2. 746 BROWN ET AL.
procedures commonly used to evaluate lac- trict, Bangladesh, the rural setting ofa variety
tation. of health-related activities of the Interna-
The difference in an infant’s weight before tional Centre for Diarrhoeal Disease Re-
and after nursing has been assumed to equal search, Bangladesh (formerly the Cholera
the weight of milk consumed, but few studies Research Laboratory). The geographic, so-
have attempted to validate this “test-weigh- cial, and demographic characteristics of the
ing” technique or to compare it with other area have been described previously (2, 3).
methods of estimating the production or con- The individual studies were performed in the
sumption of human milk. Although simple to home of each study subject.
perform in clinical wards, the test-weighing
technique itself becomes impractical for 24-h Study subjects
studies in field settings. Thus, more simplified Participants in the ward studies were re-
or abbreviated techniques of investigation cruited from a large group ofwomen enrolled
must be developed and evaluated. Further- in a social service program located in a
more, the complete assessment of lactation “bustee” area (semiurban slum) in outlying
performance must include some indication of Dacca. After learning of the purposes and
the composition of nutrients in the milk, ne- design of the study, interested women were
cessitating the extraction of a representative transported to the ward to visualize the study
Downloaded from www.ajcn.org by on August 22, 2009
sample of the day’s production. Knowledge site; consenting subjects were then accepted
of whether a given sample of milk is repre- for participation in future studies.4 The 61
sentative of the full day’s production requires women ranged in age from 14 to 39 yr (me-
the analysis and comparison of levels of nu- dian 20 yr), were all primiparous except three
trients in the milk at different times of the women, and were in good, general health.
day. Specific studies were therefore under- None was using hormonal contraceptives at
taken to address these issues in preparation the time of the study. Their infants’ ages
for subsequent studies of lactation perform- ranged from 1 to 9 months. In return for their
ance among relatively undernourished participation the subjects and their infants
women of two communities in Bangladesh. received routine preventive and curative
health services as well as remuneration in lieu
Methods and results of potential earnings. The 13 infants taking
part in studies designed to validate the test-
Study sites weighing technique were selected from inpa-
All ward studies were accomplished in an tients recovering from severe protein-calorie
eight-bed metabolic ward situated at the Chil- malnutrition at the Children’s Nutrition Unit.
dren’s Nutrition Unit (CNU) in urban Dacca, The village study subjects were selected
Bangladesh. The ward was separated from from a group of mothers who were already
the rest of the hospital, and the study room participating in a separate study of infants’
was furnished simply to appear more like a dietary intakes, and who were found to be
village home than a hospital ward. The study particularly interested in and cooperative
subjects nursed their infants openly, as is with the earlier investigations. The women
customary in Bangladesh; no special nursing were 17 to 42 yr of age (median 25 yr), in
areas were provided for the women. Meals good health and had from one to eight living
were prepared for the mothers according to children (median three), including the nurs-
fixed dietary protocols (to be described in ing infants who ranged from 6 to 29 months
detail in subsequent communications); but of age. Several additional women (usually
the women were responsible for basic relatives of field workers) were enrolled in
housekeeping chores and for the care of their the studies comparing 12- and 24-h milk con-
infants while staying on the ward. The sumption, since it was particularly difficult to
mother-infant pairs stayed in the ward for as
little as 2 days for “outpatient studies” and The research protocols were approved by the re-
spective Committees on Human Volunteers of the School
for as long as 6 wk for “inpatient studies.” of Hygiene, Johns Hopkins University, of the Cholera
The village studies were completed in sev- Research Laboratory and of the Bangladesh Medical
eral villages of Matlab Thana, Comilla Dis- Research Council.
3. STUDIES OF HUMAN LACTATION 747
arrange 24-h, home-based studies in the con- A minimum of nine validation studies was
servative village society. completed by each ofthe six clinical assistants
The methods and results of the individual responsible for the test-weighing of infants
parts of the study will be presented together during longitudinal, clinical studies of lacta-
to facilitate comprehension. A list of each of tion performance. The accuracy of the
the individual studies, as well as their number “weighed-feeding” procedure, as determined
and location is found in Table I. by the mean percentage of a known weight
of ingested milk that could be measured by
Validation of test-weighing technique test-weighings ranged from 89.4 to 98.6% for
Quantification of infants’ consumption of the individual clinical assistants (Table 2).
mother’s milk has traditionally been accom- The SDS ofthe recoveries ranged from 4.8 to
pushed by the test-weighing technique. In the 24.4% for the different assistants. The mean
present study, diapered infants were weighed ± SD overall recovery of milk ingestion dur-
before and after feedings by a clinical assist- ing all 64 validation studies was 94.9 ± 13.2%.
ant using a balance calibrated daily against The variation in the recoveries tended to be
known weights. The diapers were covered by greater when smaller amounts of milk were
plastic pants and were not changed until the consumed.
second weighing was completed. The ward Fourteen similar studies of single feedings
Downloaded from www.ajcn.org by on August 22, 2009
balance (Detecto Infant Scale) was sensitive were completed in the village subjects’ homes
to 5 g; the balance used in field studies was to define the accuracy of the test-weighing
a modified solution balance (Heavy Duty technique under field conditions. The recov-
Solution Balance, Ohaus) sensitive to I g. ery of a known amount of ingested milk
During the validation studies the infants were ranged from 56.2 to 117.8% in the field setting
fed by spoon from cups of milk that had been with a mean ± SD of 88.6 ± 14.4%.
preweighed by a second study assistant on a
balance (Dial-O-Gram Balance, Ohaus) sen- Comparison of the weights of milk consumed
sitive to 0.1 g. After the infants had been fed
and the weights of milk extracted
by the mother or by a study assistant other
than the individual responsible for weighing In order to determine whether the two
the infants, the cup of milk was reweighed. different methods of estimating the amount
The infant’s weight change was then ex- of milk production provide similar data, val-
pressed as a percentage of the change in the ues obtained from the same woman by test-
weight of milk in the cup. Three studies were weighing and by mechanical extraction were
eliminated from consideration because either compared over short intervals of time (less
a sizeable amount of milk was spilled or the than 8 days). During the period of extraction
baby regurgitated. all milk was extracted from both breasts by
TABLE 1
Summary of methodological evaluations, locations of studies,
and numbers of studies and subjects
. . No. Of mothers
Purpt.c of .tud Ltcatin ?So. of .tudie
. or ntant
Validation oftest-weighing . Urban ward 64 13
Rural homes 14 14
Comparison of test-weighing and mechanical extrac- Urban ward 167 72*
tion
Determination of density of milk Urban ward 511 60
Comparison of amount of milk consumed in 12- and Urban ward 250 28
24-h periods Rural homes 12 12
Prediction of human milk consumption from age of Rural homes 341 52
infant, frequency and duration of feedings
Determination of changes in volume and composition Urban ward 7
of extracted milk by side of extraction and time of
day
* Mothers participating as both inpatients and outpatients are considered twice.
4. 748 BROWN ET AL.
TABLE 2
Comparison of the change in infant weight before and after feedings with the weight of milk ingested
Average amount of milk Average wt change of ‘Recovery of ingested milk by
Clinical Assistant No of studies
ingested infant weighed-feeding technique
g g ‘:
1 10 160 147 91.1 ± 8.4*
2 9 113 109 89.4± 24.4
3 9 153 148 96.1±8.2
4 9 154 148 96.4 ±4.8
5 9 146 140 94.4 ± 9.8
6 18 142 141 98.6 ± 13.8
All studies 64 144 139 94.9 ± 13.2
* Mean ± SD.
a mechanical breast pump (Egnell) at 3-h Density of extracted milk
intervals and pooled for the 24-h period. Two To estimate the amount of milk consumed
groups of women were studied during either by a single infant on multiple days, the least
inpatient or outpatient periods in the clinical disruptive measurement procedure should be
ward. During the 1 18 inpatient studies of 23
Downloaded from www.ajcn.org by on August 22, 2009
utilized. Although test-weighings are a more
mother-infant pairs, the total amount of each suitable technique than extraction of milk,
day’s test-weighings measured during some extraction would be necessary periodically to
or all of the 6 days either before or just after determine the concentration of nutrients in
a day of mechanical extraction was compared the milk. To estimate the infant’s consump-
to the weight of that extracted milk. During tion of nutrients, one could then multiply the
the 3-day outpatient studies of 49 mothers concentration of nutrients in extracted milk
and infants, the data from the 2 days of test-
by the amount of milk consumed on those
weighings were compared to those for the
days when nursings were weighed. This cal-
intervening day’s extracted milk. Paired data
culation assumes little day to day variation in
were analyzed by substracting the test-weigh-
the concentration of nutrients.
ing data for a given day’s study from the Whereas test-weighings measure the
respective extraction data for the same
weight of milk consumed, the concentrations
woman during the same study period. As
of nutrients are generally expressed per units
presented in Table 3, those mean differences of volume. The amount of milk measured
were always positive and usually statistically
during test-weighings must be divided by its
significant during the inpatient studies. In density to calculate the volume consumed.
other words, mechanical extraction yielded To learn whether the density of milk is con-
slightly higher estimates of milk production stant for all women the weights and volumes
than did test-weighings. The 3-day outpatient
of extracted milk were compared during 5 11
studies yielded similar results, but the differ-
separate 24-h ward studies of 60 women. The
ences were smaller than those measured dur- mean ± SD density for all studies was 1.036
ing the inpatient studies. Whereas the mean ± 0.017 g/ml, the grand mean for individual
(± 1 SE) weight of extracted milk was 717 women was 1 .038 ± 0.01 1 g/ml. There was
(± 21) g, the weight of milk measured by test- no statistically significant correlation between
weighings averaged 7 14 (± 1 8) and 678 (± I 8)
maternal age or infant age and the density of
g, respectively, on the day before and the day
extracted milk.
after extraction. Only the difference on the
day after extraction was statistically signifi-
Comparison of the amounts of milk consumed
cant (p < 0.0 1 If the test-weighings
). from all
during 12 and 24 h
of the paired studies were corrected for the
amount of underestimation as determined in Ideally, quantitative studies of milk con-
the earlier recovery studies, the amount of sumption should take place in the subjects’
milk production as estimated by either test- homes with as little disruption as possible of
weighing or extraction would be similar. the mothers’ routine activities and feeding
5. STUDIES OF HUMAN LACTATION 749
TABLE 3
Comparison of the amount of milk produced by individual women in 24 h, as
estimated either by extraction of milk or by test-weighings, by day
of test-weighing before or after day of
“paired” extraction
Day before or after extraction
Study
I 2 3 4 5 6
One day of extraction followed
by 6 days of test-weighings
Amount ofrnilk measured 715.8 695.8 694.2 716.0 708.9 704.4
during test-weighings (g/
day)
Mean difference* 96.3t 70.3t 8 1 .9f 60. It 58.7t 68.8t
SE ofdifference 26.0 14.6 16.2 17.8 16.8 15.8
No. of paired studies 26 50 55 55 49 51
Six days of test-weighings fol-
lowed by I day of extrac-
tion
Amount of milk extracted in 694.5 7 12.3 7 10. 1 69 1 .4 689.3 729.7
paired studies
Downloaded from www.ajcn.org by on August 22, 2009
Mean difference* 48.6t 29.5 33.Ot 5l.7t 46.8t 48.2
SEofdifference 12.1 15.1 14.4 13.2 12.6 28.5
No. of paired studies 63 60 63 63 57 20
a Mean difference between extracted amount less test-weighing amount during paired studies of individual women
(g/day).
t Paired t tests (p < 0.01).
:1:
Paired t tests (p < 0.001).
patterns, and without potentially anxiety-pro- daytime ranged from 46 to 58% during mul-
yoking changes in surroundings. However, it tiple studies of 28 individual women (mean
is not logistically feasible nor socially accept- ±SD, 52 ± 3%). All but two of the SDs of the
able to accomplish large numbers of 24-h proportions for studies within individual
studies of milk intake in subjects’ homes, so women ranged between 3 and 10%.
investigations taking place in the field must
Prediction of the amount of milk consumption
be abbreviated. Therefore, preliminary stud-
ies comparing 12-h intakes to actual 24-h
f rom age of infant andfrequency and
duration offeedings
intakes were undertaken to determine what
proportion of the day’s total intake was con- Since the test-weighing technique is cum-
sumed between 6 AM and 6 PM These 24-h bersome for field studies, an attempt was
studies were completed in I 2 village homes made to identify a more easily obtainable
and the data obtained were compared to data estimate of the quantity of milk consumed in
from 250 studies of 28 women enrolled in the a 12-h period. During an 8-month period,
ward studies. three hundred forty-one 12-h weighed feed-
The proportion of the entire day’s intake ing studies were completed for 52 infants in
that was ingested between 6 AM and 6 PM the field. The field assistants timed each com-
ranged from 37 to 72% (mean ± SD, 53 ± plete feeding (both breasts) to the nearest
9%) in the 1 2 studies of village subjects. Since whole minute. The relationship between the
the number of villagers who consented to the number and total duration of feeding, as well
24-h studies in their homes was limited, the as infant age (independent variables) and the
proportions were also calculated for a sam- amount ofmilk consumed in 12 h (dependent
pling of outpatients who were enrolled in the variable) was analyzed by stepwise multiple
ward studies on at least 5 different days. The linear regression (4) with the hope of iden-
woman-specific mean proportions of the full tifying independent variables that accurately
day’s intake that were consumed during the predict the amount of milk consumed.
6. 750 BROWN ET AL.
The infant’s age, the frequency of feedings, when all three were considered simultane-
and the total amount of time at the breast ously (multiple r = 0.69), they were together
were significantly correlated with the amount able to explain only about 50% of the varia-
of milk consumed (Tables 4 and 5). The age tion in milk consumption. The SE (Sy.x) of
of the infant was negatively correlated with the estimate of the amount of milk consumed
the frequency of feeding (r = -0.2 1 ), the by an individual was 81.8 g. Therefore the
amount of time nursing (r = -0.2 1), and the 95% confidence limits of the estimate was ±
amount of milk consumed (r = -0.34). The 54.8% ofthe mean amount ofmilk consumed.
number of feedings and time at the breast
were both positively correlated with the Composition and volume of milk extracted
amount of milk consumed (r = 0.63 for both) from both breasts during 24 h
and with each other (r = 0.82). Although the In order to determine an optimal sampling
ability ofthe independent variables to predict procedure for measuring the concentration of
the amount of milk consumed was enhanced nutrients in and the volume of extracted hu-
TABLE 4
Amount of mother’s milk consumed (g) during 12-h studies by age of infant,
total amount of time s uckling, and frequen cy of nursing
Downloaded from www.ajcn.org by on August 22, 2009
Fre quency of feedi ngs (n/ I 2-h study)
. Total time at O5 6-7 8+ All
Ageoftnfant -
breast
Mean Mean Mean Mean
n n n n
amount amount amount amount
6-Il 0-59 11 224* 1 274 0 12 228
60-89 2 402 24 333 10 410 36 358
90+ 0 4 402 8 453 12 436
Total 13 252 29 340 18 430 60 348
12-17 0-59 47 228 5 287 2 328 54 237
60-89 9 318 35 343 16 328 60 334
90+ 2 261 3 370 19 414 24 396
Total 58 243 43 337 37 372 138 307
18+ 0-59 61 193 9 244 0 70 200
60-89 11 243 29 294 18 348 58 301
90+ 0 2 296 13 352 15 344
Total 72 201 40 282 31 350 143 256
All 0-59 119 210 15 260 2 328 136 217
60-89 22 288 88 324 44 355 154 327
90+ 2 261 9 368 40 402 51 390
Total 143 223 112 319 86 376 341 293
a Amount of milk consumed (g).
TABLE S
Results of multiple regression analysis relating age of infant, total amount of time suckling, and frequency of
nursing (independent variables) to amount ofmilk consumed (dependent variable) during 12-h studies* (n = 341)
Partial SD of partial Partial
Independent Mean ± SD regression regres.ston F p
variables
coefficient coefficient )df= 1.337)
<0.001
Age of infant (mo) 16.3 ± 4.9 -4.74 0.92 26.5
Amount oftime 66.0 ± 24.4 1.48 0.32 2 I.2 <0.001
at breast (mm)
No. offeedings 6.1 ± 2.1 17.30 3.80 20.7 <0.001
Dependent
Variable
Amount of milk 292.7 ± I 12.3
consumed (g) ____________ ______________
a Multiple r = 0.668, SE (Sy.x) = 81.84, p < 0.00 I.
7. STUDIES OF HUMAN LACTATION 751
a
Cu
S
t...
1 0
U
1t o
>
1 2;
.
8
#{149} t_
- 0
U
Downloaded from www.ajcn.org by on August 22, 2009
E
8
, 0
-
a
U
E
- .
a
0 0 , 0 1,
8 2 p.. #{149} O *
a a ‘‘
E
a
E
a
§
a
z
0 #{149} 0 #{149} 0
p.. ,p *
8. 752 BROWN ET AL.
man milk, milk was collected separately from Although the time-related changes for lactose
each breast every 3 h during the course of concentrations and volume were not statisti-
single 24-h studies of seven women. The vol- cally significant for the group as a whole,
umes and weights ofthe milks were measured there were significant interactions between
and well mixed aliquots were then stored at time and individuals for these two variables.
-20#{176}C before analysis for nitrogen, lactose, In other words, there were statistically sig-
and fat. Nitrogen concentration was deter- nificant diurnal changes in the concentration
mined by a semimicro-Kjeldahl procedure of lactose (F = 3.38, p < 0.001) and volume
(5) with an accuracy of 94.2 ± 8.4%, lactose (F = 1.80, p = 0.03) within individual subjects
was measured by the method of Folin (6) but the times of those changes were not con-
with an accuracy of 99. 1 ± 5.4% and total fat sistent for all individuals in the study group.
was quantified gravimetrically after extrac- The magnitude of change in lactose concen-
tion by the Rose-Gottlieb procedure (5) with tration was not great as the minimums were
an accuracy of 100.5 ± 1.9%. greater than 90% of the maximums for all but
The data for the concentration of nutrients one subject. However, some individuals had
and the volume of milk were summarized to as much as a 2-fold difference in the volume
determine changes related to the time of sam- of milk produced during different intervals of
pling (Fig. 1) and side of extraction (Table the 24-h period.
Downloaded from www.ajcn.org by on August 22, 2009
6). Analyses of variance (4) were then com- The differences between the volume of
pleted as shown in Table 7. The time-related milk produced by the right and left breast
differences were statistically significant in the were of statistical and biological significance
case of nitrogen (F = 3.82, p = 0.003) and fat (Table 6). As a group the differences were
(F = 5.30, p < 0.001). Whereas the minimum not great, but for some women the less pro-
nitrogen concentration was slightly more than ductive breast yielded only 65% ofthe volume
90% of its maximum, the minimum fat con- of the more productive one. Although the
centration was less than 65% of its maximum. concentrations oflactose and fat in milk from
TABLE 6
Effects of side of extraction on concentration of nitrogen, fat, and lactose in, and volume of extracted breast milk
from seven women
Nitrogen concentration (mg/ . .
Fat concentratton (g/dl) Lactose concentration (g!dl) volume (ml)
dl
Woman
- Left breast Right breast Left breast Right breast Left breast Right breast Left breast Right breast
1 148 156 1.79 2.14 7.31 7.50 57 67
±1 1* ±6 ±0.67 ±0.60 ±0.33 ±0.40 ±15 ±20
2 167 174 2.22 2.69 8.10 8.04 48 60
±6 ± I1 ±0.88 ±0.74 ±0.3 1 ±0.27 ± 10 ±7
3 165 147 2.23 2.15 8.03 8.06 35 54
±1 1 ±15 ±0.60 ±0.42 ±0.20 ±0.38 ±13 ±9
4 158 169 2.20 2.31 7.94 8.16 58 58
±15 ±22 ±0.62 ±0.58 ±0.46 ±0.34 ±8 ±15
5 137 138 1.74 1.77 8.28 8.25 29 35
±5 ±5 ±0.67 ±0.45 ±0.57 ±0.36 ±6 ±9
6 140 158 1.53 1.64 7.25 7.38 76 31
±20 ±15 ±0.46 ±0.67 ±0.33 ±0.29 ±12 ±9
7 171 139 2.71 3.18 7.85 8.12 35 91
±8 ±6 ±1.02 ±0.71 ±0.62 ±0.23 ±13 ±22
All 155 155 2.05 2.27 7.82 7.93 48 56
±17 ±18 ±0.78 ±0.76 ±0.54 ±0.45 ±19 ±23
a Mean ±
SD.
9. STUDIES OF HUMAN LACTATION 753
TABLE 7
Results of analysis of variance for volume and three components of extracted breast milk with factors, side of
extraction, time of day, and interactions
Factor Source ofvariation OF Mean Square F Significance
Nitrogen
Main effects Individual 6 1812 13.88 <0.001
Breast I 4 0.03 0.863
Time 7 498 3.82 0.003
2-Way interactions Individual xbreast 6 1286 9.86 <0.001
Individual x time 42 120 0.92 0.605
Breast X time 7 132 1.01 0.438
Lactose
Main effects Individual 6 2.1 16 29.91 <0.001
Breast 1 0.323 4.56 0.038
Time 7 0.666 0.94 0.486
2-Way interactions Individual x breast 6 0.065 0.92 0.493
Individual x time 42 0.239 3.38 <0.001
Breast X time 7 0.102 1.44 0.214
Fat
Downloaded from www.ajcn.org by on August 22, 2009
Main effects Individual 6 3.308 12.26 <0.001
Breast 1 1 .203 4.46 0.041
Time 7 1.430 5.30 <0.001
2-Way interactions Individual x breast 6 0. 195 0.72 0.635
Individual x time 42 0.505 1.87 0.023
Breast X time 7 0.162 0.60 0.753
Volume
Main effects Individual 6 1880 15.91 <0.001
Breast 1 1930 16.34 <0.001
Time 7 106 0.90 0.5 19
2-Way interactions Individual x breast 6 35 1 1 29.72 <0.001
Individual X time 42 213 1.80 0.030
Breast X time 7 256 2.17 0.057
each breast were also statistically different, natural setting are most likely to estimate the
the magnitude of those differences was rela- usual lactation performance with accuracy.
tively unimportant when grouped data were The issues of cost, comfort of study subjects,
analyzed. Even when individual women were and the level of technical sophistication re-
considered, the differences between milk col- quired must also be considered in planning
lected from each breast were not great, except such studies.
for two women whose milk from one breast Test-weighing is the procedure used most
contained about 80% of the concentration of commonly to estimate the amount of milk
fat as compared to milk from the other breast. consumed by the nursing infant. The proce-
dure is technically simple, requires minimal
Discussion equipment and training of personnel, and
interferes relatively little with the usual proc-
The evaluation of lactation performance is ess of nursing. However, since the amount of
of fundamental importance to nutritionists milk consumed at a single feeding may be
concerned with the health of infants and their small, the measurement error can introduce
mothers, particularly in the lesser developed variability that is a sizeable proportion of the
countries. The techniques for such evaluation amount consumed. Nevertheless, in the pres-
are limited by the central dilemma that the ent studies the mean recovery of a known
impact of the study procedures themselves on amount ofingested milk attained a fair degree
lactation remains uncertain. At best, one can of accuracy, even under field conditions, if
compare a variety of techniques and assume multiple studies were undertaken. The test-
that those procedures that are least invasive weighings consistently underestimated the
and that require the fewest changes from the amount of milk consumed, probably because
10. 754 BROWN ET AL.
small amounts of milk were spilled or regur- can be weighed during a 12-h period rather
gitated. Insensible water loss of the infant than during a complete 24-h period. On the
between weighings undoubtedly also ac- average, approximately 52% of the 24-h in-
counts for some of the underestimation. Fi- take was consumed during the daytime pe-
nally, any urine or stool passed after the nod by infants of mothers from our popula-
initial weighing and inadvertantly not in- tions, both in field and ward studies. Al-
cluded in the fmal weighing would also re- though there was considerable within and
duce the recoveries. The range of variability between subject variability in the proportion
of the technique also depends on the care of milk consumed during daytime hours, that
with which the clinical assistants perform variability could be reduced by performing
their tasks. In the present studies some assist- multiple studies. Whether such abbreviated
ants were clearly superior to others in that studies are acceptable depends on the nature
regard. and the design of research proposed. Field
The test-weighing procedure has been studies of lactation performance that were
compared to other methods of estimating completed in rural West Africa used a similar
milk production in previous studies. Coward technique of 12-h test-weighing to estimate
et al. (7) found that their technique of isotope milk consumption (10). In those studies,
dilution by breast milk ingestion yielded which took place between 7 AM and 7 PM,
Downloaded from www.ajcn.org by on August 22, 2009
higher estimates of milk consumption than less than 50% of the full day’s intake was
weighed-feedings in six infants. In the present consumed during the 12-h studies. The dif-
study we found that extraction of milk by the ferences in the proportions consumed during
mechanical pump yielded an estimate of milk daytime hours may relate to the fact that a
production that was about 6 to 7% higher different 12-h interval was studied or may
than that measured by test-weighings within indicate culture-specific differences in breast-
the same week. Hytten (8) also found that feeding practices. It would seem necessary to
mechanical extraction yielded 6% more milk estimate the conversion factor directly for any
than the sum of weighed feedings plus man- population to be studied by an abbreviated
ual extraction of residual milk in studies that technique.
took place on consecutive days (8). It is pos- The attempt to estimate milk consumption
sible that mechanical extraction recovers by indirect methods was generally unreward-
more milk than is actually consumed by the ing in that the majority ofvariability in intake
infant or that test-weighings systematically data was unexplained by the independent
underestimate the amount of milk produced. variables chosen. Unless better explanatory
Since the difference in the estimates obtained variables can be identified we believe that the
by the two techniques was similar to the estimation of milk ingestion of an individual
difference between test-weighings and actual must continue to rely on a more direct tech-
consumption observed during the “recovery nique of assessment, such as those techniques
studies,” the latter hypothesis seems more mentioned above. However, similar indirect
likely to be correct. techniques may be of value in estimating the
In order to determine the density of human amount of milk consumption by a group of
milk, the volumes and weights of multiple infants from a given population. Again, the
extracted samples were measured. The spe- usefulness of the technique will depend on
cific gravity of human milk has been reported the purpose of the data collection. The mu!-
(9) as 1.032, which is similar to the present tiple regression analysis was of additional
findings. Since the densities did not change interest in explaining some of the factors
with infant or maternal age for our sample, affecting lactation performance in Banglade-
most of the variability appears to be ex- shi mothers. For example, milk intake tended
plained by individual differences and mea- to decrease as the infants became older. In-
surement error. Therefore, when correcting creased total duration of suckling and fre-
milk weight to volume or vice versa it would quency of feedings were both independently
be preferable to use the conversion factor as associated with higher levels of milk con-
determined for the woman in question. sumption in children of all ages. This latter
To facilitate the collection of data on lac- finding is at odds with the results of the West
tation performance in field studies, feedings African studies reported by Whitehead et a!.
11. STUD1ES OF HUMAN LACTATION 755
(10). Although they stated that the variation influence the concentrations of’ f’at and nitro-
in total milk intake between individuals at gen in opposite directions. Nims et al. (13)
any given age could not be explained by the also reported that maximal fat and protein
frequency of feedings, they did report that concentrations generally occurred at different
the amount ofmilk intake fell with infant age times of day, but the actual values were not
and was accompanied by a gradual reduction presented. The differences in the productive
in the frequency of feedings. We cannot ex- capacity ofthe two breasts were most impres-
plain the differences in these two sets of data. sive for volume and fat concentrations. Be-
The primary concern of nutritionists inves- cause the differences in volume were so great
tigating lactation performance is to determine it appears that milk must be extracted from
the total amount of nutrients ingested by the both breasts if the extraction data are to be
infant and simultaneously lost by the mother. used as estimates of the volume of milk pro-
Both the volume of milk consumed and the duced. However, the concentrations of nutri-
concentration of nutrients in that milk must ents in milk from each breast were not very
be measured. Since it is the entire day’s milk dissimilar, especially if data from a group of
that is of interest, the milk that is collected subjects are of interest. However, if the con-
for laboratory analysis must be representative centration of milk fat from individual sub-
of the whole day’s production. The technical jects is the object of study, specimens should
Downloaded from www.ajcn.org by on August 22, 2009
difficulties that interfere with the accurate be obtained from both breasts. Recently, Hall
measurement of milk nutrients have been (16) reported that the concentrations of fat in
discussed by Hytten and Thomson (I 1) who the fore-milk and hind-milk from the breast
concluded that “there is no escape from the suckled second in the course of a nursing are
necessity to obtain complete 24 hour samples greater than in the respective fractions from
if misleading measurements are to be the breast suckled first. Unfortunately, we
avoided.” Hytten (12) also claimed that milk did not record which breast was extracted
must be collected from both breasts since first in our own studies. It is possible that the
substantial differences may exist in the vol- differences observed between breasts reflect
ume and composition of milk from each consistent biases in which breast was pre-
breast. sented first for extraction.
The data from the present study are con- Several recent publications have reviewed
sistent with Hytten’s claims, but some quali- the numerous proven and theoretical advan-
fications are justified. Although the concen- tages of breast feeding ( 17, 18). Because of
trations of macronutrients in milk changed our interest in the nutritional well-being of
significantly throughout the day, particularly children and their mothers, particularly in
when individuals rather than groups were the lesser developed countries, we have re-
considered, only the changes in fat concentra- cently undertaken studies of the lactation
tions were of relatively great magnitude. If performance of marginally nourished women
one were interested only in lactose concentra- in Bangladesh. The present methodological
tion, for example, a randomly obtained sam- evaluations were initiated in an effort to de-
ple would not be too misleading in that its termine the appropriate techniques for those
lactose concentration would probably be less investigations. The data reported should be
than 5% different from the daily mean con- of similar use to other individuals planning
centration for that individual. The time-re- studies of human lactation or evaluating data
lated changes in fat concentration that were from such research.
observed in this study are consistent with the The authors appreciate the collaboration of the staff
findings of other investigators (13, 14) and of the Save the Children Fund, U.K., Children’s Nutri-
emphasize the necessity of complete collec- tion Unit, and the efforts of the field workers of the
Matlab Research Area. We also thank Drs. R. B. Sack,
tions to assess the total daily fat (and energy)
G. G. Graham, W. B. Greenough, and M. M. Levine for
production accurately. On the other hand, support and advice during the planning and analysis of
Lauber and Reinhart (15) reported variations these studies. The encouragement of M. R., M. B., S. B..
in the protein and lipid concentrations rang- and N. B. is also gratefully acknowledged.
ing from 5 to 15% and chose to ignore those
References
differences in their study of women in the
1. Anonymous. WHO/UNICEF Meeting. Lancet
Ivory Coast. The time of day appeared to 1979:2:841-3.
