  |
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Journal of the New Zealand Medical Association,
24-January-2003, Vol 116 No 1168
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Ischaemic heart disease, Type 1 diabetes, and cow milk A1
β-casein
 Murray
Laugesen and Robert Elliott
In 1968, across 43 countries, Seely found that per capita
milk supply (excluding butter) was highly correlated with IHD mortality rates 1
to 4 years later, and more strongly than animal fats or
butter.2 In 1968 also,
came the first dairy science review of the genetic differences in milk proteins
between individual cattle and
breeds.3 In North Europe,
A1 was the predominant form of β-casein in cow milk (A1/β =
0.46–0.71) among the traditional black-and-white or red-and-white cow
breeds (Red Danish, Holstein-Friesian,
Ayrshire).4 Artificial
breeding during the 1970s and 1980s made use of American Holstein bulls
(typically A1/β = 0.4), often replacing indigenous cow breeds, and reducing
A1 levels in North European milk. In central and southern Europe, the A1/β
fraction was lower due to the dominance of Jersey, Simmental or Swiss Brown
cattle (A1/β mostly
<0.25),4 and virtually
nonexistent in Guernsey cattle (A1/β =
0.01).5 Except for milk
from the island of Guernsey, commercially-sold European-breed cow milk is an A1
β-casein and A2 β-casein mix.
Ischaemic heart
disease
Alerted by Elliott to the inter-country correlation between
Type 1 diabetes and A1/capita in cow
milk,1 McLachlan patented
a method to commercialise a possible relationship between A1/capita and IHD. In
2001, McLachlan published a 17-country ecological study, demonstrating a high
correlation between A1/capita in the food supply circa 1980 and IHD mortality in
1985 and 1990.6
Fonterra Research Centre (FRC, formerly the Dairy Research
Institute) scientists responded, noting that correlations found in past years
between milk protein consumption and IHD across 40 countries (some high income,
some low), were no longer found in the 1990s; observing any past correlation
appeared to have been “serendipitous”; and that the evidence was not
sufficient to warrant a change to A1-free
milk.7
In examining these competing claims, we have put aside the
question of biological mechanisms which these authors have touched on, and
confined this study to the correlations. This study of IHD is limited to
healthcare-affluent countries, to reduce inter-country IHD mortality differences
due to disparities in the availability of coronary care. For comparison, we also
examine the correlation of other food and nutritional supply variables with
IHD.
Diabetes Type
1
Diabetes Type 1 (DM-1) incidence has been increasing
globally at 3% per
annum.8 Its incidence
varies by over 300-fold across 51
countries.9
Although knowledge of genetic predisposition has increased, the nature of the
precipitating environmental factors remains elusive. Across 12 countries, milk
protein per capita and DM-1 rates were highly
correlated.10 Even when
Finnish children of the same genetic susceptibility to DM-1 were compared, those
consuming more than three glasses of milk daily remained at higher risk of DM-1
than those with a lower milk
intake.11
No one could explain why Iceland, with high milk
consumption, had a lower DM-1 incidence rate than the other genetically-related
Nordic countries. Genetically-predisposed non-obese diabetic (NOD) mice
developed DM-1 when fed milk from the European Bos taurus cow, but not if fed
milk from the Indian Bos indicis Zebu. Fed European cow milk casein, they
developed DM-1, but not if fed cow milk whey (the other main protein fraction in
milk) or soya protein. When fed the A1 variant of β-casein, the mice
developed DM-1, but not when fed A2 or (fully) hydrolysed A1 β-casein. (p =
0.002) Also, A1 β-casein had no effect when given with naloxone, a morphine
antagonist that opposes the opioid effect of
β-casomorphin-7.12
β-casomorphin-7 is a peptide formed by partial hydrolysis of A1, B or C
β-casein only, the cleavage made possible by a histidine rather than a
proline amino-acid at position 67 in these caseins. Milk-free cereal induced
DM-1 in genetically-predisposed BB (BioBreeding Laboratories, Ottawa)
rats13 and in NOD mice
also. Whole casein without cereal also induced DM-1 in NOD mice; whereas
A1/β in BB rats had only a small effect, suggesting that even if milk were
a factor in DM-1, some DM-1 would remain due to cereals in the
diet.14
In a 10-country study in 1999, Elliott found that DM-1 rates
were significantly correlated with A1/capita and particularly with the combined
A1 and B variants (of β-casein) per
capita.1 In revisiting
this study, we include nine more countries, and adjust for milk imports and
their source, for the protein yield of each breed, and for the proportion of
milk from other animals. We also estimate the A1/capita in the milk and cheese
supply separately.
MethodsCountries
selected These included all 22 countries (Tables 1 and 3) for which
published A1/β cow milk data was obtainable, after: 1) excluding the
Netherlands as simultaneous high imports and exports of milk precluded reliable
determination of the origin of milk consumed (imports 65%; exports 85% of
domestic usage in 1995); 2) for IHD only, excluding Hungary and Venezuela, as
their total health expenditure was less than US $1000 per capita in 1995 (based
on purchasing power
parities),15
leaving for study 20 “healthcare-affluent” countries (17 of which
were OECD member countries, out of 22 healthcare-affluent countries and 29, in
total, in the OECD in
199515) 3) for DM-1 only,
excluding countries not surveyed by WHO DiaMond Project or EURODIAB ACE
(Ireland, Jersey and Guernsey), leaving 19 for study.
Milk and cream
supply Milk and cream supply per capita was calculated from the
nutritional statistical databases at the FAO (Food and Agricultural
Organization) web site,16
as milk protein per capita in grams per day. This was calculated as [3.3% by
weight of ‘milk excluding butter’ + 2.7% of cream, minus 25% of
cheese]. Milk included fresh milk products – yoghurts, cream, whole and
skim milk, and milk powder – but excluded cheese and butter. We subtracted
goat and sheep milk production (FAO data, Italy 8%, Israel 2%, Hungary 2%).
Where imported milk comprised 20% or more of domestic milk usage (Germany,
Italy, Japan),17 we
adjusted for the imported tonnage and the A1/β of milk in the main
supplying countries, and similarly for cheese imports.
Nutritional data FAO
food supply data16
(unavailable for the Channel Islands) were converted to nutritional measures
using British food composition tables, from Health New Zealand’s food and
nutrition database, listing values for 77 foods and 110 nutritional measures of
national food
supplies.18
Cow breed
distributions Iceland, Norway, Jersey, Guernsey traditionally, and Israel
and Japan in recent decades, were virtually one-breed countries. For other
countries, we calculated the breed distribution from governmental animal census
data,19 from
industry,20,21 and
otherwise from national breeding programme
data.22
A1/β and other
β-casein fractions These were estimated by breed from the dairy
science literature held by the Fonterra Research Centre (FRC) for 18 countries.
In addition, factory or retail milk was tested from 11 Table 1 countries during 1998–2001. Genotype
estimates were based on published A1/β herd tests, for
Austria,23
Canada,24
France,25
Germany,26
Hungary,27
Japan,28 New
Zealand,29 Nordic
countries,30
Switzerland,31 the United
Kingdom,3 and the United
States.32 For the new
Israeli Holstein
breed,33
we assumed US Holstein averaged
A1/β values from US herd
test reports from 1968,3
1971,34 and
1989.32 For Italy,
Professor F Addeo supplied data on 23 breeds, (personal communication, June
2000).
The A1/β
fractions for breeds in Ireland
and the Channel Islands were those of the same breeds tested in mainland
Britain. In Iceland, tests were of
herds6 and of bulk milk
samples.30,35 In the
absence of recent breed estimates or market share data, we used FRC test results
on milk or milk powder from Australia (average of two samples) and Venezuela
(average of 17 brands). During 1998–2001, milk test results were obtained
from an additional nine countries: 1) from the Nordic
countries;35 and 2) from
Canada, Italy, New Zealand, and the UK, tested by FRC. For DM-1, 1990 and 1995
A1/β data were averaged to represent the 1990–4 fraction in the
national milk supply.
A1/capita in 1990 (IHD) and in 1990–4 (DM-1) was
estimated thus:
A1/capita = (cow%) x (milk protein supply/capita) x (β-casein/cow milk protein) x (national
A1/β)
Cow% is total milk production percentage minus the percentage from sheep and goat milk. Milk protein supply/capita is defined above. β-casein as a fraction of cow milk protein = 0.284.36 National A1/β = sum of the percentage contribution of each breed, weighted for its percentage of the national dairy cow population,19–22 the protein content of its milk,22 and average milk yield of its cows.22 Mortality
data
WHO37 and its website (www.who.int) supplied total cardiovascular disease
(CVD), IHD, and cerebrovascular (stroke) mortality data for 18 countries, and
Channel Islands data were supplied by their Departments of Health.
The annual mortality rate per 100 000 population in the age group
35–64 years was standardised by averaging of the rates for the component
three ten-year age groups, and then averaging male and female rates. A lag of
five years was allowed from food supply to IHD
mortality.38
DM-1
incidence
data This was provided by the
1990–94 WHO DiaMond
Project,10
except for Iceland and Switzerland which were surveyed by the EURODIAB
ACE Study group (Table
1).39 Within each
country, regional incidence results were averaged. Some were national surveys
(Table 1). Rates for age 0–14 years were
standardised by averaging six rates: 0–4, 5–9 and 10–14 year
age groups by gender.
Food variables Over
75 food and over 100 nutritional food supply variables for 1990 for all Table 1 countries (except Venezuela due to lack of data)
were obtained from FAO16
and from Health New Zealand’s food and nutrition
database18 and tested for
correlation against DM-1 in 1990–4.
Income Gross
domestic product per capita and health expenditure data for 1995 were given in
1995 US$ in purchasing power parities, base
1995.15,40
Tobacco and alcohol
availability Availability of tobacco products was from tax paid data, per
adult age 15 and over.41
Cigars were omitted as they are not as strongly related to IHD as cigarettes.
Alcohol, from FAO data, was given per capita.
Correlations were tested for significance by PEPI
software,42 and multiple
regressions by Excel 2001.
ResultsIn Table 1, countries were ranked
by IHD rate. The Japanese, ranked lowest for IHD (21.5), consumed the least milk
and least saturated fat.
The next six countries with lowest IHD mortality were from
central Europe or the Mediterranean. Their milk consumption was low (except for
Switzerland), and A1/capita was also low. France (28.8) ranked second lowest
overall for IHD; and Guernsey (40.7), where milk has been virtually A1-free for
a century or more, ranked third lowest.
To test for possible under-diagnosis or under-classification
of mortality to IHD, we also ranked countries by (total) CVD, and by CVD minus
stroke. Correlation with IHD was r = 0.92 for CVD, and 0.91 for CVD minus
stroke. For CVD, France ranked lowest, and for CVD minus stroke, second lowest;
while Switzerland ranked third lowest for both, and Guernsey ranked fourth
lowest for both.
The countries of North Europe, including countries they
populated (with their cattle) – North America and Australasia –
filled the lower half of the table and tended to consume more milk. In the last
four ranks of Table 1, countries with the highest IHD
rates all have more than 2 grams of A1/capita per day in their milk
supply.
Tobacco and alcohol
availability In 1990, neither tobacco nor alcohol was significantly
correlated with IHD five years later. Tobacco product per capita sales were
highest in Switzerland, and lowest in Sweden. Alcohol availability was highest
in central Europe – in Germany, Austria, France and the Channel Islands
– and lowest in Japan. Table 2 shows that wine supply/capita was
moderately inversely correlated with IHD.
Dietary fat factors
Dietary fat factors in univariate analysis (Table 2) showed significant
correlation with IHD – the atherogenic index, estimated from the
configuration of six dietary
fats43 (r = 0.50); and
myristic, the 14-carbon saturated fatty acid. The Hegsted score (using food
supply fats to estimate population serum cholesterol), saturated fat, and dairy
fat were significantly correlated with IHD in 1980 and 1985, but not in 1990 and
1995.
Table 2 Correlations of supply variables, with
ischaemic heart disease five years later, 1980-95, 18 countries
Table 2 countries are Table 1
countries minus Guernsey and Jersey. *p <0.05;
†p<0.01;
‡p <0.001;
ns=not significant; r=correlation coefficient; b=univariate regression
coefficient, with 95% confidence intervals; e=elasticity=% change in IHD rate
related to a 1% change in the food supply variable, estimated as b*(supply
variable mean/mean of 1995 IHD rate). The mean IHD rate was 80.8 for these 18
countries in 1995.
Cow milk protein supply
variables Milk protein supply variables were all more strongly correlated
with IHD five years later than any dietary fat variable in Table 2: A1/capita in
milk and cream (r = 0.76, p <0.0005); A1/capita in milk, cream and cheese
combined (r = 0.66); and cow milk protein per capita (r = 0.65). Across 18
countries studied over 20 years, of over 180 variables tested, A1/capita
correlated most closely with IHD five years later. A 1% change in A1/capita in
1990 was associated with a 0.57% change in IHD mortality in 1995. Correlation
between 1990 A1/capita and IHD in 1995 were stronger for male IHD (r = 0.83)
than for female IHD (r = 0.69). Length of lag was not critical to the
correlations. A1/capita in 1975, 1980, 1985 and 1990 was correlated with IHD in
1995 (r = 0.82, 0.76, 0.82, 0.76 respectively).
In 1995, A1/capita varied greatly among countries, from
about 0.3 g/day in Guernsey, to 3.0 g/day in Finland. Between 1975
and 1995, the
20-country averages decreased for milk protein and for A1/capita by 14%, and
remained the same for A1/β; IHD decreased 57%. Most of the decreases
occurred between 1985 and 1995, when A1/capita decreased 13%, while IHD
decreased more, by 37%. The correlation of the annual rates of change
between A1/capita and IHD was not statistically significant.
Cheese, butter and cream,
and the other caseins A1 consumed as cheese (after adjusting for imports)
weakened the correlations of A1/capita in milk and cream with IHD five years
later, from r = 0.76 to r =0.66. Adding A1 in cream added approximately 1% to
the strength of the correlations between milk and IHD. Butter (1.1% protein by
weight), in high-consuming countries, added 3–4% to estimated A1/capita,
but increased the correlation of A1/capita with DM-1 by 0.001 only, and was
omitted from all tables.
Per capita supply based on the B variant β-casein in
milk and cream, or on any combination of A1, B and C variants of β-casein
therein, or on including the A variant of kappa-casein, or including estimates
of A1 in cheese, decreased correlations with IHD.
Analysis On
multivariate analysis of the variables with highest univariate positive
correlation with IHD in 1995 in Table 2, only A1/capita gave a significant
result. When all 4 five-year periods in Table 2 were combined for estimating IHD
five years later, the only significant variables were the downward trend with
time, as given by the calendar year, A1/capita, and plant polyunsaturated fat
(PUF).
Table 3 Average
supply of milk protein per capita as milk or cream in 1990–4 varied more
than fourfold, from 5 g/day in Japan, to 22 g/day in Finland and
Sweden.
Table 3. The per capita supply of A1 β-casein and
milk protein, 1990–94, and incidence of diabetes mellitus Type 1 at age
0–14 years, 1990–94, 19 countries
*fresh milk equivalents, excluding cheese and butter,
including cream and yoghurt
Note: The DM-1 surveys were carried out in 1990–3 for Australia and Japan, 1989–94 for Iceland 1991-4 for Switzerland, in all others during 1990–4. Milk supply data were averaged for the same years as the DM-1 survey in that country. Source: WHO-DiaMond Project;10 for Iceland and Switzerland, EURODIAB ACE study group.39 The A1/β fraction of milk casein varied from 0.21 in
Austria to 0.53 in the UK. A1/capita supply varied sevenfold, from 0.4 g/day in
Venezuela to 3.0 g/day in Finland.
From 19 countries surveyed, 12 913 new cases of DM-1 were
detected, with a country annual average rate of 14.3 new cases per 100 000
children age 0–14 years (boys, 14.6, girls 13.9). The DM-1 rate varied
nearly 300-fold, from 0.13 in Venezuela to 36.5 in Finland.
The DiaMond Project surveys supplied age-specific DM-1 for
17 countries.10 The
country-average DM-1 rate increased from 9.3 per 100 000 at 0–4 years, to
15.9 at 5–9 years, to 18.9 at 10–14 years of age.
The correlation of milk protein with DM-1 was equally high
in all age groups (r = 0.80, 0.81, 0.81) and for 0–14 years, r =0.82. The
correlation with A1/capita was r = 0.91 for boys, and 0.90 for girls (p
<0.001), and equal at 0–4, 5–9, and 10–14 years of age. For
the five Nordic countries, correlation was similarly high (r = 0.91, p
<0.05).
For 51 countries surveyed, DM-1 was significantly correlated
with milk supply, including cheese (r = 0.70, p <0.001).
Table 4. Correlations of cow proteins per capita supply
with incidence of diabetes mellitus Type 1, age 0–14 years, across 19
countries, 1990–94
Based on Table 3.
*p <0.05; †p<0.01; ‡p <0.001, ns = not significant §DM-1 rate standardised by averaging six age-gender groups (0–4, 5–9, and 10–14 years) ║For the 17 countries surveyed by DiaMond Project r=0.82; b= univariate regression coefficient, with 95% confidence intervals; e=elasticity=% change in DM-1 rate related to a 1% change at the mean in the cow protein variable, estimated as b*(cow protein variable mean/mean of DM-1 rate) Table 4 DM-1 at age
0–14 years was correlated with the quantity of milk and cream in the food
supply, as measured by milk protein per capita (r = 0.68, p <0.001). DM-1
correlated particularly with A1/capita in the food supply in the same years (r =
0.92, p <0.001); but not with B and C β-casein per capita. A 1% change
in A1/capita was associated with a 1.3% change (elasticity) in DM-1.
Across 75 foods and over 100 nutritive food supply variables
across 18 countries, the highest positive correlation was with milk protein (r =
0.64), and with oats (r = 0.70). Latitude
was significantly correlated with DM-1 (r = 0.65, p <0.005), but not
with IHD.
None of the genotype fractions of β-casein or of
kappa-casein in milk correlated significantly with DM-1, except A1/β (r =
0.47, p <0.05). The correlations of casein genotypes were significantly
inter-correlated, but among single genotypes, only A1/capita markedly exceeded
milk protein per capita’s correlation with DM-1. The combined A1+B, and
A1+B+C per capita correlations with DM-1 were weaker than for A1/capita alone.
The A variant fraction of kappa casein per capita was significantly correlated
to A1/capita, and was correlated with DM-1 in the same range as total milk
protein. A2 β-casein in milk protein per capita was significantly
correlated with DM-1 (r = 0.47, p <0.05), but less than total milk protein (r
= 0.68, p <0.01). A1 β-casein in cheese per capita was significantly
correlated with DM-1 (r = 0.46, p <0.05), half the r = 0.92 value for A1
β-casein in cow milk.
Elliott et al found a correlation (r = 0.77) of A1/capita
with DM-1 across 10 countries based on DM-1 surveys before
1991;6 in the present
study, for 1990–4, across the same countries we confirmed this correlation
(r = 0.84, p <0.005). For the additional nine countries in this study, r =
0.90 (p <0.001). Elliott et al found that combining per capita A1 and B
β-caseins gave an improved correlation with DM-1; in this study, across the
same countries, inclusion of B β-casein weakened the correlation.
Sensitivity of the methods
of estimation The correlation of A1/capita in 1990 with IHD in 1995
varied from r = 0.75 for the 10 countries estimated from breed data only, to r =
0.69 for the 10 using both methods. This was due to the fact that the highest
and lowest values were found in the former group. Correlating for any 19
countries out of 20, the r value varied from 0.69 (deleting Ireland) to 0.80
(deleting Australia). Deletion of the smallest countries (Jersey, Guernsey and
Iceland) made no difference. Deletion of all three countries whose individual
deletions most lowered the correlation (Ireland, New Zealand and Guernsey),
lowered the correlation to r = 0.62.
For DM-1, removal of one country at a time from the
correlation with A1/capita resulted in a correlation between r = 0.89 and 0.94
for the remaining 18 countries. Deletion of the two highest data points in Figure 2 (Finland and Sweden), lowered the correlation
to r = 0.65. When the analysis was confined to the 11 countries analysed by
consumer milk tests and breed-based estimates, the correlation was r = 0.69, but
the consumer milk tests were all carried out approximately five years
later.
DiscussionCorrelation, even when
statistically significant, does not prove causation, though it may raise that
possibility. This is an ecological study with the limitations implied: its main
value is to focus further research, and is not by itself a basis for public
policy. For example, the supply of tobacco products in Table 2 was not
significantly correlated with IHD five years later. This is at variance with
individual-based studies – tobacco precipitates IHD mortality in tandem
with atheroma, and many high-IHD populations have reduced their tobacco and
saturated fat consumption. Data aggregated at national level may, however, be an
efficient means of locating effects which though small for each individual are
detectable in the population mean. In addition, extreme caution is required in
interpreting correlations involving dietary factors, which are often
inter-correlated.
Sampling errors may have occurred, due to surveying only
parts of some countries for DM-1. Testing dairy herds to characterise a national
herd risked local variations. Cow registration data may have over-represented
high-production breeds in characterising the national herd. A single retail milk
sample may not accurately characterise a country’s milk for that year.
Testing methods for A1/β vary somewhat. Japan lacked recent A1/β data.
Israeli cows or milk were not tested; instead, tests of the United States
Holsteins, from which the Israeli herd derives, were used.
A1/β
estimates were obtainable for 22 countries in total. As herd A1/β test
results may not have been representative of the national herd, we
supplemented herd tests with consumer milk tests. For New Zealand, with separate
herds milked for export and domestic supply up to 1992, any estimates were for
the latter.
Ischaemic heart
disease
In reviewing the results of the WHO MONICA (Monitoring
trends and determinants in Cardiovascular disease) Project in 21 countries
including New Zealand, the authors concluded that “the results support
prevention policies based on the classic risk factors but suggest potential for
prevention beyond
these.”44 It may be
timely to consider new concepts, including McLachlan’s A1
hypothesis.7
The IHD-A1 correlation was only valid for
“healthcare-affluent” countries. We assumed that a certain level of
health expenditure was needed to achieve a reasonable minimum chance of survival
of IHD. Per capita income was not significantly correlated with IHD mortality in
the 20 countries selected, but mortality may be correlated with national
expenditure on healthcare, for those with IHD surviving long enough to reach
hospital. Income and IHD would have correlated significantly had we included
Hungary (low GDP/capita, low health expenditure, highest IHD rate).
Similarly, the higher correlations with IHD found for
A1/capita compared with other fat variables may be true only of
healthcare-affluent countries. For example, inclusion of Hungary, with its
atherogenic index and IHD rate each exceeding any such values in Table 1, resulted in IHD being more correlated with the
atherogenic index than with A1/capita. On the other hand, low expenditure on
healthcare can be expected to decrease survival and raise IHD mortality,
outweighing dietary influences. Austria had a higher IHD mortality than
A1/capita and dietary predictors would suggest. This was not due to
misclassification of IHD (based on CVD minus stroke mortality). We were not,
however, able to compare Austria with other countries for adoption of effective
coronary care practices.
Food supply statistics have been significantly correlated
with survey data across many
countries,45 but food
supply statistics are the only feasible way to compare all countries across
time. We assumed that per capita milk supply was proportional to its consumption
in childhood or at age 35–64 years.
FAO assigns an average 3.3 g of fat and 3.3 g of protein per
100 g of milk. This standardised milk will give the same correlations for its
protein, fat, saturated fat or any other fixed component. Table 2, however,
shows that the total protein in milk was more highly correlated with IHD rates
from 1980 to 1995, than was total saturated fat, total dairy fat or butter. If
the milk effect was due to its dairy fat content, then the opposite should have
been true.
Saturated fat recorded for individuals in the 1960s had a
high correlation (r = 0.85) with IHD 10 years later across the 16-cohorts of the
Seven Country Study.46
Between 1970 and 1990, of the 18 countries listed for saturated fat in Table 1, the 11 English-speaking or Nordic countries,
with mostly high IHD rates in 1970, all lowered per capita saturated fat in
their food supply, while the other countries in Table
1 with mostly low IHD rates, increased it. This reduced the correlation
between the per capita saturated fat supply and IHD in 1990, and 1995 (r = 0.37,
Table 2).
Based
on the estimated A1/capita in milk, cheese contributed an estimated country
average of one third of the A1 β-casein in the diet, and one half or more
in France, Germany and Italy. Inclusion of cheese weakened the correlation with
IHD of A1/capita in milk and cream by 10 percentage points (Table 2). The extent
of decrease in A1/β in manufacture or during shelf life may vary by brand
or country. In any case, A1/capita in milk, whether including A1/capita
in cheese or not, was more strongly correlated with IHD 5 years later than any
other food supply variable found.
Plant foods (cereals, rice, nuts, beans, potatoes, olives,
peas, but not counting vegetables) as measured by polyunsaturated fat (PUF),
were closely associated with low national rates of IHD mortality. Of Table 1 countries, Italy, Japan and Switzerland had the
highest consumption of polyunsaturated fat in plant foods.
The Mediterranean diet has been followed most closely among
Table 1 countries by Italy and, though consuming less
olive oil, by Israel; both countries ranked low in IHD mortality. The
Mediterranean diet, as surveyed in Crete in 1948, derived half its calories from
cereals, nuts and pulses, one third from olive oil, and the rest from vegetables
and fruit.47 Milk
consumption was low (milk protein 8
g/day).48 Cretan men had
the lowest IHD mortality in the Seven Country Study, mostly attributed to low
saturated fat. An additional explanation is that A1/capita supply on Crete was
low, possibly only 0.5 g/day, due to a low milk supply, and particularly due to
low cow-milk availability. Forty per cent of milk in Greece was goat or sheep
milk,49
which contains no A1 β-casein.
The “French paradox” refers to France’s
low IHD mortality (second lowest, Table 1) despite a
high butter supply (second highest, Table 1). This was
not a misclassification error – in 1995, France also had the lowest CVD
mortality, and French women had the second highest life expectancy.
France’s low IHD mortality has been attributed to wine, garlic, plant PUF
or vitamins.50 In France
alcohol supply/capita was one half higher than in Ireland; and wine
supply/capita 16 times higher (Table 1). In Figure 1, the French IHD data point is below the 95%
confidence limits for A1, and high French alcohol consumption may explain this.
However, alcohol was not correlated with IHD (Table 2). The Irish/French IHD
rate ratio of 3.8:1 was in line with the milk protein ratios (total
protein/capita, 3.1:1; A1/capita, 4.1:1) (Table
1).
The A1 hypothesis, if confirmed, could explain the low IHD
rates in Mediterranean countries, and the Irish–French IHD
differences.
Diabetes Type
1
Of over 170 foods and nutritional variables in the food
supply in 1990, milk was the only food highly correlated (r >0.60, p
<0.01) with DM-1, apart from the northern European crops of oats and rye,
which with latitude, may merely reflect the geographical distribution of
A1/capita supply.
In this 19-country study based on 1990–4 surveys, we
confirmed Elliott’s pre-1991 findings from 10 countries that A1/capita was
highly correlated with
DM-1,6 but found that
with B (or C) β-casein added, the correlation decreased, and B or C
separately were not correlated with DM-1 (Table 4).
Elliott noted that the distinctive peptide formed mostly
from A1 β-casein and partly from B β-casein was β-casomorphin-7,
and this was possibly the active ingredient. Lack of correlation between DM-1
and B β-casein raises the possibility that B β-casein, which differs
in solubility, may be processed differently by the intestinal mucosa. Countries
with above-average B β-casein were Australia, Austria, Denmark, France,
Germany and Venezuela. Milk may be exposed to different temperature patterns at
the farm, or during processing, across countries and time periods.
The method used here and by
Elliott1 assumed that
childhood milk consumption was proportional to its per capita supply. Surveys
have since confirmed that A1 β-casein consumption of two-year-old children
was lower in Iceland (1.7 g/day) than in other Nordic countries (average 2.4
g/day).51
No such difference was found in surveys of 11–14
year-olds.51 The
correlation between DM-1 and A1/β was equally high at 0–4, 5–9
and 10–14 years of age, suggesting that early childhood exposure to cow
milk A1/β may permanently change the islet cells, making them prone to
other factors or processes that cause islet cells to die at a later age. Anti-A1
antibodies tend to be higher in DM-1 diabetics and their siblings, while anti-A2
antibodies tend to be higher in their parents and controls. This suggests a
defective immunotolerance to cow milk antigens in
DM-1,52
possibly due to
β-casomorphin-7.53
The A variant of kappa casein/capita (Table 4) was highly
correlated with A1/capita (r = 0.79), and less so with DM-1. Genes for kappa and
beta casein are situated very close together on cattle chromosome
6.54 Besides A2, B and C
β-casein, other cow proteins in the milk supply – albumin,
immunoglobulin, and lactoferrin – showed no correlation with DM-1 in
Nordic
countries.51
A1
β-casein in
cheese per capita (estimated from A1/β of the milk) did not correlate with
DM-1, certainly not as closely as A1 casein in milk and cream. First, child
consumption of cheese, more than milk, was likely to vary from adult
consumption. Second, due to wastage, cheese supply may not reflect consumption
as does milk supply. Third, its A1/β ratio was likely to vary in ways not
predicted by the A1/β of the milk it was made from. Proteolytic enzymes,
salts, temperature during manufacture, and on-shelf
ageing,55
can vary the A1/β ratio between types of cheese. Information on the market
share and on-shelf A1/β tests for each cheese type might improve the
estimate of national A1/capita from cheese as consumed.
Insulin-dependency makes for a clear definition of DM-1, and
diabetic registers and the second round of the DiaMond survey have made for very
high ascertainment. The increase in DM-1 rates with age during childhood
suggested environmental causes. From 1960 to 1996, A1/capita across 17 countries
with historical data declined 21%, or 0.6% a year, whereas the rate of DM-1
increased by an average 3% a year in 37 populations
surveyed.9 Other factors
besides A1/capita and milk supply per adult are needed to explain the global
increase
in DM-1 in children. While the
milk supply has decreased, child nutrition surveys are needed to determine
whether rising incomes
and the marketing of infant formula, coloured and flavoured milks, yoghurts, and
ice cream may have led to increased children’s consumption of cow protein,
and thereby increased A1/capita or β-casomorphin-7 in their
diet.
A DM-1 rate was not calculated for Guernsey for Table 3, as
only five DM-1 cases were found in 1990–4, and we had no information on
genetic predisposition to DM-1 among Guernsey children. Clearly, however, milk
very low in A1 does not entirely prevent DM-1 from occurring. Similarly, Jersey
residents consumed only Jersey milk (A1/β = 0.09, A1/capita 0.3 g/day), and
again numbers were small.
The low Venezuelan DM-1 rate may reflect incomplete
ascertainment but, even assuming it was 10 times higher, the correlation over
all countries was unaltered at r = 0.92. DM-1 in countries excluding Australia
had a correlation with A1/capita of r = 0.94; Australian DM-1 data were based on
one survey site in New South Wales.
In summary, from 1980 to 1995, IHD mortality in 20
healthcare-affluent countries was more highly correlated with total milk
proteins (and particularly A1) than with fats in the food supply at population
level, providing an alternative and testable hypothesis to explain the higher
IHD mortality in northern compared to southern Europe.
Across 51 countries
surveyed, DM-1 rates were significantly correlated with per capita fresh milk
protein, and in 19 countries for which data were available, A1
β-casein/capita substantially increased this correlation, from 68% to 92%.
In contrast, A2, B and C variants of β-casein in milk, and cheese proteins,
correlated less strongly with DM-1 than total milk protein.
The correlations of A1/capita with DM-1 and IHD rates raise
the possibility that intensive breeding of cows over many years may have
emphasised a genetic variant of milk with adverse effects in humans. Clinical
trials will be needed to determine whether A1-free milk can reduce the risk of
DM-1 and IHD.
Author information:
Murray Laugesen, Public Health Physician, Health New Zealand; Robert B
Elliott, Emeritus Professor, University of Auckland, Auckland
Acknowledgements:
Fonterra Research Centre, Palmerston North, for testing milk specimens; Dr
Alistair Stewart, Department of Community Health, University of Auckland for
statistical advice; and Professor Rod Jackson of the same department, for advice
on an earlier IHD draft.
Conflicts of
interest: The IHD section of this paper was funded by a grant from A2
Corporation, Auckland, to the first author, who is a minor shareholder in A2
Corporation. Both authors are directors of the NZ Milk Institute Ltd, which owns
a patent related to A1-free milk.
Correspondence: Dr
Murray Laugesen, Health New Zealand, P O Box 25-920, St Heliers, Auckland.
Email: laugesen@healthnz.co.nz
References:
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| | Current
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads) Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals |
|