2. METHODOLOGY

2.1. Standardization of Laboratory Determinations

To insure that the studies conducted at the various laboratories around the world involved in UNU-sponsored research could be compared with one another, a standardization procedure was set up, under the direction of Christine Bilmazes who is the technician in charge of the MIT Human Nutrition Laboratories. Over the past three years Ms. Bilmazes has visited most of the participating laboratories to consult with laboratory personnel and advise on the procedures in use and to be used. Additionally, during this same time, four sets of samples, each including urine, feces, and protein sources, were sent to all the laboratories for analysis. The results of these analyses are summarized in Table I and discussed as part of the general discussion of research results. More detailed summaries were prepared for the participating laboratories.

Because of the importance of this aspect of the studies, and its fundamental nature in assessing individual laboratory results, a fifth set of samples was prepared during the workshop under the direction of Dr. Sheldon Margen, to be carried home by the participants, analysed, and reported back immediately. The results are shown in Table II.

TABLE II: RESULTS OF “NEXT” STANDARDIZATION EXERCISE

2.2 Skin and Miscellaneous Losses

Most investigators participating in UNU-sponsored studies have used 3 mg N/kg/day for skin and 2 mg/kg for miscellaneous N losses, as suggested by the 1973 FAO/WHO report, (2) in calculating true N balance. These estimates were derived mainly from studies conducted under moderate ambient temperatures and correspond to obligatory losses. The applicability of this figure for N balance studies conducted under conditions of high environmental temperatures and at protein intake in the submaintenance range was questioned, since increased sweating and N intake affect skin losses. Inoue et al. (paper 1) and Atinmo et al. (paper 10) provided data concerning this problem. Bourges also reported sweat N concentrations in subjects receiving different protein intakes and undergoing exercise. Table III summarizes their findings in comparison with results previously published by Huang and Lin (3) and Calloway et al. (4).

Calloway reported recent studies (5) which indicate that nitrates formed in the body, amounting to up 100 mg N, are lost daily in urine and feces, and not measured as N by the Kjeldahl Method.

These data suggest that 5 mg N/kg/day may not be an appropriate figure to use for miscellaneous losses when estimating “true” N balance. In particular, the estimate probably should be adjusted upward when ambient temperature is above the thermoneutral range since there is evidence that higher environmental temperatures are correlated with increased N losses in sweat, and further that these losses are compensated for by lower urinary losses of Nitrogen. This results in an underestimate of the dietary N required for zero balance unless the allowance for integumental loss is appropriately increased. In theory, the same effect can result from sweat losses associated with heavy physical activity. Yet, for individuals accustomed to such work in a tropical climate, these losses appear to be less than would be predicted from experimental studies with non-adapted, non-acclimatized subjects.

2.3 Standard UNU Short-Term Protocol

Many of the investigators (papers 6–11) followed the standard short-term protocol as it was formulated at the Costa Rica meeting in 1977 (1). This procedure for estimating protein requirement for a particular protein source and population is based on feeding each of several individuals several different, constant levels of protein for successive periods of 10 to 15 days each. These data are used to estimate individual responses and individual mean requirements which in turn are used to estimate the needs of the population. Details of this design are shown in Table IV.

Even for experiments that follow, in general, the above protocol there are many factors that influence the results. Several of the more important ones are listed in Table V. Since the effects of these factors can be very important, a point-by-point discussion of them is presented.

TABLE IV: THE STANDARD PROTOCOL

A number of individuals representative of a specific population (indexed by j)

For each individual:

Four or more periods at different intakes (indexed by i)

Each period 10 days or more at constant intake I_i

U_i   =   average urinary nitrogen over the last five days of period i

S_i   =   standard deviation of urinary nitrogen over the last five days of period i

F   =   average fecal nitrogen over entire study

M   =   estimated miscellaneous nitrogen losses

O_i   =   U_i + F_i + M = nitrogen output at level I_i

NB_i   =   I_i - O_i = nitrogen balance for level I_i

O   =   a + b . I   estimated response curve; a, b estimated by weighted regression of O_i on I_i (weighted by 1/S²_i)

For each individual (a_j, b_j)

R_j   =   a_j/(1-b_j) = estimated requirement for j^th individual

R   =   average requirement over all individuals

SD   =   standard deviation of requirements

Factors generally under control of investigator:

A-1 If successive levels are studied in a sequence from high to low, the intercept values may be higher and in addition the results are likely to be more variable than if the sequence is from low to high. The best way of avoiding the biases resulting from one or the other pattern is to randomize the levels with break periods in between.

A-2 If diet composition changes so that its quality differs at different levels of intake for a single experiment then the very meaning of the response curve and the zero balance intercept is unclear, since essentially one is examining a different dietary protein at each intake level. At best, if the variation in quality is slight, and if the dietary quality is poorer at the lower levels, the resulting N balances at these levels will, on the average, be proportionately lower than at the higher levels, and the opposite is true if the quality is higher at the lower levels. In the former case, the intercept will tend to be falsely low, and in the latter case, falsely high.

A-3 It is necessary that each laboratory maintain an ongoing quality control program, and that checks be made routinely with reference laboratories. High variability of determinations within a given laboratory reduces the usefulness of its results. Consistent differences among laboratories can be corrected for, but only if the investigators are aware of them.

A-4 Urinary N loss is increased with stress, whether of psychological or infectious origin. The result is an increase in protein requirements during the subsequent recovery period when N retention increases in a compensatory manner. Moreover, stress of infectious origin not only increases urinary nitrogen excretion, but also reduces food intake due to anorexia and changes in foods offered or selected. It may also reduce absorption if the gastro-intestinal tract is involved. In studies done under usual living and working conditions, any effect of stress that is compatible with continuation of normal activities can be considered an approximate estimate of the influence of usual environmental factors on the protein requirements of that population. However, the influence of clinical disease requiring hospitalization or confinement to bed should be excluded from population estimates and dealt with as a therapeutic problem.

A-5 Underestimation of dietary energy requirements in balance studies can result in values that are abnormally high unless a compensatory reduction in physical activity can occur. Conversely, excessive caloric intakes can result in high retentions and lead to an underestimation of the protein requirements of individuals in energy balance. Although such caloric deficiencies and excesses will eventually be reflected in the corresponding weight changes, these may not be obvious over the short duration of a balance study. If they do not occur with energy intakes clearly higher or lower than habitual intakes or approximately estimated normal requirements, adaptation must have taken place.

A-6 There is no doubt that careful screening of volunteers to eliminate infections and other disease problems, together with maintaining temperature control and relieving subjects of normal work and other responsibilities, will give more uniform results than studies with free-living individuals. For some purposes, studies under closely controlled conditions are preferable or even necessary, but for estimation of population requirements; studies on free-living subjects are necessary.

A-7 Similarly, exclusion of individuals who differ markedly from the norm will tend to produce more homogenous results than would be obtained if such individuals were included. This is basically a question of the purpose of the experiment -- whether it is to obtain a good estimate of the “normal” population requirement or to gauge the range of variability that exists within the population.

A-8 Some vitamin and mineral deficiencies can impair N retention and growth and can result in incorrect estimation of protein requirements.

Factors generally not under the control of investigator:

B-1 The appropriateness of using 5 mg/N/kg in adults for integumental and miscellaneous losses is discussed in the previous section.

B-2 There is some evidence that active individuals need less protein per kg than do those not active. Studies with a high proportion of very active subjects may give lower values than appropriate for the general population. The best indicator of the inherent activity level of an individual may be the number of calories necessary for energy balance.

B-3 For some individuals and in some populations, the necessary conditions of a balance study may in themselves constitute a stress that results in an apparent increase in protein requirements. This may be unavoidable, and the possibility should be recognized.

2.4 Long-term studies

Two papers were presented that studied individuals maintained for 2-3 months on constant intakes; these statistically examined the results to describe the nature of intraindividual variability. Rand (paper 2) examined 42 individuals in five separate experiments with intakes from .73 to 1.8 g Protein/kg body weight. He found that the daily urinary nitrogen of 25 (60%) of the individuals could be satisfactorily described in terms of a constant mean and random error (with a standard deviation of .74 g Nitrogen). The other 17 (40%) individuals had significant long-term trends (linear, quadratic or cubic). Examination of the data for autocorrelation, an indication of data on successive days being more alike than data an arbitrary time apart, found only 4 individuals (10%) who showed significant autocorrelation of values. These autocorrelations were calculated after the long-term trends were removed statistically, since such trends would appear as a false autocorrelation. Obviously, if an individual's values were steadily increasing, successive data would be similar. These findings are consistent with earlier work of Rand et al. (6), but are at variance with a report of Sukhatme and Margen (7), who described autocorrelation the rule rather than the exception in 5 subjects. Margen presented data from an additional 6 subjects maintained on a .42 g protein intake in a metabolic ward. Of these individuals, 5 (67%) showed an autocorrelation; however, all took several weeks to reach an apparent steady state. The differences between the two studies were attributed to: 1) the lower variability of the subjects under the very confined conditions in Margen's laboratory, and 2) Rand's removing of any long-term trend before calculating autocorrelation.

2.5 Very short-term balance studies

Several investigators have explored the use of procedures shorter than the 50–53 days of the standard UNU protocol for estimating the amount of protein in a diet required to achieve zero nitrogen balance. This work is based on the assumption that, although a steady state may not be achieved at each intake level, the balance value measured at a level is proportional to a steady state at that level.

Scrimshaw et al. (paper 4) presented the results of two studies. The first used one-day balances at each of three intake levels and showed that the variability involved is so large as to make this design useless. The second experiment used five different levels with two-day balance periods at each level. Four individuals followed both the ascending and the descending design. The data showed that most responses could be adequately fitted with a straight line, with the variability of the ascending design much less than that of the descending design. However, the intercepts calculated from data from the ascending order were consistently below those calculated from the descending design.

Bressani et al. (paper 5) presented data showing that his ascending design with two-day balance periods at three to five levels, preceded by three days of a protein-free diet at an initial level, gives good agreement in terms of zero balance intercept with standard protocol data for the same proteins.

While these results are very promising, it was felt that the zero balance intercept values of 6–8 subjects studied for two-day periods at four levels using the procedure of Bressani, equally divided between ascending and descending patterns, needs to be explicitly compared with the standard procedure. Such agreement must be adequately demonstrated before this methodology can be accepted as a substitute for the standard procedure.