Decanalization of weight and stature during childhood and adolescence

Decanalization of Weight and Stature During Childhoodand Adolescence

JAMES LI,1 WON J. PARK,2 AND ALEX F. ROCHE3*1Computing and Telecommunications Services,2Department of Mathematics and Statistics, and3Department of Community Health, Wright State University, Division ofHuman Biology, Dayton, Ohio 45435

ABSTRACT The probabilities that children and adolescents alter theirpositions to non-adjacent canals (decanalization) on the current NCHSgrowth charts have been estimated. In this context, canals are the zonesbetween the major percentile levels (5th, 10th, 25th, 50th, 75th, 90th, and95th) that are shown on the charts. These probabilities were calculated forone- and 2-year intervals, beginning at each annual age from 3–16 years,using serial data for weight and stature from 329 male and 303 female par-ticipants in the Fels Longitudinal Study. The probabilities of particular typesof decanalizations, categorized by the initial canal and the direction of thechange in relative level (decrease/loss; increase/gain) are presented. Theprobabilities for 1-year intervals were very low and, therefore, they are notpresented. Additionally, probabilities were calculated for transitions from thecanal between the 5th and 10th percentiles to the zone below the 5th per-centile and from the canal between the 90th and 95th percentiles to the zoneabove the 95th percentile. Both these transitions are shifts to levels that areoutside the normal range, and are more common than decanalization. Theprobabilities of decanalization and of transitions from the normal range weregenerally larger for weight than for stature except for transitions from thecanal between the 90th and 95th percentiles to the zone beyond the 95thpercentile. The probabilities for both weight and stature tended to be largerfor changes toward the medians than for changes from the medians. Decana-lizations that involved increases in level were significantly related to ad-vanced skeletal maturation at ages younger than the usual age of peak heightvelocity and to retarded skeletal maturation at ages older than the usual ageof peak height velocity. Am. J. Hum. Biol. 10:351–359, 1998.© 1998 Wiley-Liss, Inc.

Weight and stature are common measure-ments in epidemiological surveys and clini-cal examinations of children and adoles-cents. These data are used to assess the nor-mality of growth for individuals and groupsby comparison with reference data such asthose presented in the U.S. National Centerfor Health Statistics (NCHS) growth charts(Hamill et al., 1977). These charts may beapplied to data recorded at a single exami-nation (status values) to judge present sizeand they may be applied to serial data toassess changes in relative levels. With thelatter approach, the changes between serialpoints are compared with the changes, dur-

ing corresponding intervals, in the majorpercentile lines (5th, 10th, 25th, 50th, 75th,90th, and 95th) that are displayed on thecharts. Children whose serial data remainin the same canal between a pair of majorpercentile lines, or deviate no further thanto the contiguous canal, are said to show

Contract grant sponsor: National Institutes of Health; Con-tract grant number: HD-12252

*Correspondence to: Alex F. Roche, M.D., Ph.D., D.Sc., WrightState University, Department of Community Health, Division ofHuman Biology, 1005 Xenia Avenue, Yellow Springs, OH 45387–1695. E-mail: [email protected]

Received 10 March 1997; Accepted 15 May 1997

AMERICAN JOURNAL OF HUMAN BIOLOGY 10:351–359 (1998)

© 1998 Wiley-Liss, Inc.

PROD #725

canalization. When the serial points changeto a canal that is not contiguous with theoriginal canal, the growth of the child is saidto show decanalization. A phenomenon thatis related to decanalization is the transitionof serial points from within the normalrange (5th to 95th percentile) to levels out-side this range.

Decanalization during infancy has beenanalyzed recently using data from many ofthe individuals included in the presentstudy (Park et al., 1997). Decanalizationduring infancy is more common for weightthan for recumbent length or head circum-ference from birth to 6 months, but not atolder ages. There are significant associa-tions between the probabilities of decanali-zation in weight and recumbent length inrelation to birth weight and mid-parentstature, respectively. It was found thatchanges toward the medians had muchhigher probabilities than changes from themedians and that the probabilities for vari-ous types of decanalization, categorized bythe initial canals and the direction of thechanges in levels, were smaller than thosefor transitions to outside the normal range.

The aim of the present analyses was toextend the previous analyses to childhoodand adolescence and to evaluate the asso-ciations between the probability of decana-lization in stature and the rate of skeletalmaturation.

MATERIALS AND METHODS

Subjects

The data analyzed were obtained fromWhite participants in the Fels LongitudinalStudy who were born in southwestern Ohioto families that differed widely in education-al level and occupation status (Roche, 1992).There were 4 sets of triplets and 14 sets oftwins among the Fels participants. The datafrom a random member of each monozygouspair were omitted from the analyses, aswere the data from 36 children with chronicpathological conditions that could have af-fected their growth. The birth dates forthese Fels participants ranged from 1930–1990 but, despite this wide range of birthdates, any secular trends were small andinconsistent when birth dates were groupedby decade. The numbers of children withdata at the beginning and the end of each2-year interval are given in Table 1.

Methods

Weight and stature were measured usingprocedures that closely matched those usedin the NCHS surveys and the recommenda-tions from a consensus conference (Hamillet al., 1977; Lohman et al., 1988). The mea-surements were made by two observersworking independently. The means (SDs) ofthe interobserver differences were 0.5 g (2.2g) for weight and 0.3 cm (0.2 cm) for stature.The means of the paired values recorded bythe two observers were used in the analyses.Interpolations were not made for missingdata. Exact percentiles, relative to the modi-fied data of Hamill et al. (1977), were calcu-lated for weight and stature using the EpiInfo computer program (Dean et al., 1994).These percentiles were used to assign thedata at each age to canals between the ma-jor percentile lines on the NCHS growthcharts (5th–10th, 10th–25th, 25th–50th,50th–75th, 75th–90th, 90th–95th) or tomarginal zones beyond these percentilelines (<5th, >95th). Wilcoxson tests showedthat there were no significant differences inselected percentile levels for weight or stat-ure between the present study sample andthe NCHS reference percentiles.

The probabilities of decanalization wereanalyzed for 2-year intervals beginning ateach annual age from 3–16 years. For easeof description, the possible types of decana-lization have been designated L-10, L-25 . . .for losses (decreases) in relative levels andG-10, G-25 . . . for gains (increases) in rela-tive levels (Table 2). These decanalizationtypes are identified by the first major per-centile line that is crossed as the relativelevels change. Thus L-10 denotes a change

TABLE 1. Number of children

Age atbeginning(years) Boys Girls3 299 2894 304 2935 306 2836 311 2917 317 2898 311 2889 301 275

10 301 27911 286 26612 286 25713 265 24714 260 23815 246 22416 244 224

352 J. LI ET AL.

from the canal between the 10th and 25thpercentiles, across the 10th percentile to thezone below the 5th percentile. Similarly, L-25 designates a change from the canal be-tween the 25th and 50th percentiles, acrossthe 25th percentile, to the canal betweenthe 5th and 10th percentiles. In addition,the designations L-5 and G-95 have beenapplied to transitions from the marginal ca-nals, 5th–10th and 90th–95th, respectively,to zones outside the normal range (<5th;>95th).

The analytic method was as follows: letXt, t 4 1,2, . . . , denote the observation ofweight or stature at age t and let Aj, j 4 1,2,. . . k, denote the percentile canals. Then thetransition probability in a 2-year time inter-val is defined by

pij (t) 4 pr{Xt+2 «Aj|Xt «Ai},

t 4 1,2, . . . , i,

j 4 1,2, . . . , k.

We estimated the transition probabilities[p̂ij (t)] by the empirical method as follows:

p̂ij ~t! = the estimate of pij ~t!

=#~Xt «Ai and Xt+2 «Aj!

#~Xt «Ai!

where # (A) is the number of subjects inSet A.

The estimates of p̂ij (t) for the proba-bilities pij (t) are on a scale of zero to 1.0where zero would indicate that the particu-lar decanalization or transition from thenormal range did not occur in any of thechildren studied and 1.0 would indicate that

this decanalization or transition occurred inall the children studied. These probabilitieswere calculated for each age interval usingdata from sex-specific groups of childrenidentified by their initial canal location.

It was considered likely that the rate ofmaturation would be related to the occur-rence of decanalization particularly in stat-ure. To investigate this possible relation-ship, the rate of maturation was judgedfrom the skeletal age of the left hand-wrist.These skeletal age assessments were madeby the Fels method (Roche et al., 1988) andwere expressed as relative skeletal age(RELSA). These RELSA values, which areunit free, were calculated as skeletal age/chronological age. A RELSA value >1.0 in-dicates maturation was more rapid thanusual and a value <1.0 indicates that matu-ration was slower than usual. The meanRELSA values for age groups were closelysimilar for the two sexes. These means var-ied from 0.99–1.03 with SDs that rangedfrom 0.06–0.13.

FINDINGSWeight

The probabilities of decanalization forweight are given in Tables 3 and 4. In thesetables, and in Tables 6 and 7, estimates areomitted for subgroups with fewer than 10participants. The probabilities of decanali-zation were arbitrarily considered to belarge when the estimated values were >0.1(10%). In the data for weight, large prob-abilities for G-10 were found for the inter-vals beginning at 4 and 5 years in males andat 3 years in females, and for L-10 duringthe interval 16–18 years in males, and forG-90 during the interval from 4–6 years infemales. Contrariwise, the probabilitieswere zero for L-50 and L-75 in each sex dur-ing all age intervals. The probabilities of de-canalization that were gains (increases inlevels) tended to be large during the age in-tervals that began at 3–5 years in each sex,at 6 years in males, and at 11–12 and 16years in females. Decanalizations that werelosses (decreases in levels) were more com-mon than those that were gains for the in-tervals 13–15 years in each sex and the in-tervals beginning at 7–9 years and 15 yearsin females. Decanalizations that were gainswere not observed in females during the ageintervals beginning at 8 and 15 years. De-canalizations that were losses were not ob-served for the intervals that began at 4–10

TABLE 2. The abbreviated designations of types ofdecanalization and movements from the normal range

Designation Early canal Later levelDecanalization

L 10 10th–25th <5thL 25 25th–50th <10thL 50 50th–75th <25thL 75 75th–90th <50thL 90 90th–95th <75thG 10 5th–10th >25thG 25 10th–25th >50thG 50 25th–50th >75thG 75 50th–75th >90thG 90 75th–90th >95th

Transition from normal rangeL 5 5th–10th <5thG 95 90th–95th >95th

DECANALIZATION IN CHILDHOOD AND ADOLESCENCE 353

years for males and at 3, 4, 6, and 12 yearsfor females.

The probabilities were compared for typesof decanalization that are changes towardthe median (L-75, L-90, G-10, and G-25) andfor those that are changes from the median(L-10, L-25, L-50, G-50, G-75, and G-90).There was little difference between theprobabilities for these groups of decanaliza-tions in females, but changes towards themedian tended to have larger probabilitiesin the males. The probabilities were alsocompared within pairs of decanalizationsthat began at equal percentile distancesfrom the medians, but differed in the direc-tions of the changes (losses; gains). Thesepairs are L-10 and G-90, L-25 and G-75, L-75 and G-25, and L-90 and G-10. In eachsex, the probabilities of decanalizationswere similar for L-10 and G-90, but theprobabilities for gains exceeded those forlosses in other pairs.

The probabilities of transitions in weightduring 2-year intervals from marginal ca-nals to positions outside the normal rangeduring 2-year intervals tended to be largerthan those for decanalization (Table 5). InTable 5, and in Table 8 for stature, the prob-abilities of transitions are given for combi-nations of age intervals. Thus, the values for2–5 years are for 2-year intervals beginningat 2, 3, 4, or 5 years, in combination. Theestimated probabilities of transitions inweight exceeded 0.20 for L-5 transitions inmales during intervals beginning at 10–13years and in females during intervals begin-ning at 6–16 years. The probabilities for G-95 transitions in weight exceeded 0.20 forintervals beginning at 2–13 years in malesand at 6–9 years in females. Large prob-abilities were more common after 9 yearsthan at younger ages in each sex for L-5transitions, but not for G-95 transitions. Inmales, small probabilities for weight (<0.10)

TABLE 3. Decanalization probabilities for weight during 2-year intervals in males

AgeDecreases in levels Increases in levels

L-10 L-25 L-50 L-75 L-90 G-10 G-25 G-50 G-75 G-903 0.06 0.00 0.00 0.00 0.10 0.09 0.00 0.02 0.03 0.054 0.00 0.00 0.00 0.00 0.00 0.17 0.06 0.05 0.03 0.005 0.00 0.00 0.00 0.00 0.00 0.20 0.04 0.01 0.04 0.056 0.00 0.00 0.00 0.00 0.00 0.08 0.04 0.02 0.01 0.047 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.07 0.00 0.008 0.00 0.00 0.00 0.00 0.00 — 0.00 0.00 0.00 0.029 0.00 0.00 0.00 0.00 0.00 — 0.03 0.00 0.00 0.04

10 0.00 0.00 0.00 0.00 0.00 — 0.00 0.00 0.00 0.0411 0.02 0.00 0.00 0.00 0.00 — 0.02 0.01 0.00 0.0512 0.02 0.01 0.00 0.00 0.00 — 0.02 0.01 0.00 0.0513 0.07 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.0014 0.00 0.03 0.00 0.00 0.00 0.09 0.07 0.02 0.05 0.0315 0.03 0.04 0.00 0.00 0.00 — 0.03 0.00 0.00 0.0316 0.12 0.00 0.00 0.00 0.00 — 0.03 0.00 0.00 0.00

The ages are those at the beginning of the intervals.In this and subsequent tables, the estimated probabilities are not presented when n < 10.

TABLE 4. Decanalization probabilities for weight during 2-year intervals in females


L-10 L-25 L-50 L-75 L-90 G-10 G-25 G-50 G-75 G-903 0.00 0.00 0.00 0.00 — 0.11 0.07 0.05 0.08 0.004 0.00 0.00 0.00 0.00 — 0.10 0.09 0.05 0.07 0.175 0.03 0.00 0.00 0.00 0.00 0.09 0.08 0.03 0.01 0.086 0.00 0.00 0.00 0.00 0.00 0.00 0.04 0.00 0.01 0.027 0.04 0.00 0.00 0.00 0.00 — 0.00 0.01 0.00 0.008 0.03 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.009 0.07 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00

10 0.06 0.00 0.00 0.00 0.00 0.07 0.00 0.00 0.00 0.0011 0.02 0.00 0.00 0.00 — 0.00 0.06 0.00 0.02 0.0012 0.00 0.00 0.00 0.00 0.00 — 0.09 0.04 0.02 0.0013 0.06 0.02 0.00 0.00 0.07 — 0.06 0.00 0.00 0.0014 0.07 0.03 0.00 0.00 0.00 — 0.03 0.00 0.00 0.0015 0.00 0.00 0.00 0.00 0.08 0.00 0.00 0.00 0.00 0.0016 0.03 0.00 0.00 0.00 — 0.09 0.03 0.01 0.00 0.00

The ages are those at the beginning of the intervals.

354 J. LI ET AL.

were found for intervals beginning at 14–16years for L-5 and G-95. In females, smallprobabilities for L-5 were noted for intervalsbeginning at 2–5 years and for G-95 for in-tervals beginning at 14–16 years.

Stature

The probabilities of decanalization forstature were zero in each age interval forL-90 in males and in females and for L-75 inmales, and they were generally small for L-50 in each sex and for L-10 and L-25 in fe-males (Tables 6 and 7). The probabilities ofdecanalization exceeded 0.10 for G-10 dur-ing intervals beginning at 4 years in malesand at 11 and 13 years in females and forG-50 and G-90 in males for intervals begin-ning at 12 and 14 years, respectively. Prob-abilities larger than 0.10 were not found forthese types of decanalization in the males.In each sex, the probabilities of G-50 andG-90 decanalizations were relatively largefor intervals that began at 11–16 years.

The probabilities of decanalization weresimilar for males and females. Consideringall types of decanalization in combination,the probabilities tended to be larger for in-tervals beginning at 14–16 years in malesand for intervals beginning at 11–14 yearsin females than for intervals beginning atother ages. Decanalizations towards the me-dians were more common than those fromthe medians in the females, but not in themales. In each sex, the probabilities of de-canalization were zero during most of theage intervals for all types of decanalization.There were few estimates for G-10 in eachsex, but many for G-75 in females and G-90in each sex.

The frequencies were compared for pairsof types of decanalization that werematched for the divergence of the initial ca-nal from the median, but differed in the di-rections of decanalization. In each sex, theprobabilities for decanalizations in stature

that were gains exceeded those for decana-lizations that were losses. Thus, the prob-abilities were larger for G-90 than for L-10,for G-75 than for L-25, for G-50 than forL-50, for G-25 than for L-75, and for G-10than for L-90.

The estimated probabilities for transi-tions in stature exceeded 0.2 for L-5 transi-tions in males during intervals beginning at10–16 years and in females for intervals be-ginning at 14–16 years (Table 8). The prob-abilities for G-95 transitions in stature ex-ceeded 0.2 in males and females for inter-vals beginning at 2–5 and 10–13 years.Transitions were uncommon (probability< 0.1) for some intervals. For stature, smallprobabilities of transitions were found forL-5 in males for intervals beginning at 2–9years and in females for intervals beginningat 2–5 years and at 10–13 years. Smallprobabilities for G-95 in stature were notnoted in either males or females.

Because the rate of maturation may berelated to the decanalization of stature, par-ticularly during pubescence, relative skel-etal ages (RELSA) were compared betweengroups in which decanalizations occurredthat were increases in the relative levels forstature (G-10, 25, 50, 75, 90; Group 1) andgroups in which there were decanalizationsinvolving decreases in relative levels (L-10,25, 50, 75, 90; Group 2). The differences inRELSA between Groups 1 and 2 were notstatistically significant for males during thepubescent period, but the participants inGroup 1 (increases in levels) tended to ma-ture more rapidly than those in Group 2 (de-creases in levels) for the intervals from 12–14 years and 10–14 years, but there was anopposite difference between these groups forthe interval from 14–16 years (Table 9). Infemales, the groups with decanalizationsthat were increases in relative levels from10–12 years and from 8–12 years tended tohave larger RELSA values than those withdecanalizations that were decreases in rela-tive levels. These differences in RELSA val-ues were statistically significant in femalesfor the interval from 8–12 years. There wasa non-significant tendency to an oppositedifference in RELSA between groups of fe-males with increases in relative levels from12–14 years and those with decreases inthese levels.

Because many of the groups used for theanalyses in Table 9 were small, other analy-ses were made after combining data for the

TABLE 5. Probabilities of transitions (L-5; G-95) forweight from a marginal zone to outside the normal

range during combinations of 2-year intervals

Age intervals(years)

Males FemalesL-5 G-95 L-5 G-95

2–5 0.11 0.26 0.08 0.136–9 0.17 0.21 0.21 0.25

10–13 0.28 0.27 0.26 0.1914–16 0.09 0.06 0.45 0.08

The intervals are identified by the age at the beginning of eachinterval.


two sexes. In these combinations of data,the age intervals chosen for males were 2years older than those for females to allowfor the sex difference in the timing of peakheight velocity (Guo et al., 1992). The differ-ences in RELSA between Group 1 (increasesin relative levels) and Group 2 (decreases inrelative levels), when data were combinedfor the intervals 10–14 years in males and8–12 years in females, showed those inGroup 1 were significantly more advancedin skeletal maturity (Table 10). There wasan opposite and significant difference in ma-turity between corresponding groups whendata were combined for the intervals from14–16 years in males and from 12–14 yearsin females. The combined data for the inter-vals from 12–14 years in males and from10–12 years in females did not show signifi-cant differences in maturity between thosewith decanalizations in stature that wereincreases in relative levels and those with

decanalizations that were decreases in rela-tive levels.

DISCUSSION

In practice, decanalization is recognizedwhen the plotted serial values for a childshift from a canal between major percentilelines on a growth chart to another canalthat is not immediately adjacent to the ear-lier canal, e.g., a change from the canal be-tween the 10th and 25th percentiles to thecanal between the 50th and 75th percen-tiles. Decanalization can also be recognizedwhen the serial points for a child cross twoor more major percentile lines on a growthchart. Additionally, transitions can occurfrom marginal canals (5th–10th percentile;90th–95th percentile) to zones outsidethe normal range which is commonlyconsidered to extend from the 5th to the95th percentile.

TABLE 6. Decanalization probabilities for stature during 2-year intervals in males


L-10 L-25 L-50 L-75 L-90 G-10 G-25 G-50 G-75 G-903 0.00 0.00 0.01 0.00 0.00 0.00 0.00 0.01 0.00 0.034 0.00 0.00 0.01 0.00 — 0.20 0.00 0.00 0.00 0.035 0.00 0.00 0.00 0.00 — 0.00 0.00 0.00 0.00 0.076 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.04 0.027 0.00 0.00 0.00 0.00 0.00 — 0.00 0.00 0.00 0.008 0.00 0.00 0.00 0.00 0.00 — 0.00 0.00 0.00 0.009 0.00 0.00 0.00 0.00 0.00 — 0.00 0.00 0.00 0.00

10 0.00 0.00 0.00 0.00 0.00 — 0.00 0.01 0.00 0.0211 0.00 0.00 0.00 0.00 0.00 — 0.06 0.00 0.03 0.0212 0.03 0.00 0.00 0.00 0.00 — 0.00 0.01 0.01 0.1313 0.03 0.00 0.00 0.00 0.00 — 0.06 0.00 0.02 0.0814 0.00 0.01 0.00 0.00 0.00 — 0.00 0.11 0.05 0.0415 0.04 0.06 0.00 0.00 0.00 — 0.08 0.04 0.05 0.0216 0.03 0.02 0.00 0.00 0.00 — 0.10 0.03 0.02 0.00


TABLE 7. Decanalization probabilities for stature during 2-year intervals in females

AgeDecrease in levels Increase in levels

L-10 L-25 L-50 L-75 L-90 G-10 G-25 G-50 G-75 G-903 0.03 0.00 0.00 0.00 0.00 — 0.00 0.00 0.03 0.094 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.105 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.036 0.00 0.00 0.00 0.00 0.00 0.06 0.00 0.00 0.00 0.007 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.008 0.00 0.00 0.00 0.00 0.00 — 0.00 0.00 0.00 0.069 0.00 0.00 0.00 0.00 0.00 — 0.00 0.01 0.01 0.06

10 0.02 0.02 0.00 0.00 0.00 — 0.00 0.01 0.00 0.0911 0.00 0.01 0.00 0.00 0.00 0.13 0.00 0.01 0.00 0.0312 0.03 0.00 0.00 0.00 0.00 0.06 0.08 0.03 0.02 0.0513 0.00 0.00 0.00 0.03 — 0.50 0.08 0.08 0.05 0.0514 0.00 0.00 0.00 0.05 0.00 — 0.05 0.02 0.03 0.0515 0.00 0.00 0.00 0.00 0.00 — 0.03 0.00 0.00 0.0016 0.00 0.00 0.02 0.00 0.00 — 0.00 0.00 0.00 0.00


356 J. LI ET AL.

In the present study, the reference per-centiles used were those displayed in theNCHS Growth Charts (Hamill et al., 1977).These U.S. National Survey data are appro-priate for the present analyses since the dis-tributions of weight and stature shown inthe NCHS Growth Charts do not differ sig-nificantly from those for the Fels Longitudi-nal Study data that were used in the pre-sent analyses. Furthermore, the rates ofskeletal maturation in the Fels sample areclose to those in NCHS national data (Rocheet al., 1971, 1974, 1976). This suggests thatthe timing of the pubertal growth spurt doesnot differ markedly between the Felssample and the NCHS cross-sectionalsample. It is important that the probabili-ties of decanalization be judged using refer-ence percentiles that are appropriate for thestudy population. If this is not the case, theestimates will be erroneous (Wright et al.,1991).

Decanalization was analyzed in relationto changes during 2-year intervals. Theprobabilities of decanalization for weightand stature in childhood and adolescencewere generally smaller than those for de-canalization during infancy, even thoughdecanalization during infancy was analyzedfor intervals of only 6 and 12 months (Parket al., 1997). This would be expected on abasis of clinical impressions, but there havebeen few, if any, previous studies of both agegroups from the same population.

The probabilities of decanalization inweight during childhood and adolescencetended to be high for L-10, G-10, and G-90and to occur during intervals that began at4, 5, 15, or 16 years. Some of this variationwith age could reflect the rather smallsample sizes and consequent selection bias.For example, G-10 can be observed only inthose between the 5th and 10th percentilesat the younger age. The larger probabilities

for L-10, G-10, and G-90 may reflect thecloser spacing of the near-marginal percen-tile levels. Decanalization involved de-creases in levels more commonly than in-creases in levels for intervals beginning at13 years in each sex and at 7 and 14 years infemales. Large probabilities were mainly forintervals beginning at 10–14 years in fe-males and for intervals beginning at 15–16years in males. This suggests an associationwith pubescence. Regression to the mediantended to be more common than decanaliza-tion from the median in the females but notin the males. The probabilities for transi-tions in weight from marginal zones to lev-els outside the normal range were largerthan those for decanalizations. This couldreflect the fact that smaller percentilechanges were required for these transitionsthan for decanalization.

The present finding that the probabilitiesof decanalization were larger for weightthan for stature at ages older than 2 years isin agreement with earlier findings for theperiod from birth to 6 months (Berkey et al.,1983; Park et al., 1997). It has been re-ported, however, that the probabilities of de-canalization do not differ between weightand length from 6–36 months (Park et al.,1997).

In the present data, the symmetry of theprobabilities of decanalization were com-pared for weight and for stature between

TABLE 8. Probabilities of transitions (L-5; G-95) forstature from a marginal zone to outside the normal

range during combinations of 2-year intervals

Age intervals(years)

Males FemalesL-5 G-95 L-5 G-95

2–5 0.08 0.58 0.09 0.246–9 0.03 0.16 0.10 0.16

10–13 0.21 0.28 0.09 0.2814–16 0.25 0.11 0.38 0.14

The intervals are identified by the age at the beginning of eachinterval.

TABLE 9. Differences within sex-specific groups inrelative skeletal age (skeletal age/chronological age;

RELSA) between those with decanalizations for statureinvolving increases in levels (Group 1: G-10, 25, 50,

75, 90) and those with decanalizations involvingdecreases in levels (Group 2: L-10, 25, 50, 75, 90)

Intervals(years) Groups N Mean (SD) SignificanceMales

12–14 1 9 1.02 (0.07) ns2 10 1.00 (0.05)

14–16 1 13 0.95 (0.05) ns2 15 1.09 (0.08)

10–14 1 9 1.01 (0.06) ns2 18 0.97 (0.09)

Females10–12 1 4 1.01 (0.12) ns

2 4 0.97 (0.08)12–14 1 8 0.96 (0.06) ns

2 4 1.03 (0.13)8–12 1 4 1.06 (0.04) 0.02

2 14 0.92 (0.10)

ns 4 non-significant at P 4 0.05.


pairs of decanalization types that differed indirection (decreases/increases) but not inthe distances from the initial canals to themedians. These pairs are G-90 with L-10,G-75 with L-25, G-50 with L-50, and G-25with L-75. The probabilities tended to belarger for the gains than the losses withineach pair. An earlier analysis of data for in-fancy did not show systematic differences inthe probabilities for types of decanalizationthat were gains and those that were losses(Park et al., 1997).

Decanalization towards the median wasmore common than decanalization from themedian for weight in males and for staturein females, but there were only small differ-ences between the probabilities of thesetypes of decanalization for weight in fe-males and stature in males. During infancy,however, decanalization towards the me-dian is much more common than decanali-zation from the median, especially duringthe first 6 months after birth (Park et al.,1997).

Relative skeletal age (RELSA) was usedas an index of maturity. It was hypothesizedthat, when decanalizations occurred, thoseadvanced in RELSA would tend to have de-canalizations in stature that were gainsearly in the pubescent period and to havedecanalizations that were losses late in thepubescent period. It was further hypoth-esized that those who are retarded in skel-etal maturity would tend to have decanali-zations that were losses early in the pubes-cent period and to have decanalizations thatwere gains later in the pubescent period.The differences in RELSA between groupscontrasted in the directions of decanaliza-tion for stature were in the expected direc-tions, but were significant only for femalesduring the interval from 8–12 years and forsome other comparisons based on data thatwere combined for the two sexes using age

groups that were 2 years younger for fe-males than for males to take into accountthe sex difference in the mean age of thepubertal spurt. These generally non-signi-ficant findings may be due to the incorpora-tion of data recorded during pubescence inthese parts of the growth charts with a dif-ferential representation of rapidly andslowly maturing children in various parts ofthe distribution (e.g., Wilson et al., 1987).

The probabilities of transitions in weightand stature from marginal canals to zonesoutside the normal range were larger thanthe probabilities of decanalizations withinthe normal range. Many of the changes thatwere classified as transitions were smallerthan those classified as decanalizations.The probabilities of transitions were esti-mated for groups of age intervals becauseonly about 5% of the sample formed the de-nominator for these estimates. These tran-sitions had probabilities >0.20 more com-monly for G-95 than for L-5 in males, butprobabilities >0.20 were more common forL-5 than for G-95 in females. These mar-ginal transitions are more likely to causeconcern when they are noted for age inter-vals when such transitions are uncommon(<0.10). For weight, these age intervals be-gan at 2–5 years in each sex and 14–16years in males. For stature, probabilities<0.10 were not noted for G-95. Such prob-abilities occurred for L-5 in males for inter-vals beginning at 2–9 years and in femalesfor intervals beginning at 2–5 years and 10–13 years.

The present results should be appliedcautiously in clinical situations. These find-ings are strictly applicable only to U.S.White children and adolescents who are freeof recognized major diseases and are mem-bers of middle class families. When a par-ticular type of decanalization is noted insuch a child, the probability of this change

TABLE 10. Differences within groups for sexes combined in relative skeletal age(skeletal age/chronological age; RELSA) between those with decanalization for

stature involving increases in levels (Group 1: G-10, 25, 50, 75, 90) and those withdecreases in levels (Group 2: L-10, 25, 50, 75, 90)

Intervals (years) Groups N Mean (SD) SignificanceMales 10–14 and 1 13 1.02 (0.06) 0.009

Females 8–12 2 32 0.95 (0.10)Males 12–14 and 1 13 1.02 (0.09) ns

Females 10–12 2 14 0.99 (0.06)Males 14–16 and 1 21 0.95 (0.06) 0.0001

Females 12–14 2 24 1.07 (0.10)

ns 4 non-significant at P 4 0.05.

358 J. LI ET AL.

in relative level can be checked using thepresent data for the same sex and age inter-val. Thus, a change in weight from a levelbetween the 25th and 50th percentiles to alevel below the 10th percentile during theinterval from 8–10 years should be regardedas clinically significant. Such a change oc-curred in only 1% of the females and not inany of the males included in the presentstudy.

LITERATURE CITEDBerkey CS, Reed RB, and Valadian I (1983) Longitudi-

nal growth standards for preschool children. Ann.Hum. Biol. 10:57–67.

Dean AG, Dean JA, Coulombier D, Brendel KA, SmithDC, Burton AH, Dicker RC, Sullivan K, Fagan RF,and Arner TG (1994) Epi Info, Version 6: A word pro-cessing, database, and statistics program for epidemi-ology on microcomputers. Atlanta, GA: Centers forDisease Control and Prevention.

Guo S, Siervogel RM, Roche AF, and Chumlea WC(1992) Mathematical modelling of human growth: Acomparative study. Am. J. Hum. Biol. 4:93–104.

Hamill PVV, Drizd TA, Johnson CL, Reed RB, andRoche AF (1977) NCHS Growth Curves for ChildrenBirth-18 Years, United States. DHEW Pub. No. (PHS)78–1650, Vital and Health Statistics, Series 11, No.165, National Center for Health Statistics, Washing-ton, DC.

Lohman TG, Roche AF, and Martorell R (eds.) (1988)Anthropometric Standardization Reference Manual.Champaign, IL: Human Kinetics Publishers.

Park WJ, Li J, and Roche AF (1997) The decanalizationof weight, recumbent length, and head circumferenceduring infancy. Am. J. Hum. Biol. 9:689–698.

Roche AF, Davila GH, and Eyman SL (1971) A compari-son between Greulich-Pyle and Tanner-Whitehouseassessments of skeletal maturity. Radiology 98:273–280.

Roche AF, Roberts J, and Hamill PVV (1974) SkeletalMaturity of Children 6–11 Years: United States.DHEW Pub. No. (HRA) 77–1642, Vital and HealthStatistics, Series 11, No. 140, National Center forHealth Statistics, Washington, D.C.

Roche AF, Roberts J, and Hamill PVV (1976) SkeletalMaturity of Youths 12–17 Years: United States:DHEW Pub. No. (HRA) 75–1622, Vital and HealthStatistics, Series 11, No. 160, National Center forHealth Statistics, Washington, D.C.

Roche AF, Chumlea WC, and Thissen D (1988) Assess-ing the Skeletal Maturity of the Hand-Wrist: FelsMethod. Springfield, IL: Charles C Thomas.

Roche AF (1992) Growth, Maturation and Body Compo-sition: The Fels Longitudinal Study 1929–1991. Cam-bridge, UK: Cambridge University Press.

Wright CM, Edwards AGK, Halse PC, and WaterstonAJR (1991) Weight and failure to thrive in infancy.Lancet 337:365–366.

Wilson DM, Kraemer HC, Ritter PL, Hammer LD(1987) Growth curves and adult height estimation foradolescents. Am. J. Dis. Child. 141:565–570.


Documents

Decanalization of weight and stature during childhood and adolescence