NORTH CAROLINA STATE LIBRARY
RALEIGH
PHSB STUDIES
JUN i
/v. c.
Doc.
5 m
A Special Report Series by the N,C. Department of Human Resources, Division of
Health Services, Public Health Statistics Branch, P.O. Box 2091, Raleigh, N.C.
No. k
May 1977
ASSOCIATIONS BETWEEN NUTRITION AND MORTALITY
IN SIX HEALTH SERVICE AREAS OF NORTH CAROLINA
(An Indirect Study)
In recent years, considerable attention has been given to studies relating
disease processes to various elements of human consumption. These studies have
experimentally implicated a number of products including cyclamates, food dye and
saccharin. In addition, various studies have related basic foodstuffs to disease
processes; for example, the high cholesterol content of eggs has been implicated
as a factor in cardiovascular disease. Still other recent reports point to the
role of diet in various forms of cancer.
A recent PHSB study (1) suggests that occupational distribution contributes
significantly to the explanation of death from acute myocardial infarction, lung
cancer and prostatic cancer; in addition, that income is explanatory for lung cancer;
education for colon-rectum cancer; and elevation for acute myocardial infarction and
prostatic cancer. The question is, what do these variables represent. ... Is diet
an important factor?
Fortunately, North Carolina is in the position of having conducted a survey
that provides dietary data for a representative sample of the household population
(2,3). Although that survey was conducted seven years ago, and time and circumstance
have undoubtedly modified eating behavior to some extent, we believe the data are
still useful indicators of the relative dietary habits of different areas of the
state and, in any event, that they afford us the unique opportunity to examine
associations between prior dietary practice and current mortality in North Carolina.
The present study uses correlation analysis to examine dietary factors that
might be affecting age-race-sex-adjusted mortality in the state's six health service
areas (HSA's), these being the smallest areas for which survey data are available.
In these analyses, intercorrelations among per capita income, a contrived elevation
variable and nutrition factors are also examined. Although the previously cited
study (l) and other investigations tend to support the hypothesis of a protective
effect of altitude upon heart function, dietary factors may be the protective agent
in North Carol ina.
METHODS AND MATERIALS
Dietary Data
The North Carolina Nutrition Survey (NCNS) was originally designed to provide
dietary data for only three regions of the state--the East, the Piedmont, and the
West (2). However, in terms of the number of households for which dietary data
were obtained, each of the HSA's appears sufficiently represented to allow for the
post-stratification used in this paper. Table 1 compares the percentage distributions
of responding survey households and household members to the corresponding distributions
obtained in the April 1970 Census.
Details concerning the survey design and procedures have been reported (2).
Briefly, the data were collected by trained nutritionist-interviewers from an adult
household member who had responsibility for meal preparation. This person was asked
Table 1
Percentage Distributions of North Carolina Households
and Household Members by Health Service Area
Total Number
Health Service Area
Households
1970 Census 1970 Sample
Enumerat ion (^ ) Survey*
1,509, 56*4
1,160
1 Western
17.8
17.8
II Piedmont
20.1
23. k
III Southern Piedmont
17.8
12.5
IV Capital
13.2
11.0
V Cardinal
13.9
19.1
VI Eastern Carol i na
17-3
16.2
Household Members
1970 Census 1970 Sample
Enumeration^ ) Survey"
"1,893,113
3.885
17.3
16.1
19.6
21 .0
I7.A
12.1.
13.0
10.9
\k.6
21.0
18.0
18.7
*Based on households providing complete dietary data. This excludes nearly 12$ of
the original sample of 1,315 households.
to recall all foods consumed from the home food supply on the day prior to interview with
food models being used to help respondents recall the amounts of foods consumed. A
computer program was then used to convert food model specifications into gram amounts and
caloric and nutrient values.
Dietary data used in the present study are described below. These data are specific
for households located in each HSA and may be obtained by contacting the Public Health
Statistics Branch.
- Data items are the per person per meal grams of food consumed from each of 20 food groups.
Details concerning the food items included in eact food group are available (3,6).
- Data items are the per person per meal amounts of 17 nutrients consumed in survey households.
In the case of water, it should be noted that this represents the water content of foods and
mixed beverages and does not include plain drinking water.
- The survey data file provides information concerning the "adequacy" of caloric and nutrient
intakes as related to age-sex-weight-specific daily standards described in a published
report (2). Data items presently used are the percentages of households meeting less than
50% of their standard for calories and each of 8 nutrients ("low" intakes) and the percent-
ages of households meeting 200 percent or more of each standard ("high" intakes). In the
determination of these percentages, standards were adjusted for meals eaten from other than
the home source (2).
- Data items are the percentages of total calories derived from protein, fats and carbohydrates.
Factors used to convert grams to calories were *t, 9 and k respectively (e.g., each gram of
protein is equivalent to b calories).
- Each household diet was rated optimal, adequate or inadequate according to published
criteria (2). The data are used here as the percentage of household diets in each diet rat-
ing category.
- Data items are the percentages of persons failing to eat the A.M., noon and P.M. meals.
Mortal ity Data
Rates used in these analyses are the 1973-75 average annual death rates, adjusted for
age, race and sex (7) and specific for the disease entities described on the next page.
Other causes of death were not studied due to the fact that death rates are presently com-
puted only for underlying causes of death, and for many diseases, the "incidence at death"
is known to be much higher. For example, hypertension, arteriosclerosis and diabetes are
considered "associated" conditions far more often than they are considered an underlying
cause. Since our primary interest is in the determinants of disease rather than the
determinants of underlying causes of death per se, analysis based soley on underlying
frequencies would not be appropriate. We also did not study causes for which the 3_year
rates appear to fluctuate randomly over time.
Causes listed below were deemed amenable to study on the basis that (i) the under-
lying frequencies are thought to closely represent the incidence at death and (ii) cor-
relations between the 1968-70 and 1973_75 HSA death rates are statistically significant
suggesting the presence of some causative agent that may invite intervention. In the case
of acute myocardial infarction, the latter condition was met only when HSA IV was eliminated;
hence, results for that cause are based on data for five rather than six HSA's. This
concession was allowed because of special interest in intercorrelat ions among elevation,
nutrition factors and myocardial infarction.
Cause ICDA Codes (8)
Acute Myocardial Infarction ^10
Colon and Rectum Cancer 153,15^
Pancreatic Cancer 157
Trachea, Bronchus and Lung Cancer 162
Cancer of the Cervix Uteri 180
Prostatic Cancer 185
Income and Elevation Data
Clearly, food utilization practices reflect many facets of living including economic
power and culturally conditioned lifestyles; hence our need to be aware of these associa-
tions when assessing apparent associations between nutrition factors and mortality.
Per capita income (9) is used in these analyses because NCNS results indicate that
food utilization is more a linear function of income than of homemaker's education (2,3).
We chose per capita income as a general descriptor that is highly correlated with other
income variables.
The elevation variable presently used is the population-weighted average of the
elevations of all county seats located in each HSA.
Correlation Analysis
Two statistical tests— Pearson' s product-moment correlation procedure and Spearman's
nonparametric procedure— were used in these analyses (10). While results are probably
stronger when both tests indicate a statistically significant correlation, the reader should
be aware that we are dealing here with a small number of observations (HSA's) such that a
single pair of values can make a lot of difference to the value of r (Pearson's coefficient)
or rs (Spearman's coefficient). Hence, results are only suggestive, not conclusive, and in
any event, they should not be taken to imply cause-and-ef feet relationships. Rather, the
correlation coefficients merely indicate the degree and direction in which two variables
change together as described in the next section.
RESULTS
Table 2 shows statistically significant coefficients for correlations between mortality
rates and each of the nutrition factors, income and elevation respectively, and Table 3
displays significant coefficients for correlations between nutrition factors and income and
elevation respectively. In these tables, the sign of the coefficient (r or rs) indicates
the direction of the association between variables; for example, Table 2 shows that
increases in acute myocardial infarction are associated with increases in egg consumption
and decreases in the use of non-confection desserts. The closer the value of r or rs to
1.00 or to -1.00, the greater the association between a pair of variables.
Mindful of the fact that evaluation of present results must be left to the nutrition
community, remarks given below merely highlight these results while pointing up other
evidence of relationships between disease and various socio-envi ronmental factors. Here,
frequent reference is made to the diet and mortality experiences of the Japanese because
(i) nutrition conditions in Japan are documented and (ii) mortality patterns in Japan
appear to vary markedly in some instances from those observed in the United States (11).
In this regard, it should be noted that the Japanese diet is being increasingly influenced
by western culture; thus, the future of that country's mortality patterns will be of great
interest.
- Present results support other evidence of a positive association between death from cardio-
vascular disease (acute myocardial Infarction) and consumption of eggs. Here, Japan's
experience seems particularly notable since that country has one of the world's lowest
recorded heart disease rates, calculated In 1965 at less than one-fourth the U.S. and N.C.
rates. The low rate Is associated with a Japanese diet that Is low (but Increasing) In
cholesterol. (11)
- The high negative correlation between myocardial Infarction and use of desserts was not
expected but may be explained In part by the fact that dessert consumption Is positively
associated with Income and use of citrus fruits while being negatively associated with use
of eggs. Also, confection-type desserts are not Included here but In a separate sugar/
sweets group. Again referring to Japan, the Japanese enjoy and use sweet foodstuffs, but
theirs have a low sugar content compared to western confections (11).
- As before (1), elevation appears protective against death from myocardial Infarction, but
this result may be confounded by the effects of various dietary factors as well as climatic
and other elevation-related factors. The previous study (1) also suggests that water zinc
may be a positive correlate of myocardial infarction death in North Carolina, but a mechanism
for any such action is unknown and levels of water zinc appear well within established limits
in this state.
- Colon-rectum cancer is shown to be positively associated with intakes of beef and negatively
associated with intakes of flour/cereal products. Intakes of protein and niacin and the
incidence of high-niacin diets are also positive correlates. Although these results may
reflect income-related factors other than diet per se, they are consistent with a low
incidence of colon cancer in Japan where diets are extremely high in carbohydrates and low
in all animal products except fish (11). The previous study (l) suggests that colon cancer
mortality in North Carolina Is positively associated with education and with water manganese;
also, death from this cause appears more frequent among divorced nonwhlte males (12).
- Little is known of the causal aspects of pancreatic cancer, but the Increasing trend In
mortality suggests that exposure to etlologlc factors has Increased (13) or that more
accurate diagnoses are being made. Present results show a positive association with three
intercorrelated variables, i.e., the incidence of high-niacin and high-vitamin C diets and
income. Pancreatic cancer is also shown to be negatively associated with the incidence of
low-protein and low-thiamlne diets. Previous findings (12) indicate that in North Carolina
higher pancreatic cancer death rates occur among divorced and widowed persons.
- Lung cancer mortality is shown to be associated with elevation and several related dietary
factors, but previous findings (1) suggest that elevation Is unimportant while income (a
negative correlate) and occupation distribution are highly important when these factors
are considered against each other and various other social and environmental conditions.
Other studies (13) have found an inverse relationship between lung cancer and educational
class, and in North Carolina, divorced males of both races appear particularly disposed to
lung cancer mortality (12).
- Present results support other evidence of an inverse relationship between cervical cancer
and socio-economic class (12,13). Sexual activity has been Implicated (13); diet may or may
not be an important factor.
- Findings for associations between prostatic cancer and consumption of protein and thiamine
are consistent with reports of a higher frequency of prostatic cancer among Japanese
Americans than among Japanese In Japan (13); the Japanese diet traditionally Includes mostly
vegetable protein and is low In thiamine relative to dietary standards In the U.S. (11,2).
Other studies indicate a higher northern than southern frequency for U.S. whites and non-
whites and a higher frequency among professional workers, relatives of prostatic cancer
decedents, and widowed and married men, especially nonwhltes (12,13). In North Carolina,
elevation-related factors may be Important (1).
ice
they
In addition to the correlation analyses reported above, the several death rates were
correlated with an "apparent" per capita dollar sale of distilled spirits (see referenc
7, page 2M). Si nee these sales involve tourists and other non-residents of an HSA the
are only gross indicators of alcohol consumption; however, the correlation with cirrhosis
of the liver was moderately high (r = .Jh) such that results for other causes may bave
meaning Among the causes presently studied, highest correlations were observed for lung
cancer (r - .68), colon-rectum cancer (r = .50) and prostatic cancer (rs = .k3) The sale
of distilled spirits was also a moderate negative correlate of elevation (r = -.63). None
of these correlations was statistically significant.
Table 2
Significant Peorson Coefficients (r) and Spearman Coefficients (r,) for
Correlations Between Mortal Itv Kales and Nutrition Factors,
Elevation and Per Capita Income Respectively
North Carolina Health Service Areas
Nutrition Factors
Per Person Per Meal Grams
Dairy Products
Beef
Pork
Sausage/Luncheon Meats
Eggs
Legumes/Nuts
Citrus Fruits/Tomatoes
White Potatoes
Fruits/Vegetables, n.e.c.
Homemade Self-rising Biscuits
Other Flour/Cereal Products
Desserts (non-confection)
Soups /Sauces
Per Person Per Meal Nutrients
Myocardial
Infarct Ion
Calories
Protein
Ca 1 c i urn
I ron
Potass i urn
Vitamin A Value
Th I amine
Riboflavin
Preformed Niacin
VI tamin C
Water
Ash
Percent Households
ui th Low Intake of:
-.96*
80
-.90*
-.70
-70
.80
- 90*
-70
-.70
-.80
-.91' -I. 00*'
-.87
Calories
Protein
Thl amine
Riboflavin
Preformed Niacin
1 ron
Vitamin A Value
Cal c I um
Vitamin C
Percent Households
uith High Intake of:
Protein
Riboflavin
Preformed Niacin
I ron
Calcium
Vitamin C
Percent Household
Diets Rated:
72
-.80
-.70
-.70
- 80
Optimal
Adequate
Inadequate
-.70
-.90
■ 90
Percent Failure to Eat:
A.M. Heal
Noon Meal
90
.70
Percent Calories from Protein
ilevat Ion
*er Cap! ta Income
-83
-.90'
-.70
80
.70
.85 -1.00'
.87 -.80
-.70
-HSA IV not Included in these correlations.
Fruits/Vegetables, not elsewhere classified,
and white potatoes.
Colon-Rectum
Cause of Death
.814* .83'
Pancreat Ic
Lung
Cancer
.80
-.94** -.83^
• 83<
.85* .83'
.92** .83
-■75 -.77
.77
.77
74
.75
.81)*
80
.74
.95** .83'
76
.77
.83*
.82*
.87* -.83*
.82* -.78
.74
-.77
.74 .77
78 .77
• 85*
.77
.77
75
80
Cervical
Cancer
78
.75
S3*
Prostatic
Cancer
77
83*
.84*
97*
-77
-I .00*
-.77
.78
.92** -.83*
71
.76
90*
.77
.77
.76 .77
.80
.7<i
.77
-.89'
.90* -.84*
-.77
77
89«
-.88* -.77
.88* -.94*
-.98** -1.00*
.83*
•89* .77
95*
.94*
I.e., excludes citrus fruits, tomatoes, high vitamin A fruits/vegetables
no superscript p
* P
. 10. two-tal led test.
.05. two-tailed test.
.01 , two-tal led test.
CONCLUSION
Although the preceding dietary findings may reflect the effects of other factors,
perhaps they still should not be dismissed lightly for the following reason. Many
processed foods have added nutrients, including niacin, thiamine and others. For some
of these products, levels of the nutrients are regulated by Federal standards, but for
many products, there are no regulations (}k) . And even for those products subject to
regulation, one must wonder if the standards have been in effect long enough to reliably
discern the long-term effects. We note, for example, that enrichment of wheat products
was initiated during World War II (15) and enrichment laws for other grain products are
;j li.. itc\
considerably younger (16)
Table 3
r
r
s
7li
.77
/I
.83*
-.6k*
83*
91*
.9A**
95**
.9<4**
78
• 77
88*
-.83*
95**
.89*
97**
1.00**
Significant Pearson Coefficients (r) and Spearman Coefficients (rs)
for Correlations Between Nutrition Factors and Income
and Elevation Respectively
Six North Carolina Health Service Areas
Nutrition Factors Correlated
with Per Capita Income
Per Meal Intake of Desserts
Per Meal Intake of Citrus Fruits/Tomatoes
Per Meal Intake of Mixed Foods
Per Meal Intake of Thiamine
Per Meal Intake of Preformed Niacin
Per Meal Intake of Vitamin C
Percent Households with Optimal Diet
Percent Households with Inadequate Diet
Percent Households with Low Iron Intake
Percent Households with Low Vitamin C Intake
Percent Households with High Preformed Niacin Intake
Percent Households with High Vitamin C Intake
Nutrition Factors Correlated
with Elevation
Per Meal Intake of Citrus Fruits/Tomatoes .77
Per Meal Intake of Fruits/Vegetables n.e.c.
Per Meal Intake of Fiber Carbohydrates
Per Meal Intake of Calcium
Per Meal Intake of Potassium
Per Meal Intake of Thiamine
Per Meal Intake of Ash
Percent Households with Low Calcium Intake
Percent Households with Low Vitamin A Value Intake
Percent Households with High Riboflavin Intake
Percent Households with High Iron Intake
Percent Households with High Calcium Intake
Percent Households with Adequate Diet
Percent Households with Inadequate Diet
Percent Persons Failing to Eat A.M. Meal
no superscript p < .10, two-tailed test.
* p<.05, two-tailed test.
** p<.01, two-tailed test.
80
8k*
78
.77
9k**
.89*
-.75
88*
• 77
-.83*
79
89*
.89*
79
80
88*
.9it**
76
-.Sk**
-.89*
Note: Persons desiring coefficients for correlations among the nutrition factors
may contact the Public Health Statistics Branch.
STATE LIBRARY OF NORTH CAROLINA
3 3091 00739 0230
One final word concerns indicators of nutritional status which have not been studied
here. These include the use of sugar vs. starch carbohydrates, animal vs. vegetable
protein and saturated vs. unsaturated fats. While the NCNS statistical file does not
identify these constituents, it does identify specific food items such that, with the aid
of nutritionists, we might estimate the needed parameters in order to perform correspond-
ing analyses. Some present results point to a need for these more detailed analyses; for
example, contrary to expectation, correlations between dietary fat and the diseases studied
here were not significant, but correlations involving saturated vs. unsaturated fats may be
quite different.
REFERENCES
(1) N.C. Department of Human Resources, Division of Health Services, Public Health
Statistics Branch. "Associations Between Mortality and Various Social,
Economic and Environmental Factors in North Carolina," PHSB Studies, No. 3,
Raleigh, April 1977.
(2) N.C. State Board of Health. North Carolina Nutrition Survey: Part One. Raleigh,
July 19, 1971.
(3) N.C. Department of Human Resources, Division of Health Services. North Carol i na
Nutrition Survey: Part Two. Raleigh, December 1 97 ** -
CO U.S. Bureau of the Census. Current Population Reports. Series P-26, No. 75"33,
U.S. Government Printing Office, Washington, D.C., June 1976.
(5) U.S. Bureau of the Census, Census of Population. 1970 General Population Charact-
eristics. Final Report PC ( 1 ) - B35 , North Carolina, U.S. Government Printing
Office, Washington, D.C., 1971.
(6) N.C. Department of Human Resources, Division of Health Services, Public Health
Statistics Branch. Food Choices of North Carolinians by Kathryn B. Surles
and James J. Palmersheim. SRS-22 10-07-02-73 , Raleigh, November 15, 1973-
(7) Department of Human Resources, Division of Health Services, Public Health Statistics
Branch. Leading Causes of Mortality, North Carolina Vital Statistics 1 973~75 ,
Vol. 2. Raleigh, July 1976.
(8) U.S. Department of Health, Education and Welfare, Public Health Service, National
Center for Health Statistics. Eighth Revision International Classification
of Diseases, Adapted for Use in the United States. Public Health Service
Publication Number 1693. U.S. Government Printing Office, Washington, D.C.,
December 1968.
(9) N.C.
(10)
(11)
Department of Human Resources, State Health Planning and Development Agency.
Selected Health Status and Demographic Data for North Carolina. Raleigh,
December 1976.
Nie
N.H.; Hull, C.H.; Jenkins, J.G.; Steinbrenner , K. ; and Bent, D.H. Statistical
Package for the Social Sciences, Second Edition. McGraw-Hill: New York, 1975.
Insull, William Jr.; Oiso, Toshio; Tsuchiya, Kenzaburo. "Diet and Nutritional
Status of Japanese," The American Journal of Clinical Nutrition. Vol.
No. 7, pp. 753-777, July 1968.
21
(12) N.C. Department of Human Resources, Division of Health Services, Public Health
Statistics Branch. "Relationships Between Marital Status and Mortality in
North Carolina," PHSB Studies, No. 2, Raleigh, February 1 977 -
(13) Lilienfeld, A.M.; Levin, M.L.; Kessler, I.I. Cancer in the United States. Harvard
University Press, Cambridge, Massachusetts, 1972.
OM U.S. Department of Agriculture. Food for Us All: The Yearbook of Agriculture 1969.
U.S. Government Printing Office, Washington, D.C., 1969-
(15) Mitchell, H.S.; Ryanbergen, H.J.; Anderson, L. ; Dibble, M.V. Cooper's Nutrition in
Health and Disease, 15th edition. J.B. Lippincott Company, Philadelphia, 19687
(16) U.S. Department of Agriculture. Composition of Foods byBerrticeK. Watt and Annabel
L. Merrill. Agriculture Handbook No. 8 (rev.), U.S. Government Printing Office,
Washington, D.C., December 1963.
Public Health Statistics Branch
Division of Health Services
Department of Human Resources
P.O. Box 2091
Raleigh, North Carolina 27602