Professional Documents
Culture Documents
Current Research
C
omprehensive, regular, country-wide dietary moni-
sumption was similar in both sexes at 17% protein, 34% toring is not a consistent feature of health surveys in
to 35% fats, and 48% to 49% carbohydrates, but differed Canada. Summary nutrition indicators, in conjunc-
by age group in women. Adequacy of micronutrient in- tion with a regular nutrition surveillance program, could
takes relative to Dietary Reference Intakes varied by age, be useful for rapid screening of diet quality, thereby con-
tributing to effective, targeted nutrition interventions.
This would be particularly helpful in population sub-
B. Shatenstein and G. Ferland are research scientists groups at risk of nutrition-related chronic disease.
with the Centre de recherche at the Institut universitaire Evaluation of a complex entity like diet requires assess-
de gériatrie de Montréal, Montréal, Canada, and associ- ment of the frequency and quantity of foods eaten, food
ate professor and full professor, respectively, with the combinations, cooking practices, and dietary variety. Al-
Département de nutrition, Université de Montréal, Can- most 3 decades ago, it was suggested that diet quality
ada. S. Nadon and C. Godin are research assistants indicators based on nutrient intakes, foods or food group-
with the Centre de recherche at the Institut universitaire ings, or a combination of nutrients and foods (1) could be
de gériatrie de Montréal, Montréal, Canada. used to quickly evaluate survey results and the efficacy of
Address correspondence to: Bryna Shatenstein, PhD, dietary programs (2,3). Indeed, it has been stated that by
Centre de recherche, Institut universitaire de gériatrie describing global dietary quality these indexes communi-
de Montréal, 4565 Queen Mary, Montréal, QC Canada cate more about dietary adequacy and disease risk than
H3W 1W5. E-mail: bryna.shatenstein@umontreal.ca the intake of single nutrients or foods (1,4,5). Diet quality
Copyright © 2005 by the American Dietetic may thus be evaluated in relation to nutrient adequacy
Association. (3), compliance with dietary guidelines (6-8), or nutrition
0002-8223/05/10508-0007$30.00/0 recommendations (ie, adequate nutrient intakes, fat and
doi: 10.1016/j.jada.2005.05.008 carbohydrate balance, dietary variety, and control of salt
© 2005 by the American Dietetic Association Journal of the AMERICAN DIETETIC ASSOCIATION 1251
and alcohol) (8,9), association with risk of chronic dis- METHODOLOGY
eases (ie, obesity, cancer, and heart disease) (6,7,10), or Instrument Development and Sampling
mortality (8,11,12), and used to assess interventions (13). The development and assessment of relative validity of
Examining the indicator’s component subscores could al- the FFQ as well as sample selection are described in
low health professionals to target specific dietary prob- detail elsewhere (25). Briefly, a 73-item self-adminis-
lems. Furthermore, when repeated over time, scores can tered, semiquantitative FFQ was developed in French
track changes in overall diet quality and compare popu- and English for Quebec adults, to assess usual food con-
lation subgroups, giving health professionals valuable sumption during the previous 12 months. The FFQ was
insight for planning nutrition interventions. modified from the Health Habits and History Question-
A nutrition indicator reflects the researchers’ a priori naire, available at the time with diet analysis software
definition of diet quality, the elements included in its (DietSys) in the public domain from the National Can-
calculation, and the source of the dietary data used to cer Institute Website (www.dccps.ims.nci.nih.gov/ARP/
compute the score. Whereas most diet quality indicators DietSys/uguide.html, accessed June 1, 2000). A list-
have been calculated from (multiple) food recalls or based approach (26) drawing on the most recent Quebec
records, some investigators have used food frequency population dietary data [the 1990 Québec nutrition sur-
questionnaires (FFQs) to develop and test their scores vey (27)] was used to build the food list and determine
(10,14). This approach to assessing usual diet quality consumption frequency options and food portion sizes. No
could be more meaningful than recalls or records as it has information is collected on supplement use. When neces-
been shown that the intra- and interperson variability of sary, the Canadian Nutrient File (28) was used to modify
indicators depends on the nature of the dietary data and the USDA food composition database supplied with Diet-
on the number and timing (ie, consecutive vs nonconsec- Sys to reflect Canadian food composition data. The FFQ
utive) of replicate dietary measurements used to generate was pretested among volunteers using a qualitative cog-
the scores (5). Interestingly, the variance ratios also de- nitive interview approach as described by Subar and col-
pend on certain sociodemographic, lifestyle, and health- leagues (29). A revised version was submitted to a second
related characteristics of the population, which are in pretest and the instrument was finalized for testing (ie,
turn known to influence the reliability and accuracy of feasibility, acceptability, time, and cost) and validation.
diet quality measures (5). It is generally agreed that The FFQ was tested in a sample of 248 adults (a target
socioeconomic status, age, and sex are consistent deter- sample size of 250 was selected to detect significant dif-
minants of food choice (15,16) and hence diet quality ferences in diet quality and provide sufficient numbers to
(8,17) and nutritional adequacy (18,19), so this is not a permit subgroup analyses) who were recruited by random
surprising observation. digit dialing using a computer-generated list covering the
The US Department of Agriculture (USDA) developed local calling area of metropolitan Montreal, which has a
the Healthy Eating Index (HEI) as a measure of overall population base of 3 million. Birthdate and sex of all
diet quality to evaluate how well Americans meet dietary household members aged 18 years or older were ascer-
guidance (9,20). It assesses the extent of compliance with tained and listed in descending order, and one person was
the Food Guide Pyramid (21), as well as several aspects of recruited per household using a random number proce-
the Dietary Guidelines for Americans (22). The HEI con- dure. Eligible subjects were aged 18 years or older, able to
read and write French or English, and living in a private
sists of 10 components, each worth 10 points. The first
household with no intention to move during their study
five components are based on consumption of foods from
participation. The study was conducted from the Institut
the five groups of the Food Guide Pyramid. The other four
universitaire de gériatrie de Montréal and approved by
components consider intakes of fats, saturated fats, cho-
the Ethics Review Board. All participants provided in-
lesterol, and sodium, as in the Dietary Guidelines for
formed consent before data were collected. Recruitment
Americans (22), which are similar to Canadian guidance began in November 2000 and ended in August 2001.
(23). The final component evaluates dietary variety. An
HEI score ⱕ50 indicates a poor diet, 51 to 80 signifies a
diet that needs improvement, and ⱖ81 reflects a good diet Data Collection and Instrument Assessment
(20). In their examination of different diet quality indexes Data collection was carried out by mail with telephone
applied to data from a Quebec nutrition survey, Dubois follow-up as FFQs were received to clarify responses or
and colleagues (24) determined that the HEI is an appro- complete missing information. The validation subsample
priate indicator for assessing the diet quality of Quebec- consisted of 94 persons from the FFQ development sam-
ers. ple who agreed to be recontacted. They were given de-
To encourage regular population diet monitoring in tailed instructions for completing their food records and
Canada, we developed and assessed the relative validity provided four 1-day, nonconsecutive food records by tele-
of a population-based FFQ to assess usual diet in Que- phone (61%), fax (18%), and E-mail (21%) during a
becers (25). As a global diet quality indicator calculated 3-week period following completion of the FFQ, with tele-
from the FFQ could provide a focus for nutrition health phone follow-up where necessary as they were received.
promotion and effective targeted interventions, we also This prospective method was used because it is indepen-
adapted the HEI to Canadian guidance and evaluated dent of memory, which eliminates difficulties in assessing
this score (the Canadian HEI) for assessing overall diet validity due to a similar error structure between test and
quality from data gathered by the FFQ. This article de- reference instruments (30). Data collection and handling
scribes the adaptation and evaluation of the diet quality were done by two bilingual (ie, French and English) re-
indicator and its application in an adult population. search dietitians.
Adaptation and Evaluation of the Canadian HEI data. The total score and the distribution of its compo-
The Canadian HEI was adapted to Canadian dietary and nents were examined. It was validated using Spearman
nutritional guidance using Canada’s Food Guide for correlation analyses (due to non-normal distribution of
Healthy Eating (31) and Nutrition Recommendations for the data) assessing agreement between nutrition and
Canadians (23) (see Figure). It has nine components, each dietary recommendations and the total score and selected
worth 10 points, except for the vegetables and fruits components of the score, by sex and age group. Nutrient
subscore (to reflect that combined food group in Canada’s recommendations were taken from the Dietary Reference
Food Guide for Healthy Eating), which has a maximum of Intakes (32) except for energy and protein, which were
20 points. The final component, dietary variety, was mod- not available at the time of analyses; vitamin A, due to
ified from the original HEI because that tool was devel- changes in determining retinol equivalents that could not
oped for diet recalls. The HEI variety subscore is based on be applied to our nutrient database; and folate, as our
the five food groups of the USDA Food Guide Pyramid nutrient database predated folate fortification of grain
(21) and counts the total number of different foods eaten products. These latter recommendations were taken from
during the 3 days of data collection contributing substan- the 1990 Canadian Recommended Nutrient Intakes (23).
tially to meeting one or more of the five food group require- Participants were ranked into quartiles by sex. This was
ments (9). In the FFQ setting this was not possible because done to examine the progression of the Canadian HEI
the FFQ provides a summary indication of usual food in- subscores as the total score improved. All statistical anal-
takes over a 12-month period. The variety subscore was yses were conducted using SPSS version 10.
thus made more reflective of Canada’s Food Guide for
Healthy Eating recommendation for variety using a variant
of the Dietary Diversity Score (18). Like the USDA HEI, the RESULTS
Canadian HEI has a maximum score of 100 points. The study sample is presented in Table 1. Two hundred
forty-eight adults aged 18 to 82 years adequately com-
pleted an FFQ. The response rate to the FFQ was 64%
Analyses (257 of 400 questionnaires mailed). Women outnumbered
FFQ dietary data were analyzed using DietSys, which men (57%). Demographic characteristics of the validation
had been modified to reflect our food list and order, por- subgroup resembled those of the full FFQ sample. A little
tion sizes, and nutrient values. A programming algorithm more than 80% were French speakers and the remainder
was written in SPSS language (version 10, 1999, SPSS spoke English, which mirrors the linguistic composition
Inc, Chicago IL) to generate the Canadian HEI from FFQ of Quebec’s population. More than 40% had a university
Table 3. Subscores and total Canadian adaptation of the Healthy Eating Index (C-HEI) (9) score calculated from self-administered food frequency
questionnaire data collected from Montreal-area residents, by sexa
Highest possible Male Female All
C-HEI components score (nⴝ105) (nⴝ143) (Nⴝ248)
total Canadian HEI was significantly associated with ergy intakes of fat (rs⫽⫺0.34 to ⫺0.46, P⬍.001) or satu-
recommendations addressing percent energy as carbohy- rated fat (rs⫽⫺0.25 to ⫺.0.36, P⬍.001).
drates (rs⫽0.64, P⬍.001) and fiber intakes (rs⫽0.43,
P⬍.001). On the other hand, subscores including Cana-
da’s Food Guide for Healthy Eating and percent energy as DISCUSSION
fat (with or without the sodium subscore) were signifi- The purpose of a dietary index is to synthesize a large
cantly inversely related to participants’ proportional en- amount of food and nutrient consumption information
Table 5. Association between Canadian adaptation of the Healthy dated FFQ tested among 248 adults aged 18 to 82 years
Eating Index (9) and selected nutrients estimated by the self- living in metropolitan Montreal, Canada (25).
administered food frequency questionnaire completed by Montreal- The Canadian HEI calculated from our FFQ showed
area residents, by sex that a relatively small proportion of our sample—approx-
Correlations (rS)a imately 20% of men and 29% of women— had a “good”
diet, defined as a score ⬎80. In fact, 76% of men and 70%
Nutrient Male Female of women scored between 51 and 80, indicating that their
diets need improvement, and 2.4% overall (3.8% of men
Energy (kcal) 0.07 0.03 and 1.4% of women) scored under 51, which reflects a
Protein (g) ⫺0.20* 0.05 “poor” diet. The overall mean Canadian HEI scores in our
Vitamin A (REb) ⫺0.20* 0.13 sample, 70.3 in men and 74.9 in women, were higher than
Vitamin C (mg) 0.47*** 0.54*** HEI results based on the Continuing Survey of Food
Thiamin (mg) 0.16 0.34*** Intakes by Individuals, 1994-96, published for Americans
Riboflavin (mg) ⫺0.06 0.16 in 1998 (20) and than those from the 1997 Michigan
Niacin (NEc) ⫺0.14 0.07 Behavioral Risk Factor Surveillance System (8), where
Vitamin B-6 (g) 0.13 0.28*** mean HEI and Michigan Healthy Diet Indicator scores
Vitamin B-12 (g) ⫺0.26** ⫺0.02
were 63.8 and 55.7, respectively. Caution should be exer-
Folate (g) 0.37*** 0.46***
cised when comparing across instruments and popula-
Calcium (mg) ⫺0.12 0.05
tions. However, as others have observed (8,20), women in
Iron (mg) 0.03 0.25**
our study scored significantly higher on the HEI than
Zinc (mg) ⫺0.21* 0.05
Fiber (g) 0.36*** 0.40*** their male counterparts. This was particularly evident for
Total fat (g) ⫺0.36*** ⫺0.22** certain subscore components, most notably in the vege-
Saturated fat (g) ⫺0.42*** ⫺0.26** tables and fruit category.
Cholesterol (mg) ⫺0.36*** ⫺0.17* The Canadian HEI did not show consistent relation-
Sodium (mg) ⫺0.18 0.05 ships with individual nutrients. Kennedy and colleagues
(9) reported positive, albeit modest and variable, correla-
a
Spearman correlation coefficients. tions ranging from 0.06 to 0.42 between the HEI and a
b
RE⫽retinol equivalents. series of nutrients. Our results varied by nutrient and
c
NE⫽niacin equivalents. sex. In general, nutrients from vegetables and fruit (ie,
*P⬍.05.
folate, vitamin C, and fiber) and from high-fiber foods
**P⬍.01.
***P⬍.001.
were strongly associated with the Canadian HEI,
whereas nutrients from animal-source foods were poorly
correlated with the Canadian HEI. This supports earlier
analyses on another indicator, the Diet Quality Index-
into a single useful indicator (33). The USDA HEI was Revised, which had shown that higher Diet Quality In-
constructed to provide a measure of diet quality for mon- dex-Revised scores were related to lower fat consumption,
itoring consumption patterns and their changes, and higher intakes of fruit and vegetables, and better intakes
serve as a tool for nutrition education and health promo- of foods furnishing iron and calcium (34). It is noteworthy
tion (9). We adapted the HEI to Canadian dietary guid- that in our sample, negative correlations between the
ance (Canadian HEI) and used it to evaluate dietary Canadian HEI and individual nutrients were primarily
adequacy based on usual intakes estimated from a vali- observed among men, and that several of these nutrients
are derived from intakes of foods from the meat and meat same populations of male and female health professionals
alternatives food group. These nutrients are also associ- as in their previous reports (6,7). They found a relation-
ated with foods providing substantial intakes of fat, sat- ship between attainment of the alternate HEI goals and
urated fat, and dietary cholesterol. In fact, in our FFQ reduced chronic disease risk, most notably due to reduc-
validation study we found that among men, poor correla- tion in cardiovascular disease risk, and they attributed
tions between protein and niacin from the FFQ and the this finding to the ability of the alternate HEI to capture
reference method (four nonconsecutive food records) certain dietary choices (eg, white vs red meat and type of
could be traced to FFQ overestimates for meat and meat fat consumed). However, they observed that it is difficult
alternatives, which ranged from 18% in men older than to target a score to chronic diseases that differ in their
age 65 years, to 36% in men aged 18 to 34 years. Such etiology and postulated that measurement error and ab-
observations suggest that a sex-based perceptual bias sence of certain details (eg, high-heat cooking practices
may confound reported intakes of certain foods, in this related to cancer etiology) could obscure relationships.
case, meats. Since assessment of the FFQ’s internal va- It was for this reason that we sought to evaluate the
lidity and its potential to represent usual diet revealed no Canadian HEI in a general dietary surveillance context
substantial omissions of regularly consumed foods (25), rather than from the perspective of disease risk. Our
variations in strength of association between the Cana- findings show that selected subscores of the Canadian
dian HEI and various nutrients could thus be attribut- HEI do indeed reflect achievement of these dietary guide-
able to reported food sources of the nutrients and the lines. Canadian HEI subscores indicating adherence to
relationship between these foods and component sub- dietary guidelines for fiber, fat, and saturated fat showed
scores of the Canadian HEI, as well as to differing intakes
robust associations with subjects’ reported intakes of
of foods among men and women.
these dietary substances on the FFQ, suggesting that the
Data from FFQs have been used by others to construct
Canadian HEI is capable of discriminating diet quality.
measures of overall diet quality based on dietary variety
We were initially unsure if we could validate the Ca-
(14). These investigators assessed adherence to Food
nadian HEI because it was not clear which reference
Guide Pyramid recommendations and validated their
food categories in relation to estimated intakes of nutri- method should be used. It was therefore reassuring to
ents associated with chronic disease risk, and the extent note that others have used a set of plasma biomarkers of
of compliance with (current) dietary guidance. The mea- dietary exposure to validate HEI scores, and that signif-
sures were found to be associated with several attributes icant correlations have been found between the HEI and
of diet quality, and compared favorably with the HEI. certain circulating plasma carotenoids, as well as plasma
However, other studies using the HEI in the FFQ context vitamin C concentrations (37). In addition, when the Diet
have had mixed results. McCullough and colleagues (6,7) Quality Index-Revised was validated using a 131-item
reported that dietary patterns determined in two large FFQ, it performed similarly to the HEI as reported in the
US populations of female (35) and male health profession- previous study (37) when compared with plasma bio-
als (36), which signaled adherence to the US Dietary chemical measurements (38). Finally, Neuhouser and col-
Guidelines (reflected by high HEI scores), showed little leagues (39) assessed the reproducibility and validity of
relationship to risk of developing major chronic diseases, the Diet Quality Index-Revised using the Women’s
although risk reduction related to following the dietary Health Initiative FFQ (40) and reported that selected
guidelines was more apparent in women than men. These concentration-based nutritional biomarkers do predict
authors proposed that the absence of clear findings could diet quality as assessed by the Diet Quality Index-Re-
be due to the fact that the HEI was developed for use with vised. These findings attest to the potential validity of a
24-hour recalls. More recently, McCullough and col- diet quality index obtained from FFQ dietary estimates.
leagues (10) developed an alternate HEI that targets food However, it is worth noting the possible discordance be-
choices and macronutrient sources specifically associated tween the FFQ usual nutrient estimates and the shorter
with reduced risk for chronic disease, and tested it in the timeframe provided by plasma biomarkers. More work