- Click here for explanation of classification scheme.

To test the frequency and degree of disagreement between measured and predicted resting metabolic rate using the Harris-Benedict standard and two more recently published standards by Owen and Mifflin in order to determine: 

• What percentage of non-obese and obese subjects are correctly predicted by each equation
• Whether errors tend to be overestimates or underestimates
• Whether any one of these equations should be preferred over another.

1. Adults: Age minimum 18 years
2. Healthy
3. Non-hospitalized
4. Volunteers
5. Able to give consent
6. Non-obese and obese individuals.

1. Younger individuals
2. Inability to ambulate
3. Steroid use
4. Known diabetes mellitus (because of fasting state required prior to IC) or thyroid disease
5. No exclusions on basis of gender or race.

Recruitment

Subjects were recruited by:

• Posting flyers in various locations in hospital and medical school
• Referrals from a weight-management clinic and a surgery clinic that evaluates morbidly obese individuals for gastric surgery
• Referrals from family members and friends.

Design

Cross-sectional study

Blinding used 

None used; not applicable

Intervention 

Not applicable

Statistical Analysis

• Minitab used
• Absolute percent difference between calculated and measured values as a percent of measured resting metabolic rate (RMR) to compare each equation with measured RMR
  • Percent difference was presented in absolute terms to eliminate the mathematical tendency for negative variation to offset positive variation, thus keeping the mean near zero
• Disagreement between measured and predicted RMR: Two values being more than 10% different from each other
  • Analysis of proportions at and<0.05 to test for significant differences in the number of disagreements between measured and predicted metabolic rate.

Timing of measurements

One measurement time

Dependent variables

Accuracy for predicted equations: Predicted RMR within ±10% of measured RMR.

Independent variables

• Measured resting metabolic rate (RMR)
  • IC type: Open-circuit with gas collection via a rigid plastic canopy
  • Rest before measurement: Amount of time not described
  • Measurement length: 30 minutes with first five minutes of data discarded
  • Fasting prior to measurement: 12- to 15-hour fast overnight (away from research center)
  • Exercise restriction prior to measurement: Subjects abstained from physical exercise during the 12- to 15-hour fasting period
  • Room temperature: Ambient controlled by subject (light blankets offered)
  • Number of measurements: One 30-minute
  • Coefficient of variation: Equal to 10%.
  • Calibration of IC machine: Stated calibrated
  • Training of measurer: Not described
  • Subject given prior explanation of protocol and IC to minimize anxiety: Not discussed
  • Dietary Measurements: Not done
• Predicted RMR:
  
  • Harris Benedict (1917) (5% of subjects had BMI=30kg/m²
  • Variation of Harris Benedict where all subjects with BMI=30kg/m² had an adjusted weight substituted into the equation: Adjusted wt (kg)=( (body weight - ideal weight)x0.25) + ideal weight
    • Ideal body weight was determined from body height and calculated with the Hamwi equation
  • Owen (1986) with subjects of varying weights; 30% had BMI=30kg/m²
  • Mifflin (1990); 47% of subjects were obese with maximum BMI 42kg/m².

[Note: Owen and Mifflin do not use weight adjustments for obesity.]

Control variables

• Body mass index: kg/m² calculated from measured height (without shoes; Stadiometer) and weight (in light street clothes and without shoes; Detecto Scales and Scaletronix for >160kg)
  • Obesity: ≥30kg/m²
  • Morbid obesity: ≥40kg/m².

 

• Initial N: Not given
• Attrition N (final N): N=130 (54 male, 76 female) non-hospitalized adult volunteers
• Age: ≥18 years (by BMI group shown below)
• Ethnicity: Sample was 98% Caucasian (5% from Australia or Europe, in US for at least one year to study)
• Anthropometrics: Range of BMI—18.8-96.8kg/m
  • Description of three BMI groups:
    • Non-obese
      • N=83
      • Men: Age mean: 41±2.9 SEM (Range 20-78)
      • Women: Age mean: 38±2.2 SEM (Range: 22-74)
      • Height (cm): Range 146-192
      • Weight (kg): Range 39-103
      • BMI (kg/m²): Range 15.9-29.6
      • RMR (kcal per day): Range 990-2,090
    • Obese
      • N=20
      • Men: Age mean: 39±4 SEM (Range 18-62)
      • Women: Age mean: 44±4 SEM (Range 26-69)
      • Height (cm): Range 154-189
      • Weight (kg): Range 77-133
      • BMI (kg/m²): Range 30.5-39.9
      • RMR (kcal per day): Range 1,214-2,654
    • Morbid Obese
      • N=27
      • Men: Age mean: 41±3 SEM (Range 28-57)
      • Women: Age mean: 37±2 SEM (Range 25-52)
      • Height (cm): Range 145-192
      • Weight (kg): Range 109-330
      • BMI (kg/m²): Range 40.1-96.8
      • RMR (kcal per day): Range 1,679-4,533
• Location: Milton S. Hershey Medical Center, United States.

Major Results

Subject characteristics

• The groups were evenly matched for height and age
• BMI was normally distributed within the obesity groupings, but skewed toward a high BMI in the overall sample
• BMI was greater than 30kg/m² in 36% of subjects with 57% of those having BMI in excess of 40kg/m²
• 56% of subjects had sedentary lifestyles; the remainder undertook physical activity for recreation rather than occupational reasons
• Subjects whose predicted RMR was inaccurately predicted did not differ in height or age from those whose RMR was accurately predicted.

Prediction of RMR by four equations

• For the sample as a whole, the Mifflin equation was accurate more often (78%) than either the Harris Benedict (67%, P=0.05) or Owen equations (65%, P=0.02)
• Harris Benedict was significantly less accurate (69%) than Mifflin (82%) and Owen (73%) in non-obese subjects
• Owen was significantly less accurate than Mifflin in all obese subjects grouped together, with the errors concentrated in the morbid obese group
  • The Mifflin equation inaccurately predicted RMR in only 25% of men and women with BMI>40kg/m²
• When errors occurred, they tended to be overestimates for the Harris Benedict equation (using actual body weight) and underestimates for all other equations
  • Focusing on morbidly obese individuals in this study, use of Harris Benedict equation indicated a 12% overestimation using actual body weight and 22% underestimation using adjusted body weight (compared to 15% and 27%, respectively found in one validation study) 
    • Harris-Benedict overestimated RMR in 67% of men with BMI>50kg/m²
    • In women with BMI>50kg/m², Harris-Benedict inaccurately predicted RMR in only 13%

  • The effect of using adjusted body weight in the Harris-Benedict resulted in 100% incidence of underestimation of RMR in subjects with a BMI≥40 kg/m²

  • Consistent underestimation was noted for the Owen equation in women with BMI≥40kg/m² (92% of women).

 

Agreement between measured and predicted RMR in non-obese and obese men and women

Equation and Group	Percent of subjects within±10% of RMR	Percent of subjects above±10% of RMR	Percent of subjects below±10% of RMR
Harris-Benedict
Not obese: (BMI<30), N=83	69a	27	4
All obese: BMI≥30), N=47 BMI 30-40 (N=20) BMI>40 N=27	64b 50 74c,d	30 40 22	6 10 4
Adjusted Harris-Benedict
All obese (BMI≥ 30), N=47 BMI 30-40 (N=20 BMI≥40 N=27	26 60 0	2 5 0	72 35 100
Owen
Not obese (BMI <30), N=83	73	6	21
All obese (BMI≥30), N=83 BMI 30-40, N=20 BMI>40, N=27	51e 75 33a	6 5 7	43 20 60
Mifflin
Not obese (BMI <30), N=83	82	10	8
All obese (BMI≥30), N=47 BMI 30-40, N=20 BMI>40, N=27	70 70 70	9 10 7	21 20 23

^aP<0.05 vs. Mifflin of same group 

^bP<0.05 vs. Harris Benedict with adjusted weight of same group 

^cP<0.05 vs. Owen equation of same group 

^dP<0.05 vs. Mifflin morbidly obese group

Mean percent difference between calculated and measured RMR with inaccurate predictions of RMR (i.e., calculated >10% of measured). Mean values ±SEM shown:

 Harris Benedict

• Not obese: 13.3±0.8
• All obese: 20.0±1.9 (P<0.05 not obese of same equation, vs. Mifflin of the same group)
• Obese: 16.4±1.4
• Morbid obese: 26.2±1.6 (P<0.05 not obese of the same equation, vs. Mifflin of the same group, vs. BMI 30-40 of the same equation, vs. Owen of the same group).

Harris Benedict (adjusted weight)

• All obese: 25.0±1.3 (P<0.05 vs. Mifflin of the same group)
• Obese: 19.5±2.1
• Morbid obese: 26.7±1.2 (P<0.05 vs. Mifflin of the same group, vs. BMI 30-40 of the same equation, vs. Owen of the same group.

Owen

• Not obese: 15.4±1.1
• All obese: 16.3±1.0
• Obese: 21.4±2.4
• Morbid obese: 16.0±1.3.

Mifflin

• Not obese: 13.7±1.0
• All obese: 13.6±0.8
• Obese: 17.4±1.6
• Morbid obese: 13.0±1.3.

 

 

• Of the equations tested, the Mifflin equation provided an accurate estimate of actual RMR in the largest percentage of non-obese and obese people, with a low rate of overestimation
• Among morbidly obese people whose RMR was not accurately predicted, the magnitude of error was smaller for the Mifflin equation than for the other equations.
• Thus, the Mifflin equation deserves consideration as the standard for calculation of RMR in obese and non-obese individuals
• The clinician can expect disagreement between measured and predicted RMR to be more common in obese individuals than in non-obese individuals, but those individuals who do not conform to standard cannot be identified without actually measuring RMR
• The usual clinical maneuver of inserting an adjusted body weight into the Harris Benedict equations corrects the tendency of that equation to overestimate RMR but replaces it with a tendency to underestimate RMR
• Although error was more likely in obese than non-obese people for all equations, there were no clinically applicable markers to identify those people whose predicted RMR should be questioned
• Using adjusted body weight in the Harris Benedict equation caused underestimation of RMR in obesity, with the magnitude of the error increasing as BMI increased. Thus, the use of adjusted body weight is not a reasonable maneuver in the clinical setting.
• In morbid obese individuals, the Harris Benedict equation tended to overestimate, while errors in the Owen and Mifflin equations tended to be underestimates.

University/Hospital:

Pennsylvania State University

Strengths

Wide range of BMI.

Limitations

• Self-selection bias: Convenience sample with volunteers
• Limited generalizability: 98% Caucasian; not representative of general population
• Traveling to research center before measurement rather than spending night in facility (?); might affect RMR; did not discuss length of time of rest before measurement (was it long enough?)
• Did the researchers control for physical activity (56% had sedentary lifestyles and 46% had recreational physical activity; intensity??)
• Small sample sizes when divided into groups
• Limitations of study not discussed.