European Journal of Clinical Nutrition (1997) 51, 297±301 ß 1997 Stockton Press. All rights reserved 0954±3007/97 $12.00 does not necessarily distort dietary surveys T Hirvonen, S MaÈnnistoÈ, E Roos and P Pietinen National Public Health Institute, Department of Nutrition, Helsinki Objectives: To study the magnitude of and trends in energy underreporting and to compare food consumption, nutrient intake and socioeconomic characteristics of underreporters to those of other Finnish adults. Design: Cardioavscular risk factor surveys in 1982 and 1992 using a 3 d food record. Underreporting was de ned as energy intake lower than 1.27*BMR, since energy intake < 1.27*BMR is improbable. Setting: Four areas in Finland, both rural and urban. Subjects: 1746 men and 1921 women, aged 25±64 y. Results: Proportion of underreporters has increased from 33% in 1982 to 46% in 1992 among women and from 27% in 1982 to 42% in 1992 among men. In a logistic regression model, BMI over 25 kg/m 2, female gender, age over 45 y and high educational level predicted underreporting. Shares of energy intake from fat, carbhoydrates, protein and alcohol remained the same whether or not underreporters were excluded. However, underreporters consumed signi cantly higher proportion of vegetables, sh, meat, potatoes, fruit and berries and less fat than others. In the 1992 data the absolute intake of most micronutrients increased and micronutrient densities decreased when underreporters were excluded. Conclusions: The proportion of underreporters has grown from 1982 to 1992. Results expressed as a percentage of energy intake are not affected by the exclusion of underreporters. In contrast, micronutrient intakes, both absolute and energy density values, were distorted by underreporting. Underreporting should be taken into account in future studies. Sponsorship: National Public Health Institute. Descriptors: dietary surveys; energy underreporting; food consumption; nutrient intake Correspondence: Dr P Pietinen, National Publich Health Institute, Department of Nutrition, Mannerheimintie 166, FIN-00300 Helsinki, Finland. Received 13 October 1996; revised 2 January 1997; accepted 7 January 1997 Introduction Underreporting is a common cause of recall bias in nutritional epidemiology. Until recent years, however, underreporting has not been of major concern. Recent studies have shown that underreporting is a large problem and might distort dietary studies (Black et al, 1991; Smith et al, 1994; Pryer et al, 1995; Ballard-Barbash et al, 1996). However, different methods and cut-off points have made it dif cult to compare the magnitude of underreporting between studies. The general conclusion has been that women and overweight people underreport more often than others. Underreporting seems to be the smallest in studies in which energy expenditure was measured in a controlled trial or in some other study design that requires high degree of co-operation from the participants (HaralsdoÂttir and SandstoÈm, 1994). Usually these people are highly motivated, have a higher educational level and lower BMI than the average population. It might be that higher involvement in a study makes people less susceptible to underreporting than lower involvement in survey-type studies (Smith et al, 1994; Ballard-Barbash et al, 1996). Therefore underreporting might be a greater problem in population surveys than in experimental studies. The aims of our study were to examine the magnitude of underreporting in the Finmonica study in 1982 and 1992 and its association with food and nutrient intake and to nd out predictors of underreporting. Subjects and methods This study is based on two dietary surveys carried out in 1982 and 1992. The conduct of these surveys has been described in detail elsewhere (Uusitalo et al, 1987; Kleemola et al, 1994). The 1982 study was carried out in three areas: the provinces of Northern Karelia and Kuopio in eastern Finland and in the city of Turku and Loimaa rural area in southwestern Finland. The 1992 survey included also the Helsinki-Vantaa area. For each survey an independent random sample of men and women aged 25±64 y was drawn from the population register. The sample size was 250 men and 250 women in each ten-year age group and area. The subjects were rst invited to the local health centre for an examination and then participated in the dietary survey. In 1982 the participation rate was 80% and in 1992 it was 76%. All subjects born between the 7th and 12th (1982), or between the lst and 11th (1992) day of each month were chosen for the subsample of the dietary survey. The diet subsample was 2267 people in 1982 and 2822 people in 1992.
298 Study participants were instructed to keep a food record of all food consumed over a 3 d period, starting the following morning. The records were kept on blank forms like a diary. A 63-page picture booklet was used in estimating portion sizes (Pietinen et al, 1988). Completed food records were sent back by mail. 87% of those instructed to keep the records completed the work in 1982 and 82% in 1992. Thus the nal response rate was 60% (653 men and 695 women) in 1982 and 66% (870 men and 991 women) in 1992. The food records were processed using the software and the food and nutrient database developed at the National Public Health Institute (Ovaskainen et al, 1996). Both 1982 and 1992 surveys were used in studying macronutrient shares and predictors of underreporting. In those analyses, Helsinki-Vantaa area was excluded from the data. The whole 1992 survey was used in the analyses concerning micronutrient intake and food densities. Weight and height were measured by trained nurses at the examination. Body mass index (BMI) was computed as weight (kg) divided by the square of height (m 2 ). Educational levels were formed by dividing subjects into tertiles within each birth year based on their total years of education. Basal metabolic rate (BMR) was estimated using WHO equations which take into account weight, age and sex (World Health Organization, 1985). Underreporters were identi ed using the cut-off value of 1.27*BMR. It has been shown that among free-living persons energy expenditure levels below 1.27*BMR are rare (Goldberg et al, 1991). People whose energy intake were below this cut-off point were de ned as underreporters. Table 1 Predictors of underreporting (energy intake < 1.27*BMR). The model includes all the variables simultaneusly Variable Odds ratio 95% CI Year, 1982 1.00 1992 1.85 1.57±2.19 Gender, Male 1.00 Female 1.48 1.25±1.75 Age, 25±34 1.00 35±44 1.06 0.82±1.36 45±54 1.65 1.28±2.11 55±64 1.46 1.14±1.86 Area, North Karelia 1.00 Kuopio district 1.12 0.91±1.37 Southwest 1.55 1.27±1.89 Educational level, low 1.00 medium 1.19 0.91±1.38 high 1.25 1.02±1.54 BMI, < 25 kg/m 2 1.00 25 < BMI < 30 2.04 1.69±2.47 > 30 3.72 2.93±4.72 Statistical methods Statistical analyses were made using SAS programs (SAS Languages and Procedures, 1989). A logistic regression model was tted to assess predictors of underreporting. T- test was used in between-group comparisons of means. Results In 1982, 33.4% of women and 26.7% of men were underreporters. The respective numbers were 46.4% and 41.7% in 1992. Thus underreporting has become more common since 1982, especially among men. Women underreport more frequently than men. In a logistic regression model, year of study, gender, age, area, education and BMI were strong independent predictors of underreporting (Table 1). Of these, BMI was the strongest factor. Among those whose BMI was over 30 kg/m 2 underreporting was twice as common as among those with BMI 25 kg/m 2 or less. (Figure 1) Older people underreported more often than young people. However, the association with age was not uniform over gender and study year. In addition, underreporting was more common in western Finland (Turku-Loimaa) than in North Karelia and among highly educated people (Table 1). Energy intakes for women and men in 1982 and in 1992 are presented in Table 2. The shares of macronutrients from energy differed very little depending on whether or not underreporters were excluded (Figure 2). The proportion of energy from fat decreased signi cantly from 1982 to 1992 and that of protein, carbohydrates and alcohol increased. Figure 1 Proportion of underreporters by BMI, year and gender. When underreporters were excluded energy intake from saccharose increased signi cantly from 1982 to 1992. Differences in the micronutrient intake in 1992 are more apparent (Tables 3 and 4). The absolute intake of almost every nutrient increased when underreporters were excluded. These differences were statistically signi cant, except for vitamin A among men and thiamin among women. Among women the intake of zinc rose from below recommendations level to above level. The situation changed radically when the intakes were calculated per MJ. The intake of every nutrient, except for potassium decreased. Also these differences were statistically signi cant, except calcium intake among men, vitamin E and thiamin intakes among women and vitamin A, vitamin D and ribo avin intakes among both genders. Figure 3 presents the food consumption densities (g/mj) of underreporters compared to the others in 1992. Both male and female underreporters consumed sign ciantly higher proportion of vegetables, sh, meat, potatoes, fruit and berries than the others. The consumption of cheese and bread among women was statistically different between underreporters and others but not in men. On the other hand underreporters consumed less fats used in cooking and baking and as spreads and salad dressing. There were no statistically signi cant differences between underreporters and others in the consumption of sausages, meat, milk, soft drinks, sweets and ice cream.
Table 2 Energy intake (MJ) in 1982 and in 1992 299 1982 1992 All Energy intake Energy intake All Energy intake Energy intake < 1.27*BMR 1.27*BMR < 1.27*BMR 1.27*BMR Men 11.77 (n ˆ 653) 8.18 (n ˆ 174) 13.04 (n ˆ 479) 10.56 (n ˆ 876) 7.82 (n ˆ 365) 12.52 (n ˆ 511) Women 8.56 (n ˆ 695) 6.30 (n ˆ 232) 9.71 (n ˆ 463) 7.86 (n ˆ 991) 6.10 (n ˆ 460) 9.38 (n ˆ 531) All 10.11 (n ˆ 1348) 7.09 (n ˆ 406) 11.42 (n ˆ 942) 9.11 (n ˆ 1867) 6.85 (n ˆ 825) 10.90 (n ˆ 1042) Figure 2 Proportion of macronutrients of total energy intake. t-test for differences between years 1982 and 1992, when underreporters were excluded, *P < 0.05, **P < 0.01, ***P < 0.001. Discussion The proportion of underreporters was one of the largest among dietary surveys (46% for women and 42% for men in 1992). The growing proportion of underreporting from 1982 to 1992 was somewhat expected because of increasing health consciousness (Karisto et al, 1993). In previous studies the proportion of underreporters vary between 19% and 52% (Lichtman et al, 1992; Ballard-Barbash et al, 1996). In a study conducted in Australia 29% of people were underreporters (Smith et al, 1994). The data were collected by a food frequency questionnaire. In a British study using 7 d weighed record 30% of men and 47% of women were underreporters (Pryer et al, 1995). In a Danish study, which used dietary history, estimated mean energy expenditure was 13.5 MJ for men and 9.4 MJ for women and respective energy intakes were 10.2 MJ and 7.3 MJ (Heitmann and Lissner, 1995). Ballard-Barbash et al (1996) compared energy expenditure estimated from anthropometric data to intake of recall of four nonconsecutive days. All participants were women. They found that 52% of women were underreporters. However, comparison between studies is dif cult since methods are different. Also cut-off points vary between 1.06*BMR and 2.10*BMR. It can be argued that the cut-off point for underreporting in our study was too low (1.27*BMR), because in this level physical activity is equivalent to lying on a bed. There could be a problem in using the same cut-off point in 1982 and in 1992, because the physical activity level should also be the same. In Finland, energy expenditure during work and during moving to and from work have decreased from 1982 to 1992 both in men and women. Because of increased leisure time exercise, the total physical activity level in 1992, however, was only 0.6% lower among women and 3.0% lower among men than in 1982 (Fogelholm et al, 1996). Energy expenditure was estimated from anthropometric data in all these previous studies. Another approach is to compare reported intake and measured energy expenditure. HaraldsdoÂttir and SandstoÈm (1994) compared energy expenditure in a strictly controlled trial and calculated intake from a 7 d weighed record. All subjects were of normal body weight. Reported intake was only slightly lower than measure energy. In our study underreporting was more common among overweight than among people with normal body weight. This is in accordance with previous studies (Prentice et al, 1986; HulteÂn et al, 1990; Heitmann and Lissner, 1995). Ballard-Barbash et al (1996) reported an inverse association between BMI and self reported energy intake. The study of Lissner et al (1989) makes an exception. They Table 3 Mean micronutrient intake among men in 1992 according to energy intake (EI) levels All EI < 1.27*BMR EI > 1.27*BMR a Recomendations b Nutrient Total /MJ Total /MJ Total /MJ Total Iron, mg 17.4 1.70 13.5 1.74 20.4*** 1.66* 10 Calcium, mg 1200 115 914 116 1420*** 115 600 Potassium, g 4.46 0.440 3.66 0.471 5.08*** 0.416*** 1.9 Magnesium, mg 442 43 348 44.7 514*** 41.8*** 350 Zinc, mg 14.6 1.41 11.3 1.45 17.1*** 1.39* 12 Selenium, microg 84.6 8.27 68.0 8.72 97.5*** 7.93*** ± Vitamin A, microg 1550 156 1370 179 1680 138 1000 Betacarotene, microg 2190 224 1980 261 2350* 196*** Ð Vitamin D, microg 4.90 0.48 4.01 0.514 5.59*** 0.454 5 Vitamin C, mg 135 13.6 117 15.2 148*** 12.3*** 60 Vitamin E, mg 10.5 1.02 8.16 1.04 12.3*** 0.997* 10 Folate, microg 323 31.8 256 33.0 375*** 30.6*** 200 Ribo avin, mg 2.57 0.251 2.05 0.263 2.97*** 0.241 1.6 Thiamin, mg 2.00 0.915 1.57 0.201 2.32*** 0.190*** 1.4 a t-test for differences between underreporters and others, *P < 0.05, **P < 0.01, ***P < 0.001. b Food and Nutrition Board, 1989.
300 Table 4 Mean micronutrient intake among women in 1992 according to energy intake (EI) levels All EI < 1.27*BMR EI > 1.27*BMR a Recomendations b Nutrient total /MJ Total /MJ Total /MJ Total Iron, mg 13.4 1.74 11.0 1.82 15.6*** 1.67*** 12±18 Calcium, mg 996 129 802 136 1170*** 125** 800 Potassium, g 3.61 0.473 3.07 0.514 4.08*** 0.440*** 1.9 Magnesium, mg 342 44.9 290 48.3 388*** 41.8*** 300 Zinc, mg 11.0 1.43 8.91 1.48 12.8*** 1.37*** 12 Selenium, microg 62.5 8.08 50.4 8.32 73.2*** 7.84*** Ð Vitamin A, microg 1343 176 1140 194 1520* 161 800 Betacaroten, microg 2600 350 2400 411 2770* 299** Ð Vitamin D 3.75 0.485 3.10 0.511 4.31*** 0.464 5 Vitamin C, mg 147 19.5 132 22.0 159*** 17.3*** 60 Vitamin E, mg 8.44 1.09 6.66 1.10 10.0*** 1.7 8 Folate, microg 272 35.9 232 38.7 308*** 33.2*** 200 Ribo avin, mg 2.10 0.270 1.69 0.277 2.46*** 0.266 1.2 Thiamin, mg 1.5 0.196 1.23 0.203 1.74 0.186 1.0 a t-test between underreporters and non-underreporters, *P < 0.05, **P < 0.01, *** P < 0.001. b Food and Nutrition Board, 1989. Figure 3 Food intake of underreporters compared to food intake of others. ttest for differences between underreporters and others, *P < 0.05, **P < 0.01, ***P < 0.001. found no association between adiposity and underreporting. However, their study was not a population survey but an experimental study. In another experimental study (Lichtman et al, 1992) the energy expenditure measured by doubly labeled water was compared to weighed record. In this study all subjects were overweight. They were divided into two groups depending on whether or not subjects failed to lose weight on a low calorie diet. 47% in the diet resistant group and 19% in the non diet resistant group were underreporters. It can be argued that we have not found underreporters but people who happened to eat little. Reasons for this may be several: illness, temporary lack of money, traveling, low basal metabolic rate etc. Ballard-Barbash et al (1996) has suggested that overweight people are more likely to be dieting or restricting their food intake than others. This could be one cause for the higher prevalence of underreporting among overweight persons compared to others. Another argument could be that three consecutive days is not a good way to study long term eating habits of individuals. However, it is very unlikely that 40% of persons were eating little only by chance. Proportion of underreporters was also high (30% among men, 47% among women) in a British study, although a longer recording period (7 d) was used (Pryer et al, 1995) It should be mentioned that there were also overreporters, but with this method we were not able to identify them. Furthermore, for energy overreporting there are no standard cut-off points. The problem of overreporting was investigated by Mertz et al (1991). In their study 266 volunteers recorded their food intake for at least 7 d and thereafter they were fed diets adjusted in amounts to maintain their body weight for more than 45 d. Only 8% of people overestimated their food intake whereas 21% were accurate within 419 KJ of their maintenance requirement and remaining 81% reported their intake below their maintenance requirment. This suggests that overreporting is not a large problem. Part of the decline of energy intake from 1982 to 1992 seems to be real. This can be seen from food balance sheets (Agricultural Economics Research Institute, 1985, 1995). Energy intake was 11.93 MJ=person in 1982 and 11.52 MJ=person in 1992 (decline 3.4%). According to the food balance sheets energy intake from fat has decreased from 36.5% in 1982 to 35.4% in 1992 (in our study from 38.3% in 1982 to 34.7% in 1992). This is in accordance with our survey data. However, part of the apparent decline in energy intake was due to increased underreporting. Even though the proportion of underreporters was large, excluding them did not change the proportion of fat, protein and carbohydrates of energy intake. In total fat intake the lower proportion of energy from fat spreads and oils among underreporters compared to others was compensated by the higher proportion of energy from meat (among male underreporters) and cheese (among female underreporters). This is an encouraging result, since macronutrient intake is one of the main focuses in these dietary surveys. This implies that despite underreporting, results can be valid at least as shares of energy intake. The intake of micronutrients, however, changed when underreporters were excluded. Thus the effect of excluding underreporters should be examined whenever population means and distributions of nutrient intake are shown and the results are being interpreted. Conclusions Underrporting has become more common since 1982. The magnitude of underreporting is large, especially among women, middle aged or older and overweight people. However, underreporting did not distort the main conclu-
sions of the dietary surveys, when looking at macronutrient intakes. In micronutrient intakes underreporting caused signi cant bias. This should be taken into consideration in future studies. AcknowledgementsÐWe are grateful to M.Sc. PaÈivi Kleemola for checking the food records and M.Sc. Jukka Lauronen for technical help. References Agricultural Economics Research Institute (1985): Balance sheet for food commodities. Year 1982. Agricultural Economics Research Institute, Helsinki. Agricultural Economics Research Institute (1995): Balance sheet for food commodities. Year 1992. Agricultural Economics Research Institute, Helsinki. Ballard-Barbash R, Graubard, I, Krebs-Smith SM & Schatzkin A, Thompson FE (1996): Contribution of dieting to the inverse association between energy intake and body mass index. Eur. J. Clin. Nutr. 50, 98±106. Black AE, Goldberg GR, Jebb SA, Livingstone MBE, Cole TJ & Prentice AM (1991): Critical evaluation of energy intake data using fundamental principles of energy physiology: 2, evaluating the results of published surveys. Eur. J. Clin. Nutr. 45, 583±599. Fogelholm M, MaÈnnistoÈ S, Vartiainen E & Pietinen P (1996): Determinants of energy balance and overweight in Finland 1982 and 1992. Int. J. Obes. 20, 1097±1104. Food and Nutrition Board (1989): Recommended dietary Allowances, 10th edn, National Research Council. Goldberg GR, Black AE, Jebb SA, Cole TJ, Murgatroyd PR, Coward WA & Prentice AM (1991): Critical evaluation of energy intake data using fundamental principles of energy physiology: Derivation of cut-off limits to identify under-recording. Eur. J. Clin. Nutr. 45, 569±581. HaraldsdoÂttir J & SandstroÈm B (1994): Detection of underestimated energy intake in young adults. Int. J. Epidemiol. 23, 577±582. Heitmann BL & Lissner L (1995): Dietary underreporting by obese individualsðis it speci c or non-speci c? Br. Med. J. 311, 986±989. HulteÂn B, Bengtsson C & Isaksson B (1990): Some errors in a longitudinal dietary survey revealed by the urine nitrogen test. Eur. J. Clin. Nutr. 44, 169±174. Karisto A, PraÈttaÈlaÈ R & Berg M-A (1993): The good, the bad, and the uglyðdifferences and changes in health related lifestyles. In: Kjaernes U, Holm L, EkstoÈm M, FuÈrst EL, PraÈttaÈlaÈ R (eds). Regulating Markets, Regulating PeopleÐOn Food and Nutrition Policy Novus forlag: Oslo, p 186. Kleemola P. Virtanen M & Pietinen P (1994): The 1992 dietary survey of Finnish adults. National Public Health Institute Publication B2, 1994. Lichtman SW, Pisarska K, Berman ER, Pestone M, Dowling H, Offenbacher E, Weisel H, Heshka S, Matthews DE & Heyms eld SB (1992): Discrepancy between self-reported and actual caloric intake and exercise in obese subjects. N. Engl. J. Med. 327, 1893±1898. Lissner L, Habicht J-P, Strupp BJ, Levitsky DA, Hass JD & Roe DA (1989): Body composition and energy intake: do overweight women overeat and underreport? Am. J. Clin. Nutr. 49, 320±325. Mertz W, Tsui JC, Judd JT, Reiser S, Hallfrisch J, Morris ER, Steele PD & Lashley E (1991): What are people really eating? The relation between energy intake derived from estimated diet records and intake determined to maintain body weight. Am. J. Clin. Nutr. 54, 291±295. Ovaskainen M-L, Valsta L & Lauronen J (1996): The compilation of food analysis values as a database for dietary studiesðthe Finnish experience. Food Chem: 57, 133±136. Pietinen P, Harman AM, Haapa E, RaÈsaÈnen L, Haapakoski J, Palmgren J, Albanes D, Virtamo J & Huttunen JK (1988): Reproducibility and validity of dietary assessment instruments I: A self-administered food use questionaire with a portion size picture booklet. Am. J. Epidemiol. 128, 655±666. Prentice AM, Black AE, Coward WA, Davies HL, Goldberg GR, Murgatroyd PR, Ashford J, Sawyer M & Whitehead RG (1986): High levels of energy expenditure in obese women. Br. Med. J. 292, 983±987. Pryer J, Brunner E, Elliot P, Nichols R, Dimond H & Marmot M (1995): Who complied with COMA 1984 dietary fat recommendations among a nationally representative sample of British adults in 1986-7 and what did they eat? Eur. J. Clin. Nutr. 49, 718±728. SAS Language and Procedures (1989): Version 6, Ist edn. SAS Institute, Inc: Cary, NC. Smith WT, Webb KL & Heywood PF (1994): The implications of underreporting in dietary studies. Aust. J. Public Health 18, 311±314. Uusitalo U, Pietinen P & Leino U (1987): Food and nutrient intake among adults in east and southwest FinlandÐA dietary survey of the FINMO- NICA project in 1982. National Public Health Institute Publication, B1 1987. World Health Organization (1985): Energy and protein requirements. Report of a joint FAO/WHO/UNI consultation. WHO Technical Report No. 724, Geneva. 301