• Assessment
    In healthy individuals, what is the prediction accuracy and maximum overestimation and underestiamtion errors compared to measured resting metabolic rate when using the Dietary Reference Intakes for Energy (DRI) or RDA?
    • Conclusion

      There are no studies reporting the accuracy within 10% using the DRI or RDA.  When the TEE factors are released in the 2002 DRI publications, a re-evaluation of the evidence will be needed.  (Grade V)

      Basal metabolic rate is approximately 60% of total energy expenditure (TEE) in healthy individuals.  There is limited evidence to suggest that current RDA guidelines overestimate the RMR portion of total energy expenditure.  If RMR is overestimated, then factors used to account for energy expenditure above RMR (i.e., physical activity) increase the estimation error.  (Grade IV -prior to June 2003)

How did we identify the best research?

Research Definitions

Adjusted body weight

Can be calculated by the following formula:   Adjusted wt (kg) = [(actual body wt – ideal wt) X 0.25] + ideal wt.

Age classifications:    

Adults – Is different from Medical Subject Heading (MeSH) “adult” and will represent persons, ages 19 – 79 years.

Young, Middle-aged or Older adults – In the evidence summaries, these terms reflect a particular study population and when used will report the age range or mean plus/minus standard deviation of the particular study so identified.

Older adults-When used in the Conclusion statement, this term will represent studies that identified persons ages 65 years and older as a dominant characteristic of the population (or a group within a sample) studied.

Aged is defined under the MeSH term “Aged” and indicates 65-79 years.

Oldest Old is reflected under the MeSH term “Aged, 80 and older.”

Design Quality:

Studies identified as “high quality” or “strong design” (i.e., a “plus” quality rating) had to identify or discuss individual characteristics of weight and age since both are intervening variables that are on the pathway that determines accuracy of measured or predicted RMR.  In addition, diseases allowed or excluded needed to be identified since certain diseases are associated with metabolic rate variations.  In addition, studies had to address indirect calorimeter protocol adherence in the following areas:

  1. machine calibration
  2. 20-30 minute rest before measurement if traveling to a measurement center or to discuss procedures prior to single measurements (e.g., machine acclimation measurements)
  3. steady state (e.g., pre-determined group mean covariance, elimination of erratic measurements and/or ongoing monitoring; depending on study discussions reporting “adherence to study protocol” was potentially eligible)
  4. measurement length
  5. exercise restrictions in healthy adults the day prior to measurements or identifying/monitoring movement restrictions/restlessness in critically ill patients
  6. fasting (ideally, specifying fasting length > 8 hrs but accepted “overnight”) with an exception for studies including patients on IV, parenteral or enteral feedings.

Because of the health ranges of the populations studied, evidence analysts had difficulty distinguishing certain aspects of initial steady state measures and any ongoing monitoring that occurred but may have been unreported in the written study description.  Thus, mentioning steady state qualified as meeting this criterion in the methodologically-sound protocol evaluation.  In addition, an exception to the exercise criterion was given to studies measuring energy expenditure in older adults because of a reduced likelihood of participating in significant physical activity the day prior to an energy expenditure measurement.  All studies evaluated reported a procedure that excluded subjects taking “medications that affect metabolic rate” or identified stability in post-menopausal women using hormone therapy.

Expected years of life lost

The difference between the number of years a person would be expected to live if he/she was not obese and the number of years expected to live if the person were obese.

Statistics

The differences between measured and predicted resting metabolic rate (RMR) can be reported with various statistics.  Some statistics evaluate the difference between predicted and measured RMR at one point in time; other statistics are concerned with the difference between predicted and measured RMR changes over time. Additionally, some studies use statistics to report how the measured RMR of groups of people differ from the predicted rate, whereas other studies are concerned with how the measured RMR of individuals in a particular group will differ from the predicted RMR.  Simplified ways in which studies have represented the differences between predicted and measured RMRs are discussed.

These notations are used in the equation discussions of that follow:
Mg is the measured group average RMR
Mi is the measured individual RMR
Prmr is the predicted RMR (group or individual)

Absolute Simple Differences

Reporting absolute differences between measured and predicted metabolic rates provides information about the measurement range of differences (i.e., overestimates and underestimates).  When reporting RMR, the difference is translated into kcals/d.

An absolute simple individual difference is calculated by subtracting one amount from the other (see below).

A statistical approach that reports group mean prediction error uses a Student’s t-test to identify the bias (i.e., the accuracy of each individual’s prediction) and gives a standard deviation (i.e., the absolute precision or consistency of the difference).  This represents an estimate of how far off any given member of a group’s estimated energy needs will be.  This approach averages overestimates and underestimates.  Thus, individual factors influencing over- and under-prediction errors (e.g., group member characteristics, conditions of how RMR was measured) are important to understand when comparing group mean prediction errors.

Individual

Group

Mi- Prmr (ind) or Prmr (ind) -Mi

Mg- Prmr (group) or Prmr (group) -Mg

Relative Error Differences (Proportions) and Percents

Relative measures allow comparisons without regard to the magnitude of the measurement range yet these measures give information about the importance of the error.  This type of measure is a proportion because the observation of the difference between two measures in an individual (or group mean measures) that are reported in the numerator is linked to the denominator by a measure of the same individual or group at a certain time point.

A relative error difference of predicted RMR from measured RMR gives significance about the magnitude of the error within an individual or compared to a group of individuals with different RMRs.  For example, a 200-kcal/d error may be less important in a healthy subject who requires 2,700 kcals/d (i.e., a proportion of 200 kcals /2,700 kcals) whereas this error becomes more important if it occurred in someone on an 800-kcal/day weight reduction diet (i.e., a proportion of 200 kcals /800 kcals).

Relative errors are sometimes expressed as percentages.  To convert a relative error into a percentage error, multiply by 100.  In the healthy subject described above, the relative error would be approximately 7% of total daily kcalories whereas the error for the individual on an 800-kcals diet would be 25%.

Individual

Group

(Mi- Prmr (individual)) / Prmr (individual)

(Mi- Prmr (individual)) / Mg  

 (Mi- Prmr (individual)) / Prmr (individual) *100

(Mi- Prmr (individual)) / Mg *100

 (Mg- Prmr (group)) / Prmr (group)

(Mg- Prmr (group)) / Mg

 (Mg- Prmr (group)) / Prmr *100

(Mg- Prmr (group)) / Mg * 100

 

In summary it is important to distinguish between the following characteristics:

  • Is the measured group RMR or the measured individual RMR the main concern?
  • Is the divisor the measured group RMR or the predicted RMR?

Researchers determine how to analyze the data based on whether they want to evaluate the accuracy of a predictive equation or the accuracy of a measurement and whether they want to apply the results to individuals or a population (i.e., groups of individuals).  Careful analyses of these differences are needed when reviewing the evidence analysis worksheets. 

Relative Ratio

A ratio is the value obtained by dividing one quantity by another and the numerator and denominator do not have to be mutually exclusive.  Measured individual RMR is divided by predicted RMR of that individual or a measured group mean RMR can be divided by the predicted group mean RMR.  In evaluating measured to predicted ratios in groups, this is reported as “mean to prediction ratio” (or percent mean to prediction ratio if multiplied by 100). 

Individual

Group Difference

Mi / Prmr (individual) or Mi / Mg

Mg / Prmr (group) or Prmr (group) / Mg

95% Limits of Agreement Analysis (Also known as 95% Probability of Individuals within a Reference Interval)  represents a reference interval for individual resting metabolic rate differences (measured and predicted) and identifies the 95% expected probability that new individuals selected from a similar population would have difference scores (i.e., measured minus prediction) within the limits of agreement range.  Sample sizes of at least 50 individuals are needed in a study for the sample limits of agreement to be precise for the population.

Root-mean squared prediction error (RMSPE) determines how close the predicted RMR value for each individual is to the actual measured RMR.  This approach more accurately presents the accuracy of the prediction equation for each person (i.e., individual prediction error bias), rather than as a mean of the entire sample.  Thus, this approach more accurately reflects individual over- and under-estimations.

WEIGHT CLASSIFICATION DEFINITIONS

Classification of Overweight and Obesity by BMI for Young – to Middle-aged Adults (19-59 y)*

 

BMI (kg/m2)

Obesity Class

Underweight

<18.5

 

Normal

18.5-24.9

 

Overweight

25.0 – 29.9

 

Obesity

30.0 – 34.9

I

 

35.0 – 39.9

II

Extreme Obesity

>/= 40

III

*      National Heart, Lung, and Blood Institute.  NHLBI Obesity Education Initiative Expert Panel. Clinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults.  The evidence report.  Bethesda, MD: National Institutes of Health, National Heart, Lung, and Blood Institute (NHLBI), June, 1998.

Classification of Overweight and Obesity by Percent Body Fat Body Fat Guidelines from the American Dietetic Association

Classification

Women
(% Fat)

Men
(% Fat)

Normal

15-25

10-20

Overweight

25.1-29.9

20.1-24.4

Obese

> 30

> 25

Extremely Obese*

>40

>39

*The distinction between obesity and extreme obesity was clarified by use of the same proportion used between normal and extreme obesity in the BMI classification:  Extreme Obesity (BMI 40) / upper limit of BMI Normal weight (BMI 24.9) = 1.6.