to be similar in effect on serum cholesterol to equal amounts of the di-ene, linoleic acid, which acccunts for almost all of P in all normal nonfish diets.*Bg Third, another series of 18 sets of controlled experiments with 12 to 22 men in each setlo gave almost identical results in a new multiple regression analysis in which, again, mono-ene had no significant effect and the characteristic of the dietary fat that affects the cholesterol level proved to be not significantly different from that previously observed, i.e., a function of 2.7S1.3P or, roughly, 1.35( 2S-P). Test with the results of controlled experiments in other laboratories is the fourth, and critical, test. Ahrens et a1.i’ agreed that their own data fitted closely to our prediction but argued that “2.74( S)-1.31 (L) . , is synonymous to iodine value.“1* Apart from the fact that their term L (for linoleic acid) is more restricted than our term P (for total poly-ene), their argument is in error. They point out, correctly, that if icdine value = V and if s, m, and p are, respectively, percentages of the fat represented by glycerides of saturated, mono-ene and di-ene, then V = 90 - 0.9( s-p). But s-p # 2s-p, let alone 2S-P, and the distinction is critical; if s-p were the controlling variable, then equal amounts of saturated and polyunsaturated fatty acids would have opposite but equal effects. An example may help clarify the distinction. Consider 2 diets, both providing 40 per cent of calories from fats, but the fat in diet A is made up of equal parts of palmitoleic and oleic acids while in diet B there are equal parts of palmitic and linoleic acids. In both diets the iodine value is identical but ecluation (2b) predicts a rise of about 27 mg. of cholesterol/100 ml. of serum if the diet is changed from A to B. The expression 2S-P = F(2s-p), where F is the proportion of total dietary calories provided by fat and s and p are percentages of fatty acid glycerides in the fat. For many mixed food fats there is a high correlation between iodine value and 2s-p but there is not necessarily any correlation between iodine value and F; when total fat is a constant percentage of total dietary calories, 2s-p can be used for comparisons; this is not possible unless F, = FZ. The experiments of Malmros and Wigand12 nicely test the validity of 2S-P though they underestimate the effect of changes in dietary fats; dietary periods of I to 2 weeks are less than the minimum length needed to approach the ultimate effect on serum cholesterol. Figure 1 summarizes the results with their data; observed cholesterol change was related to n (2S-P) with r = 0.96, though only “book” values were available for fatty acid composition of the fats fed. The solid circles are from comparisons involving rapeseed oil as the fat in one of the 2 diets compared; the circles with crosses similarly involve olive oil. These points from diets containing fats dominated by the mono-enes erucic and oleic acid (rapeseed and olive oil, respectively), correspond to the general correlation and they indicate that, as concluded earlier, natural monoenes can be neglected in estimating effects of diet fat on serum cholesterol. Turpeinen et al. I3 studied the response of patients in mental hospitals to a measured change in the diet. For 59 patients whose average serum cholesterol
750
KEYS, ANDERSON
0
10
20
30
40
50
60
70
00
AND GRANDE
5
A(2S-P)
Fig. l.-Relationship between observed effect on serum cholesterol of a change in fats in the diet and the dietary change expressed in terms of percentages of calories from saturated (S) and polyunsaturated (P) fatty acids. Average values of
groups of persons studied by Malmros and Wigand. was 268.7 mg./lOO ml., initially, the average observed value after dietary change was 243.7; using our detailed formulationlo they predicted a value of 241.4. For 58 patients with an initial average of 202.1, the subsequent observation was 187.3 and the prediction was 188.7. Data and Analysis of Gunning et al. Recently, Gunning et al., R from data on 4 patients, stated: “a high index of correlation (r = 0.9) was found to exist between the square-root of the iodine number of the dietary fat and the plasma cholesterol level,” and, further, “there is a rough ‘fit’ if one uses the formula suggested by Keys (but) it is not as good as the square-root of the iodine number. ” Both of the above statements are in error. In the first place, actual calculation shows that the correlation coefficient is not r = 0.9; it is about r = -0.82. Gunning et al. list different values for the iodine value of one fat (avocado oil) in their tables 1 and 2; with the value in table 1, r = -0.818; from table 2, r = -0.829. In the second place, it seems that Gunning et al. made no calculations with “the formula suggested by Keys.” Table 1 summarizes the data on the 10 fats used by Gunning et al.,” including the values for 2s-p, where s and p refer to percentages of the fat made up of saturated and polyunsaturated fatty acids. Calculation shows that the coefficient of correlation between 2s-p and the square-root of the iodine value is r = -0.98 or -0.99, depending on which value is taken for avocado oil. Obviously there can be no significant difference between these 2 measures of dietary fat in regard to the relationship to serum cholesterol; actual calculation shows that both the (negative) square-root of the iodine value and 2s-p are correlated with the serum cholesterol with r = 0.8. Figure 2 shows the relationships of both fat variables with the serum cholesterol concentration. It should
751
SERUM CHOLESTEXiOL RESPONSE TO CHANGES IN DIET
Table l.--Fats
and Oils Used by Gunning et aL3
Iodine Fat
Value
(I.V.)“2
2s-_p
Iodine Value
Fat
(I.V.)‘/?
2s-p
Avocado Beef Butter
90 40 27
9.5 6.4 5.2
16 114 125
Corn Cottonseed Mix UEF 178
115 106 90
10.7 10.3 9.5
-30
Chicken Coconut
81 13
9.0 3.6
41 176
Safflower Soybean
142 119
11.9 10.9
-51 -26
1 6
Table 2 1)
2)
3) 4) % Diet Calories
No.of (I.V.)‘/Z
:
5)
6)
7) Chol.. mx.%
Expt.
Men
ABMF
23
53
7.3
34
17
3
31
214 C 6.4”
ABLF
23
55
7.4
13
6
1
11
185 2
5.7
ABC
23
112
10.6
44
10
20
0
158 f
5.4
ACMF
24
54
7.3
35
17
3
31
TotalFat
S
P
(2S-P)
ACLF
24
60
7.7
12
5
1
9
ADM
23
83
9.1
41
11
9
13
213 4 6.4 180 *
5.9
188 -c 6.8
ADB
23
43
6.6
41
24
3
45
ADRO
23
96
9.8
41
7
9
5
187 2 7.1
AGMO
23
140
11.8
42
15
11
19
199 ti 7.3
AGVO
23
77
8.7
42
16
11
21
195 -c 5.5
PLF
16
59
7.7
9
4
1
I
168 -c 5.0
233 i
9.1
PO0
16
80
8.9
36
8
4
12
182 2 5.3
PSO
16
69
8.3
18
8
4
12
182 & 5.8
Summary glycerides the
I.V.
in -_
serum.
lS.E.
of of
13
experiments
saturated
“I.V.”
=
and iodine
on
physically
polyunsaturated value
of total
healthy fatty
diet
men
acids
in
a
locked
respectively.
metabolic “Chol.”
=
unit.
S
P
are
total
cholesterol
and
in
fat.
of mean.
be noted that with this material the iodine value itself is just as good (r = -0.8) as the square root of that value. The data of Gunning et al. are, in fact, grossly inadequate for a critical analysis of these questions. Their data cover only 17 patient-diet trials and for one patient the data refer to a single diet so no calculations can be made about response to diet fat. Among the other 16 trials, in 7 cases dietary cholesterol was also a variable though it was unmeasured and ignored. There remain a total of 9 diet trials on 2 patients, one of whom (“controlled hypothyroidism”) could hardly be termed metabolically “normal.” But apart from these limitations, the fats selected do not include the kinds that differentiate sharply between iodine value and S-p; for that purpose it is desirable to include fats extremely high in mono-ene, such as olive oil and rapeseed oil, and fats such as fish oils which contain fatty acids with many double bonds. Gunning et ah3 ignored differences in the total fat in the diet which, in effect, implies that high-fat and low-fat diets are equivalent so long as iodine value of the separated fat is constant. However, a more reasonable relationship may be postulated: if Y is serum cholesterol, mg./106 ml., F is percentage of total calories provided by total fats, Q is the iodine value (or its square-root), and a and b are constants,
752
KEYS, ANDERSON AND GRANDE 34c
,- 0
320
0
t 8 3oc
,.
& CI
0 0 200
I =u.A 0
0 180
1
L
12
11
IO
9
(IODINE
8 VALUE I
7
6
5
3
3
k!
Fig. Za.-Relationship between observed serum cholesterol and the square-root of the iodine value of the dietary fat. Data from Gunnjng et al.:’ 340 0
320
300 2 E 8
280
z ”
260
0 0 0
? 0
0 0 0 0
r=D.8
0 0 J
180 -40
0
40
80
120
160
2co
29P
Fig. 2b.-Same data as in figure 2a, but the dietary fat characterized by 2s-p, where s and p are percentages of the fat accounted for by glycerides of saturated (s) and polyunsaturated (p) fatty acids. Y = a + bFQ, and AY = b(F<,Q., _ - - FIQ,).
(3) (3a) Later,
we shall examine
this formulation
with more adequate
tlata.
METHODS Below we present some data subjects were physically healthy
that will be reported in greater detail elsewhere. adult male mental patients in a locked metabolic
The unit
SERUM
CHOLESTEROL
with special
kitchen
RESPONSE TO CHANGES
and dining
facilities.
eaten under supervision. The methodology was identical Each
series
of experiments
used
753
IN DIET’
These
men received
no other food besides
with that used in many studies previously a low-fat
basal
diet
of rotating
menus
that
reported.6s14J5 to which
was
added the experimental fat or oil, purchased in large quantity and stored at -20 C. so as to have unchanged materials throughout the series of experiments. All foods served to each man at each meal were measured and plate waste was estimated. Besides the analytical data on the test fats and oils and the data computed from tables of composition for the foods in composite as fed, matter in the lipid Each series of individually
the basal diet, chemical analyses were periodically made on the diet these analyses covering water, ash, protein, lipid extract, nonsaponifiable extract, and fatty acids, by gas liquid chromatography, in the glycerides. experiments was preceded by prolonged control on. a standardized and per
cent
of calories from total fats, 18 and 4 per cent being from glycerides of saturated polyunsaturated fatty acids, respectively. The caloric metabolic balance point for
measured
version
of the regular
hospital
diet
which
averaged
and each
man was estimated and his individual ration level was set accordingly; thereafter tended to gain or lose weight his diet was adjusted weekly. All calorie adjustments
if he were
made by changing the amounts of fat-free bread and jelly in the rations. came ill or changed in weight more than 3 Kg. in an B-weeks experiment from the analysis of results, Venous
40
Men who bewere dropped
blood was drawn from each man on 2 or more days at the close of each dietary
period. The separated serum was analyzed in duplicate portions our standardized version of the method of Abel1 et al.16 RESULTS
OF CRITICAL
for total
cholesterol
by
EXPERIMENTS
Table 2 summarizes the results from 5 sets of experiments with diets designed to allow critical analysis of relationships among fats in the diet and serum cholesterol. In each series, subgroups of the men, matched in serum cholesterol level at the end of a preceding period on a standard control diet, subsisted on the experimental diets in different sequences. The experimental diets consisted of a basic low-fat diet of fixed composition covering 55 to 60 per cent of calorie needs to which was added the experimental fat and/or simple carbohydrate (bread made without fat plus sugar as fruit-flavored jelly) to bring total calories to energy balance. In the AB series the iodine value of the fat was almost identical in 2 experiments but there was almost a threefold difference in 2S-P. The third diet had an iodine value double that of the 2 diets while the value of 2%P was practically zero. The serum cholesterol values proved to be parallel to the values of 2S-P but not to iodine value, square-root of iodine value or total fat in the diets. The AC series consisted primarily of a comparison of a low-fat and a moderately high-fat diet, the iodine values and fatty acid compositions in the 2 diets being similar. Serum cholesterol values mirrored 2S-P and total fat, In the AD series, fat provided 41 per cent of the calories in all 3 experiments but 2S-P covered a wide range. The experimental fat in experiment ADRO was rapeseed oil which is remarkable because the mono-ene erucic acid makes up at least half of the total fatty acid. But the results show no great difference from what would be expected from an equal amount of oleic acid, the usual mono-ene in our diets, or equal calories as carbohydrate. The AG series provided a particularly critical test because the 2 diets, both
754
PEYS,
ANDERSON
AND
GRANDE
providing 42 per cent of calories from total fats, were closely similar in 2S-P but greatly different in iodine value .g In experiment AGMO the experimental fat was menhaden oil, specially prepared and supplied to us by Technological Laboratory (Seattle) of the Bureau of Commercial Fisheries, LJ. S. Department of the Interior. The serum cholesterol values indicate, as concluded earlier, that in effect on serum cholesterol of man, the poly-enes in fish fat are similar to linoleic acid.“BlO The P series was primarily a test of the theory that oleic acid is substantially neutral in regard to serum cholesterol and that, in effect, a mixture of equal parts of saturated fatty acid and linoleic acid is not the equivalent of oleic acid. The results are in full accord; the olive oil diet (POO) provided 24 per cent of total calories from oleic acid but it produced exactly the same serum cholesterol as the diet (PSO) m . which vegetable oils low in oleic acid were mixed to give the same value of 2S-P but only one-fourth as much oleic acid (6 per cent of calories) in the diet. The data in table 2 may be used to make a new least-squares solution for equation ( 1 ), above. The result is, again, that mono-ene fails to have a coefficient significantly different from zero and the answer is: (Ia)
Y = 166.2+ 3.07s - 1.76P.
The coefficient of correlation between observed serum cholesterol and the value predicted by (la) 1s . r = 0.954. But if the coefficients 2.7 and -1.3, previously obtained, are used instead of 3.07 and -1.76, the predicted values are not significantly different. Accordingly, there is no reason to change the previous coefficients and we prefer (lb)
? x 167.8+ 2.7s -1.:3P.
Figure 3 shows the agreement between the mean observed cholesterol values and those predicted from equation ( lb ) . Th e 1owest point (circle) in figure 3 represents experiment ABC in which corn oil provided 30 per cent of total calories. As reported previously,‘“.” corn oil consistently gives slightly lower serum cholesterol values than predicted from the fatty acid composition alone, presumably because it contains much more plant sterol than the other fats generally used. When corn oil provides 30 to 40 per cent of total calories, an extra cholesterol depression of about 6-12 mg./lOO ml. is the usual finding. The cross (x) in figure 3 shows where the point for Experiment ABC should be if this allowance is made. In experiments summarized in table 2 and figure 3, corn oil was included in large amounts only in experiment ABC. The analysis resulting in equation ( lb), and the predicted values in figure 3, may not give the best possible correlation because it is implied that, on the average, the men in all 5 sets of experiments are similar, as groups, in their intrinsic cholesterol characteristic; i.e., on a given diet, the average serum values would be the same for all 5 groups of men. Actually, though many of the subjects participated in several sets of experiments and a few in all of them, it cannot be proved that the S groups of men are all random samples from a common population.
SERUM
CHOLFZSTEROL
RESPONSE
TO CHANGES
IN DIET
755
240
230 E 0 0 22Q& a
9; 167.0 + 2.75 -1.3 P
Fig. 3.-Observed serum cholesterol versus the value predicted tion Y = 167.8 + 2.7%1.3P. Data from table 2.
from the equa-
However, with no such assumption, the analysis can be made equally well in terms of differences between experiments within sets; that was the method used in previous analyses where the samples of men were not always so well matched.5J0 There are, then, from table 2, materials for 11 sets of comparisons: ABMF vs. ABLF, ABMF vs. ABC, ABLF vs. ABC, ACMF vs. ACLF, etc. When these differences, AY, are predicted from 2.7AS-MAP, the result closely agrees with the observed values; r = 0.95. The data of table 2, provide materials for critical testing of the hypothesis that the square-root of the iodine value is a reliable predictor of the serum cholesterol value. As indicated above, it is na’ive to neglect the proportion ,of total calories provided by total fat in the diet. If the effect of a dietary fat is directly related to the iodine value of that fat, the more of that fat there is in the diet, the greater will be its effect. Accordingly, equation3 above can be examined with the data from table 2, comparing the average serum cholesterol level on a given diet with the product of Q, square-root of the iodine value of the diet fat, F, the percentage of the total calories provided by that fat. Figure 4 summarizes the results. There is no significant correlation (r = O-10), and the postulate must be rejected. However, as noted above, the serum cholesterol level does tend to be correlated with the iodine value, or its square-root, when total dietary fat is constant, provided that the fats contain no significant amount of fatty acids more unsaturated than linoleic acid, and the proportion of mono-ene does not vary widely.
756
KEYS,
ANDERSON
AND
GHANDE
0
0
00
0 0
180
0
0
0 160.
15OL 0
100
200
300
400
500
E 10
FC W”2
Fig. 4.-Observed Data from table 2.
serum cholesterol
versus the square-root
of the iodine value.
DISCUSSION
The question as to the effects of mono-enes are crucial to consideration of 3S-P versus the iodine value, or its square-root, in the prediction of serum cholesterol response to change of dietary fat. Previous data, as well as those presented here, make it evident that in this respect oleic acid has little or no effect different from equal calories of carbohydrate in the diet. The case for erucic acid being similar is supported by a critical experiment reported her?, but there are also confirmatory data in the literature. We noted earlier the results of the experiment of Malmros and Wigand’” with rapeseed oil which did not appear to depart from expectations if its large content of mono-ene (erucic acid) is ignored. Linko’g has also provided rAvidence on this point from 15 patients who consumed 100 Gm. of rapeseed oil daily for 14 days following a control period. Cholesterol values varied in the serum of those patients but on the average there was a decIine of only about 5 mg. of cholesterol/100 ml. This may seem remarkable if attention is focussed on the iodine value of 104 of the rapeseed oil. But 100 Gm. of rapeseed oil containing 21 Gm. of polyunsaturated and 7 Gm. of saturated fatty acids means that at a total calorie level of 2400 the estimate of 2.7S-1.3P is about -4 mg./lOO ml. for expected change. The diet is certainly not the only factor that influences the serum cholesterol level in man and in the diet there are other variables to consider besides the percentages of total calories provided by the glycerides of the 2 broad classes of saturated and of polyunsaturated fatty acids. However, for direct relevance to free-living men, it is useful to focus on the nutrients, and their quantitative
SERUM
CHOLEXI’JlROL
RESPONSE
TO CHANGES
IN DIET
757
variations, as they occur in natural, spontaneously chosen human diets. Highly abnormal diets can be artificially concocted and fed in special experiments; the interesting results in these situations, and from studies on experimental animals, must not be confused with the situation in long-time real-life subsistence of man. Gross deficiencies of some nutrients may produce some special effects; here we are concerned only with the state of caloric balance and the absence of gross deficiencies of proteins, vitamins and minerals, as well as of special metabolic disorders, and diets in which total fats provide from about 5 to 55 per cent of total calories. In such diets, which surely must cover the range of almost all natural variation, at least in the so-called Western World, besides S and P, on which we have focussed so far, it is necessary to consider several other items. Plant sterols are ubiquitous but they are ingested in only minute amounts except in special experiments, and by the few persons who deliberately incorporate great amounts of some oils, notably corn oil, in the diet. Beveridge et al.1° estimate that around 300 mg. of sitosterol/lOOO calories will produce a statistically significant decrease in serum cholesterol and that a diet providing 15 per cent or more of calories in the form of corn oil could have this effect. Most other vegetable food oils are much lower in sitosterol content. We have reported previously that with experimental diets containing much corn oil it is necessary to allow for this special effect. I7 For each 10 per cent of calories supplies by corn oil, perhaps it is reasonable to estimate an average fall of about 2 to 3 mg. of cholesterol/100 ml. of serum independent of the effect of the fatty acids in the corn oil. Other dietary items to be considered include exogenous cholesterol and the effects of particular saturated fatty acids; these will be discussed separately in later parts of this report. Finally, there is the important fact that people differ intrinsically in their cholesterol metabolism, both as shown by their serum levels on a given diet and in their response to a given dietary change. We have examined this matter previously, I” but propose to discuss it further in Part III of this communication. ACKNOWLEDGMENTS Many persons aided the work reported here but special thanks are due to the volunteer aides from the Church of the Brethren, who literally lived with the subjects at the Hastings State Hospital and more recently at the Faribault State School and Hospital, and to the staff members of those hospitals, particularly Mrs. Helen Williams. Mrs. Mary Jane Schweich and Miss Judy Erickson. For the dietary analyses we are indebted main11 to Miss hlargaret Neibling, Dr. Paul Kupchs and Mrs. Ethel Owens Sowl. Blood analyses in recent years were mainly done by Mrs. Gayle Frohlich Olson and Mrs. S. Nishimoto, among others. Mrs. Nedra Foster did much work on the dietary designs as well as on supervision of the chemical work. Willis Parlin, Norris Schulz and h4iss Joan La Riviere greatly aided the statistical analyses.
REFERENCES 1. Kinsell, L. W.: Some thoughts regarding the P:S ratio concept. Amer. J. Clin. Nutr. 12:228, 1963. 2. Grande, F.: Prediction of serum choles-
terol changes produced by dietary fats. Amer. J. Clin. Nutr. 13:122, 1963. 3. Gunning, B., Imaichi, K.. Splitter, S. D.,
758
KEYS,
and Kinsell. L. W.: Effects of different dietary fats on plasma-lipid levels. Lancet 2:336, 1984. 4. Ahrens. E. H.. Jr,. Hirsch. J.. Insull. W.. Jr.. Tsaltas, T. T.. Blomstrand. R.. and Peterson. h4. L.: The influence of dietary fats on serum-lipid levels in man. Lancet 1:943. 1957. 5. Keys. A., Anderson. J. T.. and Grande. F.: Prediction of serum chole: terol responses of 111an to changes in fats in the diet. 1 ,ancet 2:939, 1957. 6. -,
-, and -: Efl’.sc t on serum cholxterol in man oE mono-ene fatty acid (oleic acid) in the diet. Proc. Sot. Exp. Biol. Med. 98:387. 1958. 7. Grande. F.. Mataumoto. Y.. Anderson. J. T.. and Keys. A.: Effect of dietary rapeseed oil on man’s serum lipids. Circnlalion 26:653, lS62. 8. Ahrens. E. H.. Jr.. 11~~11. W.. Hirsch. J.. Stoffel. W.. Peterson. L., Farquhar. J. W.. >4iller, T.. Thomasson. H. J.: The effects human sermn-lipids of a dietary highly unsaturated. hut poor in
Jr,. hl. and on fat. es-
sential fatty acitls. Lancet 1: 115, 1959. 9. Grande, F., Ancler:.on. J. T.. and Keys, A.: A comparison of the effect of fish oil polyunsaturated fatty acids and of linoleic acid on man’s sermll lipids. Amer. J. Clin. Nutr. 12:331. 1963. 10. Keys. -4.. Anderson. J. T.. and Grand?, F.: Serunl cholesterol in man: diet fat and intrinsic responsiveness. CirIl.
culation 19:201. 1959. Ahrens. E. H.. JI-._ Hirsch. J.. Insull. W.. Jr.. and Peterson. M.: Chemistry of Lipides as Related to Atherosclero-
ANDERSON
AND
GHANDE
sis. Ed. by I. H. Page. Springfield Ill.. Charles C Thomas. 1938. p, 222. 12. Slalmros. A.. and Wigand. G.: The effect on serum-cholesterol of diets containing different fats. Lancet 2: 1 1957. 1:3. Turpeinen. 0.. Roine. I’.. Pekkarincn. M., Karvonen. M. J.. Rarttanen. Y.. Runeberg. J.. and Alavirta. I’.: Efi’cct of serum cholesterol level of reglac~nlcnt of dietary milk fat by soybra oil. Lancet 1: 196. 1960. 1-I. Anderson. J. T., (:rande. F., Chlouverakis. C.. Proja. ?\I.. and Keys. .4.: Effect of dietary cholesterol on scr,mI cholesterol level in Nan. Fed. Proc. 21: 100. 1962. -. Mats~mioto. Y.. and Keys. A.: Glucose. sucro:x and lactose in tlrc: diet and blood lipid in man. J. Nlltr. 79: 349. 19ci:3. 16. -. ant1 Keys. A.: Cholesterol in serlun alld lipoprotein fractions: Its measurement and stability. Clin. Chem. 2:
15. -_
145, 1956. 17. (Zrande. F.. Anderson.
J,
T.. and Ke1.s.
A.: Serunl cholesterol in man ;kntl unsaponifiahle fraction of corn oil in the diet. Proc. Sot. Exp. Biol. Med. 98:436, 1958. 18. Linko. E.: Vegetable oils and sermil experiments cholesterol. Short-term with rapeseed and sunflower seed Acta med. Stand. 159:475. oils. 1957. 19. Beveridge, J. M. R.. Haust. II. Connell. W. F.: Magnitude hypocholesterolemic effect of sitosterol in man. J. Nlltr. 1964.
Ancel K~Js, Ph.D., Professor, School of Public Health, Director, Laboratory of Physiological Hygiene, Vniversity of Minnesota, Mi.nneapolis, Minn. Joseph T. Anderson, Ph.D., Professor, School of Public Health, University of Minnesota, Minneapolis, Minn. Francisco Grade, M.D., Professor, School of Public Health, University of Minnesota, Director, Jay Phillips Research Laboratory, Mount Sinai Hospital, Blinncapolis, Minn.
L.. and of the dietary 8:): 119.