- Click here for explanation of classification scheme.

• Examine the biological predictors of RMR in healthy older men
• Develop age-specific equation to predict RMR from easily measured variables
• Cross-validate age-specific equation in an independent sample
• Compare the study’s measured RMR values with previously used generalized and age-specific prediction equations.

• Understand and give written consent
• Diseases in subjects that were allowed.

• Refusal to consent
• Clinical evidence of CHD (ex. ST-segment depression greater than 1mm at rest or exercise) or cardiomyopathy)
• Hypertension (blood pressure >140/90mm Hg)
• Medications that could affect cardiovascular function or metabolic rate
• Medical history of diabetes or obesity
• Instability of body weight during previous year(change of more than 2kg)
• Exercise-limiting non-cardiac disease (arthritis, peripheral vascular disease, cerebral vascular disease, etc.).

Recruitment

Procedures not specified

Design

Cross-sectional study

Blinding used 

Not applicable

Intervention 

Not applicable

Statistical Analysis

• Assessment of linearity of data: 
  • Semi-partial F tests were performed to determine whether quadratic models explained a significantly greater amount of variance in RMR above that explained by linear regression models. No curvilinear function was found to contribute significantly to the model above that found to contribute by linear function
• Pearson-Product-Moment correlation coefficients: to assess degree of association between variables
• Stepwise multiple regression analysis: To predict RMR from laboratory and easily measured variables
  • All-possible-subsets regression analysis results were similar; only results of stepwise regression analysis presented
  • Potential laboratory predictors included: Standing height, body weight, body surface area, body mass index (BMI), fat mass, fat free weight (FFW) four skinfold measurements, Vo₂-max (L/min), leisure time activity (LTA), daily energy and macronutrient intake, age and plasma thyroid hormone levels
  • A second practical equations was generated that included only easily measured variables: Body weight, standing height, age, skinfold measurements and LTA
• Cross-validation groups were developed by randomly assigning subjects to a validation group (N=61) or a cross-validation group (N=28)
• Measured and predicted means of RMR of the cross-validation group were compared using a dependent T test
• Sample correlation coefficient (r) between predicted RMR and measured RMR was compared with the multiple correlation coefficient (R²) obtained from the regression equation of the validation group by an independent z test
• P<0.05; means=SD or std error of the estimate (SEE).

Timing of Measurements

One data collection point

Dependent variables

• Resting metabolic rate (RMR)
  • Measured: Indirect calorimetry; energy expenditure (kcal per minute) calculated from Weir equation
    • IC type: Ventilated hood
    • Equipment of Calibration: Not specified
    • Coefficient of variation using std gases: No
    • Rest before measure (state length of time rested if available): Not reported
    • Measurement length: 45 minutes
    • Steady state: No steady state mentioned
    • Fasting length: Yes; overnight
    • Exercise restrictions XX hr prior to test? 48 hours after bout with exercise
    • Room temp: Not reported
    • Number of measures within the measurement period: Not reported
    • Were some measures eliminated? Not reported
    • Were a set of measurements averaged? Not reported
    • Coefficient of variation in subjects measures? Not reported
    • Training of measurer? Yes, by Poelman and is in a clinical research facility
    • Subject training of measuring process? Yes the night before.

Independent variables

• Weight: Method not specified in this report (reference for laboratory methods given)
• Height: Method not specified in this report (reference for laboratory methods given)
• Body fat content: Estimated from body density as measured by underwater weighing (Keys and Brozek formula)
• Fat-free weight (FFW) total body weight minus fat weight
• Skinfolds: Triceps, chest, abdomen and thigh using Lange skinfold caliper
  • Measured on right side to the nearest 0.5mm; each value represents the mean of three consecutive measurements taken by same investigator
• Leisure time activity (LTA): Structured interview using the Minnesota LTA Questionnaire
• VO₂max: Measured by progressive and continuous treadmill test to volitional fatigue; highest oxygen uptake for one minute during the test was recorded as the maximal VO₂
• Energy intake and macronutrient composition: recorded from three-day food weight food diaries for two weekdays and one weekend day; Nutritionist III used for analysis
• Plasma hormone: Thyroxine (T₄), free T₄, and 3, 5, 3'-triiodothyronine (T₃) measured using clinical assay kits; free T₃ measured using an analogue assay.

 

 

 

• Initial N: N=89 males
• Attrition (final N): M=89
• Age: 63±8 years (range: 50-78)
• Ethnicity: Not reported
• Other relevant demographics: None specified
• Anthropometrics:

	Mean±SD	Range
Weight, kg	77±9	60-99
Height, cm	177±7	162-191
Body fat, percent	20±6	8-33
BMI	21.6±2.7	17.0-29.0
Fat-free body mass, kg	61±6	48-74
Fat mass, kg	16±6	5-28

•  Location: United States.

 

 

The measured RMR range in this sample was kcal per day: 1,645±83 (1,397-1,868)

Correlations (r) of measured variables with RMR:

• Weight: 0.73 (P<0.01)
• Height: 0.46 (P<0.15)
• FFW: 0.82 (P<0.01)
• Fat weight: 0.26 (P<0.05)
• Age: -0.16 (P<0.05)
• VO₂max: 0.54, (P<0.01)
• T₃: 0.26 (P<0.05)
• Free T₃: 0.37 (P< 0.01)
• Chest skinfold: 0.11 (NS).

Biological equation

• Fat-free weight (FFW) was the strongest predictor of RMR as it explained 85% of the total variance
• FFW, VO2 max, and free T₃ explained 87% of the variance (P<0.05) 
  • VO2max and free T₃ explained a small but significant 1% of the unique variance
  • Other predictor variables were considered in the analyses but were not entered into regression equations were body weight, height, age, percent body fat, fat weight, LTA, skinfold thicknesses, dietary intake data, body surface area, T₃, T4 and free T₃.

Stepwise Multiple Regression Analysis for Biological Equation in 89 Healthy Older Men Aged 50 to 78 Years

 

Dependent variable	Step	Independent variable	R2 (percent)	P
RMR	1	FFW	85	<0.01
RMR	2	VO2max	86	<0.05
RMR	3	Free T3	87	<0.05

Practical equation

• There were no significant differences between groups (cross-validation and validation) for all variables measured
• In the validation group,
  • The best set of predictors of RMR from validation group data were weight, chest skinfold thickness, LTA and BMI. (r=0.90, R²=0.81, SEE=37kcal per day, P<0.01.
  • There were no significant differences between the means of measured (1,644±87kcal per day) and predicted RMR from the practical equation (1,642±69kcal per day)
• In the total sample (pooled validation and cross-validation group):
  • The practical equation included weight, chest skinfold, LTA and age and the cumulative R²=76% (P<0.05).

Stepwise Multiple Regression Analysis for Practical Equation in 89 Health Older Men (Aged 50-78 years)

 

Dependent variable	Step	Independent variable	R2 (percent)	P
RMR	1	Weight	54	<0.01
RMR	2	Chest skinfold	69	<0.01
RMR	3	LTA	73	<0.01
RMR	4	Age	76	<0.05

 

Practical equation validation background information:

The means for the residual spread (predicted minus measured RMR) from the equations:

• WHO underestimated by 3.5% (SD±5%)
• HB underestimated by 4% (SD±6%)
• Mifflin underestimated by 5% (SD±5%)
• Fredrix overestimated by 3% (SD±5%).

The relative range of predicted RMR from measured values varied from -17% to 7% for WHO; -19% to 9 % for HB, -3% to 14% for Fredrix and -18% to 5% for Mifflin, suggesting a loss of accuracy when applied to individuals.

Comparison of Original Prediction Equations of the Four Studies Used for Cross-Validation Analyses and the Practical Equation From the Present Study

 

Study	Original Equation
Mifflin et al (N=50 men>50 years)	RMR (kcal per day)=10 (W)=6.25 (H) -5(A)+5 (R2=0.71)
Harris-Benefict (N=5 men>50 years)	RMR (kcal per day)= 5(H)+13.7 (W)-6.8 (A)+66 (R2=NA)
Fredrix et al (N=18 men>50 years)	RMR (kcal per day)=(10.7[W]-9[A] -203)+1,641 (R2=0.84)
WHO/FAO/UNU (N=? men>60 years)	RMR (kcal per day)=8.8(W)+11.3(H)-1,071 (R2=0.71)
Arciero and Poehlman (N=89 men > 50 years)	RMR (kcal per day)=9.7(W)-6.1(CSF)-1.8(A)=0.1(LTA)+1,060  (R2=0.76)

W: Weight (kg); H: Height (cm); A: Age (year); CSF: Chest skinfold (MM); NA: Not available; R²: squared correlation coefficient from the original study

As stated by the author in body of report:

• “In our study, body weight alone best predicted RMR, accounting for 54% of the explained variance in RMR in older men. This finding is in agreement with other investigations that have shown a high correlation between body weight and RMR. The use of body weight alone would permit an estimation of RMR within a residual error of±42kcal per day.”
• “Our results confirm earlier findings that FFW is the best predictor of RMR in older men, explaining 85% of the variance (R2=0.85). Based on a measurement of RMR we could predict RMR in older men within an error of ±32kcal per day. An additional 2% variance was accounted for by adding free T3 and VO2 max to the model.”
• Our three-variable biological equation would predict RMR within an error of±30kcal per day
• The LTA Questionnaire was found to accurately estimate energy expenditure from physical activity (r=0.83, P<0.01)
• This study found no relationship between RMR and age in males up to 40 years. However, in men older than 40 years, a linear decrease in RMR and age was noted. Based on this finding the development of a generalized equation with linear regression techniques across a wide age range would be less precise for use in older individuals due to the curvilinear relationships between age and RMR.
• “Harris Benedict results in a wide range of under-predictions and over-predictions of individual RMR values (-19% to 9%). The most plausible reasons for errors are differences in indirect calorimetry equipment, brief measurement period (five to 15 inches) and paucity of older individuals.”
• “Our equation to predict RMR may be useful in predicting daily energy requirements in the elderly if it is used in combination with an activity questionnaire.”

Government:	NIA, NIH, American Diabetes Association
Not-for-profit	American Association of Retired Persons Andrus Foundation

Strengths

• A large sample of older men
• Used cross-validation techniques
• Reported that “all measurement techniques were highly standardized to ensure optimal basal conditions; including inpatient settings and 48-hour post-exercise session.”

Generalizability/Weaknesses 

• “Self-selection sample that is highly motivated to keep a food diary, participate in underwater weighing and complete a leisure time physical activity questionnaire
• “References to three studies to describe the indirect calorimetry protocol by one of the researchers makes it difficult to ascertain the specific variables that affect REE measurement accuracy”
• “No limitations of study discussed and in discussion refer to a study by one of the authors that contains N=13 (six women; seven men); as well as a study completed in children in a clinical setting -which does not apply to elderly”
• Limitations include lack of subjects greater than 80 years old
• Comment in JPEN J Parenter Enteral Nutr. 1993; 18 (2): 193-194 states:
  • “The equation by the authors may be readily applied to ambulatory patients, but there is still a great deal of variation in predicting total energy needs because of the extremes of daily activity levels.”