Accuracy of predictive equations for evaluation of resting energy expenditure in Brazilian patients with type 2 diabetes.


 Background Evaluation of the resting energy expenditure (REE) is essential to ensure an appropriate dietary prescription for patients with type 2 diabetes. The aim of this study was to evaluate the accuracy of predictive equations for REE estimation in patients with type 2 diabetes, considering indirect calorimetry (IC) as the reference method.Methods A cross-sectional study was conducted in 62 patients (31 men and 31 women) with type 2 diabetes. Clinical and laboratory variables were evaluated, as well as body composition by electrical bioimpedance. The REE was measured by IC (QUARK RMR, Cosmed, Rome, Italy) and estimated by predictive equations. Data were analyzed using Bland–Altman plots, paired t-tests, and Pearson’s correlation coefficients.Results Patients in the sample had a mean age of 63.1 ± 5.2 years, median diabetes duration of 11 (1–36) years, and mean A1C of 7.6 ± 1.2%. Body composition analysis revealed a mean fat free mass of 35.2 ± 11.8 kg and fat mass of 29.1 ± 8.8 kg. There was wide variation in the accuracy of REE values predicted by equations when compared to those measured by IC. For women, the FAO/WHO/UNO equation provided the best REE prediction in comparison to measured REE (-1.8% difference). For men, the Oxford equation yielded values closest to those measured by IC (-1.3% difference).Conclusions In this sample of the patients with type 2 diabetes, the best predictive equations to estimate REE were FAO/WHO/UNO and Oxford for women and men, respectively.


Introduction
Diabetes mellitus (DM) is a chronic disease that affects a signi cant proportion of the world population [1]. Type 2 diabetes is the most common form of DM, usually occurring in adulthood, and is associated with obesity in about 80% of cases [2]. The primary strategy for treating obese patients with type 2 diabetes is the loss of body mass through lifestyle changes [2], which has been associated with improvement in glycemic control [2]. Among these interventions, an appropriate dietary prescription with the goal of reducing body weight, taking into account each patient's daily energy needs, is essential [3].
The main component of energy requirements is the total energy expenditure (TEE); calculating the TEE requires knowledge of the resting energy expenditure (REE) [3].
Variability in REE may depend on several factors, such as sex, ethnicity, age, physical activity, genetic factors, body composition, caloric intake, and the presence of diabetes or obesity [11]. Several studies have evaluated REE using predictive equations across different populations [16][17][18] and ethnicities [19][20][21][22][23][24][25][26]. Studies considering sex have shown that REE is lower in women than in men [27][28][29]; one such study found that REE measured by IC was 23% higher in men [27]. These data contributed to a follow-up study conducted in obese men and women, which also demonstrated a signi cant difference (REE higher in men by approximately 335 kcal/day) [29].
In addition, the presence of diabetes is also associated with REE. Previous studies demonstrated that patients with diabetes and poor glycemic control had higher REE [9,25,26]. Data on the use of REE predictive equations in patients with type 2 diabetes have been described elsewhere [9,10,14,21,22,[24][25][26][30][31][32][33][34][35]; however, data on Brazilian diabetic patients are still scarce [34,35]. A cross-sectional study of obese Brazilian women with type 2 diabetes showed that some predictive equations underestimated REE by approximately − 2.6%, while others overestimated it by 10.6%, when compared with IC measurement [34]. A recent survey of Brazilian patients with type 2 diabetes of both sexes demonstrated wide variation in REE values evaluated by predictive equations. The FAO/WHO/UNO equation showed the best accuracy when compared to measured REE, but still underestimated it by -5.6% as compared to IC a difference of 100 kcal/day [35].
Considering that sex is an important variable in REE evaluation; that data in Brazilian patients with type 2 diabetes are insu cient; and that poor glycemic control has been associated with an increase in REE, evaluating the performance of predictive equations for REE in this population is essential to ensure that adequate dietary interventions are being prescribed for diabetic patients. Within this context, the aim of the present study was to evaluate the accuracy of the main predictive equations used in clinical practice for the calculation of REE in a sample of Brazilian patients with type 2 diabetes, strati ed by sex, considering IC as the reference method.

Study design and patients
This cross-sectional study included 62 patients (31 men and 31 women) with type 2 diabetes. Type 2 diabetes was de ned by age > 30 years at onset, no previous episode of ketoacidosis or documented ketonuria, and insulin treatment (when necessary) only 5 years after diagnosis. The inclusion criteria were not having received dietary counseling by a nutritionist in the preceding 6 months, age < 70 years, serum creatinine < 2 mg/dL, normal thyroid function tests, and absence of severe liver disease, decompensated heart failure, or any acute disease. The study protocol was approved by the Research Ethics Committee (Approval number: 15.0625), and all subjects provided written informed consent for participation.

Clinical evaluation
Blood pressure was measured with a digital sphygmomanometer (Blood Pressure Monitor, model HEM-705CP, Omron Healthcare Inc., Bannockburn, IL). Two measurements were obtained, 2 minutes apart, and the mean recorded for analysis. Patients were considered hypertensive in case of systolic blood pressure ≥ 140 mmHg on at least two occasions, history of hypertension, or current use of antihypertensive drugs.
The anthropometric parameters used to assess nutritional status were body mass (with participants barefoot and wearing lightweight clothing) and height, both measured with a calibrated anthropometric scale (Filizola®). The body mass index (BMI) was calculated as the body mass (in kg) divided by the height (in m) squared. Body composition analysis by electrical bioimpedance (InBody® 230, Seoul, South Korea) was performed for determination of fat mass (FM) and fat-free mass (FFM), both in kg.
Habitual physical activity was measured objectively by step counting with a pedometer (HJ-321, Omron Healthcare Inc.) and classi ed into ve levels: sedentary (< 5000 steps/day), low active (5000-7499 steps/day), somewhat active (7500-9999 steps/day), active (≥ 10000-12499 steps/day) and highly active (≥ 12500 steps/day) [36]. Participants wore the pedometer for 7 days, attached to the waistband of their clothing during waking hours, except when bathing or swimming. Participants were encouraged not to alter their usual physical habits during the protocol.

Resting energy expenditure measurement
Objective measurement of REE was performed by IC. The IC protocol consisted of 10 min of rest on a gurney in the supine position, followed by 30 min of collection of exhaled gases using the canopy dilution technique and a coupled collection device. An open-circuit calorimeter (QUARK RMR, Cosmed, Rome, Italy) was used to determine VO 2 (oxygen consumption) and VCO 2 (carbon dioxide production). To calibrate the equipment, the volume of the turbine owmeter was rst calibrated electronically by the system, followed by calibration of the collector plates using a known gas concentration. This process was repeated for each test to standardize measurement. The rst 10 min of gas collection were excluded from the analysis; thus, VO 2 and VCO 2 (L/min) obtained during the nal 20 min of each collection (mean value) were used for REE calculation. The equation proposed by Weir [37], which incorporates a correction factor and thus obviates the need to consider protein metabolism, was used to obtain values in kcal/min: [(3.9 x VO 2 ) + (1.1 x VCO 2 )]. The result in kcal/min was multiplied by 1,440 min to obtain the 24-hour REE.
Subjects were asked to refrain from all moderate-or high-intensity physical activity during the 24 h preceding the test, and not to consume alcohol or caffeine. Smokers were instructed not to consume any tobacco products for at least 12 h before the day of REE measurement. Additionally, the subjects were instructed to fast for 12 h prior to the test (water freely allowed) and to have a good night's sleep (at least 8 hours). Finally, all subjects either drove or were driven to the test site to avoid any energy expenditure before determination of REE. All tests were performed between 06:30 and 08:00, in a temperature-controlled (23 °C) and sound-controlled room, under low luminosity. All participants continued to take their usual medications during the study period; those who had morning doses to take received them after IC.

Selection of equations for estimating resting energy expenditure
The REE was estimated by eleven predictive equations, which were selected after a search of previous publications on the theme [4][5][6][7][8][9][10][11][12][13][14]. To be included, the equations had to have been developed for adult men and women and should be based on body weight, height, age, sex, and/or FM. Equations derived only for speci c ethnic groups or for individuals with BMI ≥ 40 kg/m² were not included (Supplement 1).

Statistical analysis
Sample size calculation was based on a study wherein the variability of REE in relation to glycemic control, weight, age, and sex-particularly in male patients-demonstrated a multiple correlation coe cient of 0.9 [25]. Considering a study power of 80%, alpha error of 5%, and 20% attrition rate, 62 patients would be required.
The means of estimated REE and measured REE were compared by a paired Student's t-test. Agreement between estimated and measured REE was examined graphically by plotting the differences between the predicted and the measured REE against their mean values, with 95% limits of agreement (mean difference ± 1.96 standard deviation) [38]. Pearson's correlation coe cients were used to assess the correlation between estimated and measured REE. Results are expressed as means and standard deviations or medians and interquartile ranges. Data were analyzed using SPSS version 23.0, while Bland-Altman plot values were analyzed in R version 3.3.3 (R Project for Statistical Computing, Vienna, Austria). A p value of < 0.05 was considered signi cant.

Results
A total of 62 patients with type 2 diabetes were included in the study (80.6% white; mean age, 63.1 ± 5.2 years; median disease duration, 11  years; mean BMI, 30.1 ± 4.0 kg/m²). A ow diagram of patient selection is shown in Fig. 1. Men had greater body mass (89.9 ± 13.8 vs. 74.2 ± 11; p < 0.001) and FFM (38.6 ± 12.1 vs. 31.7 ± 10.7; p = 0.009) when compared to women. Regarding physical activity, the median number of steps/weeks was 5522 (1496-18097), thus classifying the majority of participants as less active. All participants (100%) had hypertension. Most had a lipid pro le within normal limits; however, fasting blood glucose and A1c levels were abnormal, as expected in a sample of patients with diabetes.
All were on oral antihyperglycemic agents (100%) and antihypertensive agents (100%), while 67.7% (n = 42) also took lipid-lowering agents. The pro le of the sample is described in Table 1. -Chi-square test impossible because 100% of the sample is hypertensive, on hypoglycemic agents, and on antihypertensive agents.     [35]. In healthy Chilean individuals of both sexes, the Oxford equation also seems to be the best alternative for calculation of REE [39].
In our study, most predictive equations underestimated REE when compared to the reference criteria (-9.1 to -2.4% difference). In addition, we found a wide difference between measured and estimated REE, since the equations cannot estimate values with the same consistency and magnitude as IC. Similar discrepancies were also observed in other studies of patients with type 2 diabetes [34,35].
Sex is a factor that has been associated with REE [27][28][29]. When comparing the FAO/WHO/UNO equation in men and women, we found that it underestimated REE in both (-1.6% vs. -1.8%, respectively). Conversely, in a study of French patients with type 2 diabetes, this equation overestimated REE in both sexes [30]. In another study of Brazilian women with type 2 diabetes, the equation also overestimated REE when compared to IC [34].
The Harris-Benedict equation is that most used in clinical practice to determine energy requirements [4].
However, studies have shown that it may not be appropriate to estimate REE in both sexes [40,41]. In men and women without diabetes, the equation overestimated REE by 9% [40] and 14% [41], respectively. Most of the equations evaluated in this study were originally developed in healthy, eutrophic populations [4,[6][7][8]10]. Thus, the differences we observed may have been due to the presence of obese patients (BMI > 30 kg/m²) in our sample, as well as to the fact that, in individuals with diabetes, insulin resistance is associated with abnormal metabolic reactions [43]. In fact, the presence of diabetes per se in uences REE [9,10,14,26,33]. Studies conducted in Japan have shown that obese individuals with type 2 diabetes have a higher REE than their obese counterparts without type 2 diabetes, and that fasting blood glucose levels can be one of the main determinants of this increase [14,26]. More recently, a study also performed in Japanese patients with type 2 diabetes showed that REE correlated signi cantly with plasma glucose and HbA1c [33]. The reasons for this phenomenon are not yet well established, but factors such as increased gluconeogenesis [9], increased protein turnover [44], increased glycosuria [9], and elevated levels of glucagon [45] may all in uence REE in patients with diabetes.
In 2002, Gougeon et al evaluated the REE of women with type 2 diabetes and proposed an equation for predicting REE that included plasma glucose, HbA1c, and FM as independent variables [9]. As already noted, studies have shown that the presence of diabetes is an important variable that must be considered when evaluating REE [9,25,26]. In our study, however, these variables did not correlate signi cantly with REE in patients of either sex. Moreover, the equation proposed by Gougeon

Declarations
Ethics approval and consent to participate The study protocol was approved by the Research Ethics Committee of the Hospital de Clinicas de Porto Alegre #protocol number 1506.25, and all subjects provided written informed consent for participation.

Consent for publication
The manuscript has not been published (in full or in part) before and is not being considered for publication in any other journal while under consideration for Nutrition & Metabolism. The authors understand that, if accepted, this manuscript must not be published elsewhere in similar form, in any language, without the consent of this Journal.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.