Impact of body mass index and age on the relative accuracy of self-reported energy intakes among Japanese patients with type 2 diabetes

Abstract

Dietary questionnaires have been used to ascertain food or nutritional intakes in many studies; however, the extent and characteristics of measurement errors in patients with diabetes have not been examined. This study examined the measurement errors from self-reported dietary history questionnaires (DHQ) in Japanese patients with type 2 diabetes (T2D). Fifty-nine patients with T2D underwent a 24-h urine collection and 3-day dietary record (DR), and completed the DHQ. Intakes of energy, protein, sodium, and potassium were calculated from the DHQ. The estimated energy intake was calculated from the DR, and estimated intakes of protein, sodium, and potassium were determined from the 24-h urine samples. Energy intake values from the DHQ were lower than those from the DR by 12.5% in men and by 14.6% in women, which was significant only in men. In women, protein intake values from the DHQ were 19% higher than those from the 24-h urine. Multivariable linear regression analysis showed that energy intake ratio (DHQ/DR) was significantly negatively associated with body mass index (BMI) in both sexes and significantly positively associated with age only in women (all p < 0.05). Protein intake ratio (DHQ/24-h urine) was positively associated with duration of diabetes only in men (p < 0.05); however, this relation disappeared in the multivariable model. No factors showed association with sodium or potassium intake ratio. The DHQ showed under-reporting of energy intake by approximately 15% in Japanese patients with T2D. This was associated with obesity in both sexes and with younger age in women.

Introduction

Dietary therapy is highly important in the treatment of diabetes; thus, it is essential to ascertain the patient’s dietary intake and provide appropriate guidance to facilitate adequate glycemic control. The self-administered diet history questionnaire (DHQ) [1, 2] and its simplified version, the brief-type self-administered diet history questionnaire (BDHQ) [3], were developed to ascertain the food intake and nutritional intake in Japanese individuals. The validity of these questionnaires has been evaluated in previous studies [1, 2, 4, 5], and they have been recently used for several epidemiological and human nutritional studies. However, diet and nutrition surveys using dietary questionnaires are prone to reporting errors because of information being collected on the basis of participant self-reports.

The most frequent reporting error that requires attention is under-reporting of energy intake. In a study targeting healthy Japanese individuals, the mean under-reporting rate of energy intake was 11% for men and 15% for women [6]. Moreover, healthy obese individuals tend to under-report their dietary intake [6–8].

Patients with type 2 diabetes (T2D) have also been predicted to have higher rates of reporting error because they have better awareness of “divergence” and “habits” in their meals than do healthy individuals. However, up until now, the extent and characteristics of errors from existing DHQ [1] have not been investigated in patients with diabetes. This study aimed to examine the accuracy of dietary and nutritional intake self-reporting in Japanese patients with T2D who completed the DHQ.

Methods

Participants

We recruited 60 Japanese patients with T2D who visited the Diabetes Center of Tokyo Women’s Medical University Hospital and were aged 20–75 years, had a hemoglobin A1c (HbA1c) of 8.0% or less, were not taking oral diuretics, and were without diabetic nephropathy (estimated glomerular filtration rate: ≥ 60 ml/min/1.73 m²; urinary albumin–creatinine ratio: ≤ 30 mg/gCr). Of these patients, one patient did not complete the study, and the remaining 59 patients (33 men and 26 women) were included for the data analysis.

Urine collection

A 24-h urine collection test (24-h urine) was conducted for all patients. The patients were explained the method for urine collection over 24 h and were provided three or four 1-L plastic containers for urinary collection; they were requested to bring them back to the hospital on completing urine collection. They were also requested to void the entire urine volume in the bladder at the beginning of urine collection, record estimated urine volumes if they failed to collect urine, and record the time at which they started and completed urine collection.

Diet history questionnaire (DHQ)

When 24-h urine collection was conducted, information on dietary and nutritional intake over the past month was obtained from each patient using a DHQ. The DHQ [1, 2] is a validated 16-page questionnaire developed to quantitatively determine the intake of nutrients and food in detail. Answering the questionnaire takes at least 40 min, as it is the most detailed questionnaire for meal assessment among Japanese individuals. With the DHQ, it is possible to calculate the intake of approximately 40 types of nutrients and 150 types of foods using a dedicated nutritional value calculation program. These calculation programs have been found to strongly reflect the data from a 16-day food diary.

Three-day dietary record

After the DHQ and 24-h urine collection test, all patients received personal nutritional instructions from a dietician (YT). Before that, they were requested to complete a three-day dietary record (3-day DR) to ascertain their meal content and meal weight, either consecutively or non-consecutively for 3 weekdays. For those items that could not be weighed, it was requested that the approximate or standard amount be recorded.

Biochemical analysis

All biochemical assessments were conducted at a commercial laboratory (SRL or BML, Tokyo).

Urinary osmolarity was measured by the cryoscopic method (OM-6060, Arkray, Kyoto, JAPAN). Urinary creatinine (enzyme method, Cygnusauto^®), protein (pyrogallol method, Micro TP-AR^®), urea nitrogen (Urease UV method, Quickautoneo BUN^®), sodium (Ion selective electrode), and potassium (Ion selective electrode) levels were measured using LABOPSPECT008 (Hitachi High-Technologies Clinical analyzer, Tokyo).

Venous blood samples were obtained. Glycated hemoglobin (HbA1c; high-performance liquid chromatography method) values and serum concentrations of urinary nitrogen (Urease UV method, Quickautoneo BUN^®), creatinine (enzyme method, Cygnusauto^®), sodium (Ion selective electrode), potassium (Ion selective electrode), total protein (Biuret test, Qualigent TP^®), and albumin (BCP method, Qualigent ALB^®) were measured using LABOPSPECT008 (Hitachi High-Technologies Clinical analyzer, Tokyo) in all patients when 24-h urine tests were conducted.

Anthropometric measurements

Anthropometric measurements and body composition were evaluated on the basis of the participants’ height and weight when 24-h urine collection was conducted for each patient. Height was measured to the nearest 0.1 cm with the participant standing without shoes, using a digital stadiometer (AD-6400; A&D, Tokyo). Weight was measured to the nearest 0.1 kg, with the participant dressed in light indoor clothing, using a digital scale (WB-150; Tanita, Tokyo). BMI was calculated as [weight/height²] (kg/m²).

Information including duration of diabetes, use of medications, the presence of diabetic retinopathy and neuropathy, and comorbidities were examined from the patients’ clinical records and questionnaires.

Analysis of covariates

First, intake of energy, protein, sodium, and potassium was calculated from the DHQ for each patient.

Next, the estimated energy intake obtained from the 3-day DR was calculated by the dietician (YT) according to the Japanese Food Standard Ingredients Table 2015 (7th Edition) [9] and Handbook for Carbohydrate Amount [10]. For foods and drinks with no reference amount, the estimated amount for one meal was calculated with reference to the booklet.

The estimated intake of protein, sodium, and potassium was assessed using data from the 24-h urine test. Protein intake estimates were calculated using the Maroni formula (daily urea nitrogen excretion [g] + 0.031 [g/kg] × body weight [kg] × 6.25 + urinary protein) [11]. Sodium and potassium intake estimates were on the basis of the assumption that, on average, 86% of sodium intake and 77% of potassium intake is excreted in the urine; thus, the following formulae [12] were used: daily urinary sodium (mEq/day) × 23/0.86; daily urinary potassium (mEq/day) × 39/0.77.

The differences in the energy intake amounts estimated from the DHQ and from 3-day DR and the difference in protein, sodium, and potassium intake amounts estimated from the DHQ and from 24-h urine were calculated.

The percentage differences were calculated as gap values (values estimated from the DHQ subtract the 3-day DR or 24-h urine) divided by values estimated from the DHQ. Similarly, the ratio of the estimated value from the DHQ and from the 3-day DR for energy intake (hereinafter referred to as DHQ/DR) and for protein, sodium, and potassium intake (hereinafter referred to as DHQ/24-h urine) were calculated. These values were used to assess under- or over-estimation of energy, sodium, and potassium intake due to self-reporting.

Statistical analysis

Data are shown as median (25th, 75th quartile) or percentage (%). The Mann–Whitney U test was used for comparing the independent 2-group median values, and the Wilcoxson rank sum test was used for comparing the dependent 2-group median values. Fisher’s exact test or the chi-square test was used for percentage comparisons.

Univariable and multivariable linear regression analyses were performed to assess linear relationships between clinical parameters (age, duration of diabetes, BMI, and HbA1c) as independent variables and energy intake ratios (DHQ/DR) or protein, sodium, and potassium intake ratios (DHQ/24-h urine) as dependent variables in men and women, separately.

The Statistical Package for Social Science for Windows (Chicago, IL, USA) ver. 24 was used for statistical analysis, and p < 0.05 was set as a significant difference in two-sided tests.

Results

Clinical characteristics in men and women with T2D (Table 1)

Table 1.

Medians and their inter-quartile ranges or proportions for clinical characteristics in men and women with type 2 diabetes

	Men	Women	p value
Number	33	26
Age (years)	60 (50, 68)	55 (41, 66)	0.114
BMI (kg/m2)	24.0 (22.8, 34.3)	25.8 (22.7, 35.9)	0.536
Obesity (%)	39.4	50.0	0.441
Duration of diabetes (years)	6 (1, 13)	5 (1, 17)	0.635
HbA1c (%)	6.9 (6.6, 7.6)	6.9 (6.7, 7.4)	0.437
Diabetic neuropathy (%)	40.0	66.7	0.124
Diabetic retinopathy (%)	28.0	39.0	0.521
Anti-diabetes medications (%)
Biguanides	70.4	66.7	0.759
α-glucosidase inhibitor	17.2	14.7	0.813
DPP-4 inhibitor	31.3	26.5	0.569
Sulphonylurea	37.0	27.3	0.578
Thiazolidine	18.8	10.1	0.215
Insulin	18.5	12.1	0.718

BMI body mass index, Obesity BMI ≥ 25 kg/m²

p value: Mann Whitney U test, Fisher’s exact test or Chi-square test

There were no differences in the median age, BMI, duration of diabetes, or HbA1c between sexes. In addition, there was no difference in the proportion of patients with obesity or diabetic neuropathy or retinopathy between sexes. There were no sex difference in the proportion of patients under various anti-diabetes medications between sexes.

Difference in energy, sodium, and potassium intake between two estimation methods in men and women with T2D (Table 2)

Table 2.

Differences in energy, protein, sodium, and potassium intake between two estimation methods in men and women with type 2 diabetes

	Estimation			p value
	DHQ	DRa or 24-h urineb	cDifference (%)
Men (n = 33)
Energy (kcal/day)	2081 (1780, 2264)	2340 (1975, 2821)a	− 12.5	0.009
Protein (g/day)	75 (59, 90)	69 (60, 78)b	+ 8.0	0.085
Sodium (mg/1000 kcal)	1948 (1645, 2638)	2236 (1791, 2775)b	− 14.8	0.348
Potassium (mg/1000 kcal)	1137 (886, 1394)	1142 (878, 1452)b	− 0.4	0.796
Women (n = 26)
Energy (kcal/day)	1879 (1598, 2163)	2154 (1670, 2501)a	− 14.6	0.657
Protein (g/day)	70 (55, 89)	57 (48, 70)b	+ 18.6	0.025
Sodium (mg/1000 kcal)	2402 (1760, 2758)	2257 (1703, 3171)b	+ 6.0	0.829
Potassium (mg/1000 kcal)	1290 (1053, 1592)	1241 (865, 1754)b	+ 3.8	0.949

BMI body mass index, DHQ diet history questionnaire, DR dietary record; 24-h: 24-h

^aData for DR

^bData for 24-h urine

^cThe percentage differences were calculated as gap median values (median values estimated from the DHQ subtract the DR or 24-h urine) divided by median values estimated from the DHQ. p value: Wilcoxson rank sum test

The median energy intake values estimated from the DHQ were 12.5% lower in men and 14.6% lower in women than those estimated from the 3-day DR. However, a statistically significant difference between the two methods was shown only in men.

The median protein intake values estimated from the DHQ were 8% higher in men and 19% higher in women than those estimated from the 24-h urine collection. However, a statistically significant difference between the two methods was shown only in women.

As for median sodium or potassium intake, differences in estimated values between the DHQ and 24-h urine collection were not statistically significant in both sexes.

Factors associated with energy intake ratios (DHQ/DR) in men and women with T2D (Table 3)

Table 3.

Factors associated with energy intake ratio (DHQ/DR) in men and women with type 2 diabetes

	Univariable model			Multivariable model
	Beta	SE	p value	Beta	SE	p value
Men
Age (1 year)	0.006	0.005	0.254	0.001	0.006	0.928
Duration of diabetes (1 year)	0.005	0.010	0.622	− 0.008	0.010	0.442
BMI (1 kg/m2)	− 0.012	0.005	0.025	− 0.013	0.006	0.049
HbA1c (1%)	0.077	0.080	0.352	0.059	0.080	0.477
Women
Age (1 year)	0.021	0.006	0.007	0.021	0.004	0.007
Duration of diabetes (1 year)	0.013	0.007	0.096	0.002	0.004	0.676
BMI (1 kg/m2)	− 0.023	0.010	0.036	− 0.027	0.006	0.004
HbA1c (1%)	− 0.035	0.094	0.717	0.043	0.059	0.490

DHQ diet history questionnaires, DR dietary record, SE standard error

The univariable regression analysis showed that BMI in men and age and BMI in women had linear relationships with energy intake ratios (DHQ/DR). The significant relationship with BMI in men and age and BMI in women persisted in the multivariable model. The energy intake ratio (DHQ/DR) decreased significantly by 1.3% in men and 2.7% in women at 1 kg/m² increase of BMI and decreased significantly by 2.1% in women at 1-year decrease of age. These results indicated that under-reporting of energy intake was associated with obesity in both sexes and with younger age in women only.

Factors associated with protein, sodium, and potassium intake ratios (DHQ/24-h urine) in men and women with T2D

The univariable linear regression analysis showed that there were no significant relationships between factors and protein intake ratio (DHQ/24-h urine), except for duration of diabetes in men (β = 0.025, standard error = 0.007, p = 0.002). However, the significant relationship for duration of diabetes disappeared in the multivariable model. As for sodium or potassium intake ratios (DHQ/24-h urine), there were no significant linear relationships with factors for either sex.

Discussion

Accumulated data have shown that obesity is associated with underreporting in healthy individuals, although study methods differed from study to study [6–8, 13–17]. As for DHQ, previous studies showed that obese or younger healthy men and women tend to under-report their energy intake when compared to that in DR or estimated energy requirement (EER) according to the Food and Agriculture Organization/World Health Organization/United Nations University [6, 8, 18]. To the best of our knowledge, this is the first study to evaluate the accuracy of existing self-reported DHQ among Asian patients with diabetes. Our study showed that DHQ, which was developed to measure nutritional intake for healthy Japanese individuals, under-reported energy intake by 12.5% and 14.6% for men and for women with T2D, respectively. The extent of under-reporting increased with increases in BMI in both sexes and it increased with decrease in age in women.

The National Health and Nutrition Survey III has shown that people with diabetes under-reported energy intake in the 24-h DR (in comparison with the basal metabolic rate evaluations), more than that by people without diabetes [18]. However, the extent of under-reporting in our study was comparable to the results in healthy Westerners [19]. The BMI distribution range in randomly selected patients with diabetes resembled the range seen in general Westerners, which suggests that the range in BMI variation may explain the extent of under-reporting, irrespective of race or the presence of diabetes. Our hypothesis needs to be verified in future large-scale studies.

A previous study on healthy individuals used EER to examine the accuracy of energy intake by DHQ [8]. This is because energy intake, energy requirement and energy expenditure are considered equal for many adults as well as obese and undernourished people, as long as their weight and body composition remain unchanged [19]. Our study patients had stable glycemic control and complications, with a weight change of ± 0.5 kg over the past few months. However, there were no supporting data that EER can be read as energy intake for Japanese patients with diabetes under dietary treatment. Moreover, the doubly labeled water (DLW) method is the gold standard in calculating energy expenditure [20, 21]; however, it is expensive, difficult to measure for large numbers of people, and is not practical in clinical practice.

Overall, 81 studies on healthy individuals have shown that the energy intakes obtained by five different dietary assessment methods (diet recall, food frequency method, diet history, diet recall, and observation by an outsider [a third party]) were generally lower than the total energy expenditure measured by the DLW method [19]; besides, the extent of the underestimation increased with an increase in BMI [19]. The estimation based on the observation by an outsider showed the least gap, compared with that of the DLW method [19]. In our clinical study, it was impossible to use the DLW method or to perform a dietary assessment by the observation of an outsider. Although it is difficult for any dietary assessment to avoid the issue of under- or over-reporting, the 3-day DR was adopted in our study. This was because the 3-day DR was feasible in our routine clinical practice, and one Japanese study of healthy individuals showed that the minimum number of days necessary for estimating nutrient consumption was at least 2 days [22].

Due to the cross-sectional nature of the study, it is not known which of underreport or obesity is cause and outcome: obese patients are shame of being obese and underreport their intakes, or patients who underreport their intakes because of lack of meal memories overeat and become obesity. Within the group of patients with diabetes in our study, there may have been obese patients with eating disorders, such as bulimia or sleep-related eating disorder, which may have caused unintentional intake of foods or nutrients, although we could not assess their eating disorders during the study. One previous study showed that the extent of under-reporting of energy intake was not associated with higher BMI but was associated with higher depression scores in obese American women without diabetes [23].

Psychological factors [14, 16, 24, 25] influence under-reporting of energy intake via body weight change, originating from both the person’s own negative perception and society’s negative perception of obesity, with both parties believing that the higher the BMI, the more the person should restrict their diet; this leads to the under-reporting of energy intake. Under-reporting of energy intake may have been intentional or unintentional; however, we were unable to examine the issue in the current study.

In our female patients with T2D, younger age was independently associated with under-reporting of energy intake. The tendency to under-report energy intake in younger females has already been reported in studies targeting healthy individuals [6, 14]. This may reflect the strong desire by younger women to lose weight because they consider themselves to be obese even at a normal weight.

In contrast, in our male patients with T2D, shorter duration of diabetes was significantly associated with the under-reporting of protein intake. Although the reason for this was unclear, the findings might reflect our clinical experience that Japanese male patients with T2D appear to be, in general, less interested in dietary counseling and cooking than their female counterparts. Since no studies have analyzed data for diabetes using the same statistical approach as our study, previous results of sex-based reporting discrepancies in persons without diabetes (healthy individuals) are conflicting, with some reports suggesting no clear difference between men and women [13]; while, other reports suggest that women tend to under-report [26]. Further research will be needed.

Previous reports have indicated a positive correlation between BMI and under-reporting of protein and potassium intakes [27, 28]. In this study, we used energy-adjusted nutrient intake to eliminate the effect of under-reporting of dietary intake in the DHQ as much as possible, which may be the one of the reasons for the differences in the results. In another study [8] that used the same energy-adjusted nutrient intake as this study, the results were almost the same as our results, with no significant correlation found between BMI and the estimated intake ratio of protein, sodium, and potassium in the DHQ or urine collection results.

We found larger variations in energy intake than those shown in a previous study [29]. This may have been influenced by the small number of patients with diabetes and differences in individual comprehension skill. We were concerned that the explanation provided to participants in each case could induce bias by guided interrogation. Thus, the participants read the instructions attached to the DHQ and answered them.

The limitations of this study should be noted. First, we used the 3-day DR as a comparison, which may also have gaps from the true values of energy intake from participants. As discussed above, currently, an adequate method to measure true value of energy intake does not seem to exist. Second, the protein, sodium, and potassium intakes were estimated on the basis of a single 24-h urine collection, similar to that in previous studies, which assessed the validity of the estimated data from DHQ [8, 29]. It should be noted that at least three non-consecutive episodes of complete 24-h urine collection would be needed to accurately estimate usual dietary sodium at the individual level [30]. Single complete 24-h urine sample collection is recommended for assessing the current 24-h dietary sodium ingestion in the population [30]. As we focused on the sodium intake among the groups in this cross-sectional observation study, single complete 24-h urine collection was adopted. Moreover, mean deviation in estimated protein intake was relatively stable at 13% in a report with multiple urine collections [31]; thus, the single collection may not have significantly changed the results. Third, the small number of patients reduced the statistical power in the stratified analysis. Furthermore, the current study excluded patients with serious diabetic complications or HbA1c ≥ 8.0% (poor glycemic control) and this resulted in selection bias. Thus, our result should be interpreted with caution and the generalizability of our findings may be limited. There is an urgent need for verification of self-reported DHQ in large-scale studies for Japanese patients with T2D. Fourth, psychological and cultural factors may have influenced over-reporting and under-reporting in the DHQ as discussed above. However, there is little evidence on the extent by which they affect reporting [26]. Unfortunately, we have not examined these factors, and further research will be needed.

In conclusion, DHQ showed under-reporting of energy intake by approximately 15% in Japanese patients with T2D, and under-reporting of energy intake was associated with obesity in both sexes and with younger age in women.

Funding

This project was fully supported by the 24th Takako Satake Research Grant from Tokyo Women’s Medical University to TN.

Compliance with ethical standards

Conflict of interest

YT, TN, JO, CU, YT and SS have no conflicts of interest to disclose. TB is supported through unrestricted research funding from Baxter, Boehringer Ingelheim, Chugai, Daiichi Sankyo, Santen, Mitsubishi Tanabe, MSD, Nipro, Novartis, Novo Nordisk, Ono, and Sumitomo Dainippon.

Human rights statement

All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1964 and later versions.

Informed consent

Written informed consent was obtained from all patients included in the study. This study was approved by the ethics review board of Tokyo Women’s Medical University (Approval number: 3272R, October 22, 2017, 5481, February 10, 2020).