Seasonality of growth and the relationship between weight and heightgain in children under three years of age in rural Malawi

K Maleta1,2,3, SM Virtanen2,3,4, M Espo2, T Kulmala2 and P Ashorn2,5

College of Medicine1, University Of Malawi, Blantyre, Malawi; Paediatric Research Centre2 and School of Public Health3, University ofTampere; Department of Epidemiology and Health Promotion4, National Institute of Public Health, Helsinki; Department of Paediatrics5,Tampere University Hospital, Finland

Maleta K, Virtanen SM, Espo M, Kulmala T, Ashorn P. Seasonality of growth and the relationshipbetween weight and height gain in children under three years of age in rural Malawi. Acta Pædiatr2003; 92: 491–497. Stockholm. ISSN 0803–5253

Aim: To describe the seasonal pattern of growth and analyse the relationship between weight andheight gain in children under 3 y of age. Methods: A population-based cohort of 767 children wasprospectively followed from birth until 36 mo of age in rural Malawi, southeast Africa. Weightand height measurements were collected at monthly intervals until 18 mo of age and quarterlythereafter. Gains in weight and height and prevalence of malnutrition in different seasons werecalculated. The relationship between weight and height gain was analysed using a series of corre-lation analyses. Results: Both weight gain and linear growth velocity showed an age-dependentseasonal pattern. After infancy, periods of maximal or minimal height increments systematicallyoccurred 3 mo after those for weight gain. The prevalence of malnutrition also followed a seasonalpattern, peaking a few months after periods of reduced growth. Despite the overall pattern, weightgain and subsequent linear growth were not correlated on an individual level. At any point,however, a child’s weight for height was directly, albeit weakly, correlated to height gain in thesubsequent 3-mo interval.

Conclusion: Growth of children under 3 y of age followed an age-dependent seasonal pattern.The poor correlation between children’s weight and height increments suggests that seasonalityaffected weight gain and linear growth through different mechanisms.

Key words: Cohort, growth, longitudinal, seasonality, Sub-Saharan Africa

Per Ashorn, Paediatric Research Centre, University of Tampere Medical School, Finn-Medi 3Building, FI-33014 University of Tampere, Finland (Tel. �358 3 215 8410, fax. �358 3 215 8420,e-mail. per.ashorn@uta.fi)

Childhood growth has been suggested to follow aseasonal pattern, both in low- and high-income coun-tries. According to most of the available evidence,weight gain tends to peak in autumn and winter andheight gain in spring and summer (1–6). In low-incomecountries, this phenomenon has usually been attributedto seasonal variation in food availability and theprevalence of infectious diseases (7). In the indus-trialized world, the emphasis has been on other climaticand environmental factors, such as changes in the lightand dark cycle and its influence on physical activity andgrowth hormone production (8).

Understanding the seasonality of growth and itsorigin would be especially important in low-incomecountries, where childhood malnutrition is often a majorpublic health problem. The seasonality itself wouldsuggest that some months in the year contribute more tothe burden of malnutrition, and that effective inter-vention strategies should take this into account when

addressing the problem. As periods of best or worstweight gain seem to precede those of maximum orminimum height increments, interventions addressingweight gain might also improve linear growth, i.e.alleviate stunting. Some studies have suggested thatweight and height gains peak and dip concurrently,instead of following each other (9). Furthermore, arecent Chinese study failed to demonstrate a temporalrelationship between weight and height gain at anindividual level, although the pattern was observed in apopulation of children aged 0 to 26 mo (10).

To analyse the seasonality of growth and therelationship between weight and height gain further,we longitudinally followed a population-based cohortof rural Malawian children from birth until 3 y of age.The aim was to describe the seasonal pattern of growthat different age groups and to investigate whetherchildren’s height gain was somehow associated withtheir earlier size or growth.

2003 Taylor & Francis. ISSN 0803-5253

���� ������ �� �� �� ����

Materials and MethodsStudy area and participantsThe study was carried out in Lungwena, southernMalawi, Southeast Africa. Approximately 17000inhabitants occupied a total of 4200 households in 23villages which formed the catchment area of a publichealth centre providing free primary health care.Subsistence farming and fishing formed the maineconomic occupations. The climate was monomodaland the staple food, maize, was grown during the rainyseason between November and March.

The subjects were the liveborn singleton offspring ofa cohort of 795 women recruited between June 1995 andSeptember 1996 into a child survival study. Owing to avery high enrolment rate, the study cohort comprisedapproximately 95% of all newborn children in the area.Informed consent was obtained verbally from eachpregnant woman before enrolment. The research pro-tocol was reviewed and approved by the MalawiNational Health Science Research Committee. Thedetails of recruitment and collection of data on socio-economic background, delivery events, feeding, mor-bidity and mortality have been described elsewhere(11–15).

Anthropometric measurementsEight trained research assistants made home visits tocollect anthropometric measurements at monthly inter-vals from birth up to the age of 18 mo and at 3-mointervals thereafter. Weight was measured using aspring scale (reading increments 10%) until the age of12 mo and thereafter with a battery-operated digitalbathroom scale (100 g increment). Length/height andmid-upper arm circumference (MUAC) were measuredwith locally constructed length/height boards and non-elastic tape measures, having reading increments of5 mm, and 1 mm, respectively. For length/height, thechildren were measured supine until they were able tostand, and erect thereafter.

The scales and length/height boards were checkedand calibrated regularly with internal standards. Toverify data accuracy, a random sample of 5–10% of themeasurements was taken twice by two independentresearch assistants.

Statistical analysisData entry and analysis were done with Microsoft Excel2000 and SPSS 10.0 computer programs. Anthropo-metric indices: weight for age Z score (WAZ), heightfor age Z score (HAZ) and weight for height Z score(WHZ) were calculated with EPI-INFO 6.04b software,using the World Health Organization (WHO) adoptedNational Centre for Health Statistics (NCHS) referencepopulation (16). Since birth measurements (as measuredwithin 48 h of birth) were available for only one-fourth

of the babies, the analysis was restricted to the agesfrom 1 to 36 mo.

For each child and time-point, the velocities ofweight, height, WAZ, HAZ, or WHZ gain werecalculated by dividing inter-measurement incrementby the time between two consecutive measurements(velocity expressed as gain/year). The calendar monthcontaining the mid-point date between two visits wasconsidered the month of measurement. Time pointswhen two consecutive anthropometric data were notavailable were not included in this analysis, i.e. nointerpolation was performed.

To test the temporal relationship of height, weightand weight for height gains, a series of zero order andpartial correlations were applied to the individual data.Anthropometric increments over 3-mo periods wereused in this analysis. The importance of variouscorrelates was assessed using Pearson’s correlationcoefficient (r). The aount of variation explained byany correlate was assessed using the coefficient ofdetermination (r2).

ResultsSuccess of follow-upOf the 767 liveborn children participating in the study,146 died and 62 dropped out before they reached 36 moof age. Measurements collected before their loss tofollow-up were, however, included in the analysis. Themedian (range) numbers of children measured andincluded in the analysis at each age were 532 (504 to562) and 487 (454 to 516), respectively.

Growth pattern of the childrenFigure 1 shows the growth pattern of the study childrenand the WHO reference population (16). Compared tothe NCHS children, the newborns in Lungwena were on

Fig. 1. Growth pattern of children younger than 3 y of age inLungwena. The solid lines indicate the median weight for age(WAZ), height for age (HAZ), and weight for height (WHZ) Z scoresfor Lungwena compared to the NCHS reference population (brokenline). The median (range) number of monthly measurementsincluded in the analysis at each age was 487 (454–516).

492 K Maleta et al. ACTA PÆDIATR 92 (2003)

average short but of reasonable weight. During infancy,there was a marked growth deviation from the referencecurves.

Linear growth faltering started immediately afterbirth, whereas weight gain faltering became evidentonly after 3 mo of age. Between 13 and 36 mo, medianweight for age and height for age remained atapproximately 1.8 and 2.5 SD units below the reference,respectively. Median weight for height was abovereference at birth, declined between 6 and 12 mo ofage and paralleled that of the reference thereafter (Fig.1).

Seasonality of growth and prevalence of malnutritionInitially, we analysed the mean weight and heightincrements in different calendar months at six 6-mo agegroups. The data on 13 to 36-mo-old children was thenpooled since all groups in this age-bracket followed anidentical seasonality pattern.

The mean (SD) rate of weight and height gain among1 to 6, 7 to 12 and 13 to 36-mo-old children was 6.3(6.4), 2.3 (5.9) and 2.0 (3.9) kg/y, and 23 (18), 13 (15),and 8 (9) cm/y, respectively. The corresponding meanrates for change in anthropometric indices were �1.4,�1.9, and 0.1 units/y for WAZ, �1.7, �0.9 and 0.0units/y for HAZ, and 0.2, �2.3, and �0.1 units/y forWHZ.

Figure 2 demonstrates the seasonal variation inWAZ, HAZ and WHZ increments in different agegroups. As shown, the seasonality pattern was agedependent. Among 1 to 6-mo-old babies, both WAZand HAZ declined most rapidly between December andApril (rainy season) and least in June–July, whereasWHZ increments varied from one month to another. At7–12 mo of age, no obvious seasonal pattern wasnoticed, although May and June still appeared to bethe months of best growth. Between 13 and 36 mo, bothweight and height increments followed a seasonalpattern, albeit different from each other. Weightincrements peaked in May–July and dipped in Decem-ber—January, whereas height increments were highestin August–October and lowest in March–April (Fig. 2).

The prevalence of malnutrition also followed anannual pattern when seasonality of growth was ob-served. The highest proportion of underweight (WAZ �2), stunted (HAZ � 2), or wasted (WHZ � 2) childrenwas noticed after several months of small weight,height, or weight-for-height increments, respectively.This seasonality was most obvious for wasting and leastremarkable for stunting. Among children aged 13–36 mo, the prevalence of wasting was approximately6% in March, but less than 3% in August to December(Fig. 2).

Relationship of height and weight gainTo analyse the predictors of height gain on individuallevel, we calculated Pearson’s correlation coefficients

between height, weight and weight-for-height valuesand gains at different 3-mo age intervals. Height gain atany interval was directly correlated to concurrentweight gain, weight for height Z score at start of thesame age interval and weight for height gain in theprevious interval (Table 1). In contrast, there was nocorrelation between height gain and weight gain in theprevious interval but height gain was inversely corre-lated to previous height gain. The Pearson’s correlationcoefficients, though statistically significant, were rela-tively weak, i.e. ranging from 0.14 to 0.40 (Table 1).

When the above variables were controlled for eachother (partial correlation), height gain in any ageinterval showed a statistically significant correlationonly to concurrent weight gain and the weight for heightZ score at the start of the same interval (Table 1).Calculating from the correlation coefficients, concur-rent weight gain (r = 0.24–0.44) explained 6–19% of thevariation in height gain in any age interval. Similarly,weight for height Z score at the beginning of the interval(r = 0.10–0.38) explained 1–14% of the variation inheight gain.

To identify the correlates of weight for height atvarious ages, we repeated the above analyses usingWHZ score as the constant variable in the correlations.As shown in Table 2, weight for height at any age wasdirectly correlated to weight and weight for height gainin the preceding interval and inversely correlated withprevious height gain. When controlled for all threevariables simultaneously, a positive correlation wasobserved only between weight for height Z score andweight gain in the preceding interval. Calculating fromthe correlation coefficients, weight gain in the preceding3-mo interval explained 7–38% of the variation inchildren’s weight for height at any age.

The correlation results remained essentially similar,when the analyses were restricted to individual 3-moseasons (Jan.–Mar., Apr.–June, July–Sept., Oct.–Dec.)(data not shown).

DiscussionIn the present cohort of rural children under 3-y of agein Sub-Saharan Africa, growth velocity showed age-dependent seasonal variation. Among the under 6-mo-old children, both weight and height gains were greatestbetween May and September. Between 6 and 12 mo ofage, there was no clear growth seasonality. After 12 mo,weight gains peaked in May to July and height gains3 mo later, between August and October. On anindividual level, height gain was associated with thechild’s weight for height at the start of the observationperiod but not directly with preceding weight gain.These associations were independent of the season orthe child’s age.

The impact of age on growth seasonality has beeninvestigated in a few studies elsewhere (10, 17, 18).

ACTA PÆDIATR 92 (2003) Seasonality of growth in rural Malawi 493

Fig

.2.S

easo

nalit

yof

grow

than

dm

alnu

triti

onin

thre

eag

egr

oups

.The

velo

city

ofw

eigh

tfo

rag

e(W

AZ

),he

ight

for

age

(HA

Z),

orw

eigh

tfo

rhe

ight

(WH

Z)

gain

indi

ffer

ent

cale

ndar

mon

ths

isill

ustr

ated

with

bars

(lef

tax

is)

and

the

prev

alen

ceof

unde

rwei

ght

(WA

Z�

2),s

tunt

ing

(HA

Z�

2),o

rw

astin

g(W

HZ�

2)w

ithlin

es(r

ight

axis

).T

hem

ean

(ran

ge)

mon

thly

num

ber

ofm

easu

rem

ents

incl

uded

inth

ean

alys

isis

234

(148

–346

)in

the

1-to

6-m

oag

egr

oup,

248

(124

–383

)in

the

7-to

12-m

oag

egr

oup

and

945

(889

–990

)in

the

13-

to36

-mo

age

grou

p.

494 K Maleta et al. ACTA PÆDIATR 92 (2003)

Most of the existing data are, however, compatible withour results and suggest that the pattern changes withage. Thus, periods of maximal weight and height gainscoincided with each other among Gambian infants (9).Among older children, height gains have been found tofollow weight gains, with a 3 to 4-mo lag period(1, 3, 10, 17, 19).

It is likely that the changing pattern of growthseasonality is influenced by the babies’ increasingdependency on the external environment and itsseasonal fluctuation as they grow older. During the first

6 mo of life, the growth of a breastfed baby is mostlyinfluenced by the child’s foetal conditions or themother’s nutritional status, and not so much on currentexposure to the environment. Seasonality is present, butit is tempered and modified, because it is mediatedthrough maternal health. After this period of initialgrowth, breastfeeding alone gradually becomes inade-quate to meet the metabolic needs of the child.Unfortunately, the complementary foods in rural Sub-Saharan Africa are often nutritionally poor and theconcomitant increase in the incidence of diarrhoea and

Table 1. Association of height gain at any age period and weight gain and weight for height.

Height gaininterval

Zero order correlation coefficients Partial correlation coefficients

Weight gainin previous

interval

Height gainin previous

interval

WHZ gainin previous

interval

Weight gainin interval

of height gain

WHZ atstart of

height gain

Weight gainin previous

interval

Height gainin previous

interval

WHZ gainin previous

interval

Weight gainin sameinterval

WHZ atstart of

height gain

1 to 3 mo 0.35** 0.26** 0.44** 0.38**3 to 6 mo 0.01 �0.31** 0.24** 0.23** 0.29** 0.06 �0.04 0.06 0.35** 0.28**6 to 9 mo 0.05 �0.32** 0.27** 0.16** 0.36** �0.03 �0.01 0.03 0.26** 0.29**9 to 12 mo 0.17** �0.27** 0.32* 0.08 0.39** �0.02 �0.00 0.02 0.23** 0.29**

12 to 15 mo 0.09 �0.15** 0.15** 0.13** 0.29** 0.08 �0.09 �0.08 0.24** 0.26**15 to 18 mo 0.14** �0.20** 0.23** 0.19** 0.26** 0.09* �0.11* �0.08 0.25** 0.19**18 to 21 mo 0.04 �0.14** 0.11* 0.24** 0.16** 0.06 �0.08 �0.05 0.30** 0.18**21 to 24 mo 0.02 �0.20** 0.12** 0.27** 0.30** 0.08 �0.11* �0.08 0.32** 0.27**24 to 27 mo 0.06 �0.18** 0.09* 0.21** 0.22** 0.05 �0.10* �0.04 0.26** 0.10*27 to 30 mo �0.04 �0.14** 0.02 0.25** 0.21** �0.03 �0.02 0.01 0.31** 0.29**30 to 33 mo �0.02 �0.26** 0.12** 0.20** 0.23** 0.04 �0.08 �0.04 0.26** 0.28**33 to 36 mo 0.11* �0.19** 0.20** 0.20** 0.20** �0.03 0.00 0.04 0.28** 0.22**

WHZ: weight for height Z score.Pearson correlation coefficients of height gain in a particular 3-mo interval to weight gain, height gain and weight for height gain in the previous

interval, weight for height at the start of height gain interval and weight gain in the same interval of height gain. Partial correlation: each variablecontrolled for the other variables.

* p-value of Pearson correlation coefficient less than 0.05.** p-value of Pearson correlation coefficient less than 0.01.

Table 2. Association between weight for height at any age and weight gain, height gain and weight for height gain.

Age of weight forheight measurement

Zero order correlation coefficients Partial correlation coefficients

Weight gain inprevious 3 mo

Height gain inprevious 3 mo

WHZ gain inprevious 3 mo

Weight gain inprevious 3 mo

Height gain inprevious 3 mo

WHZ gain inprevious 3 mo

3 mo 0.48** �0.20** 0.53** 0.46** �0.42** �0.26**6 mo 0.46** �0.27** 0.52** 0.62** �0.60** �0.53**9 mo 0.51** �0.23** 0.56** 0.52** �0.51** �0.47**

12 mo 0.48** �0.18** 0.48** 0.44** �0.42** �0.39**15 mo 0.42** �0.12** 0.42** 0.27** �0.25** �0.21**18 mo 0.45** �0.14** 0.47** 0.21** �0.18** �0.14**21 mo 0.41** �0.12** 0.44** 0.34** �0.33** �0.30**24 mo 0.49** 0.04 0.13** 0.88a �0.81a �0.84a

27 mo 0.44** �0.04 0.40** 0.32** �0.28** 0.25**30 mo 0.31** �0.02 0.31** 0.05 �0.03 �0.0133 mo 0.22** 0.01a 0.30** 0.35** 0.09* 0.40**36 mo 0.36** 0.02 0.18** 0.43** �0.14** 0.30**

WHZ: weight for height Z score.Pearson correlation coefficients of weight for height at a particular age to previous weight gain, height gain and weight for height gain. Partial

correlation: each variable controlled for the other two variables.* p-value of Pearson correlation coefficient less than 0.01.** p-value of Pearson correlation coefficient less than 0.05.a Disjunction in the reference population result in very high correlations between weight, weight for height gain and WHZ.

ACTA PÆDIATR 92 (2003) Seasonality of growth in rural Malawi 495

other diseases puts a further strain on the children.Beyond 12 mo of age, because these external environ-mental factors affect the children directly, the season-ality of growth becomes much more evident. In thepresent study, weight gains peaked after the harvest anddipped in the rainy season, when infections were mostcommon and food security poor.

The seasonality of growth among 13 to 36-mo-oldtoddlers offered a possibility to analyse the relationshipbetween weight and height gains among the studypopulation. Whereas the seasons of highest or lowestheight gain occurred on a population level system-atically 3 mo later than those for maximum or minimumweight increments, a similar relationship did not existon an individual level. These results agree with thoserecently observed in urban China (10), and suggest thatweight increment per se was not directly associatedwith height gain. Instead, the children’s height gain wasassociated with their weight for height at the start of theobservation period. However, preceding weight gainalso played a part since it was associated with thechildren’s weight for height at any point, but there wasstill considerable variation in height gain, which wasnot associated with variation in either weight gain orpreceding level of weight for height.

Previous studies on malnourished children haveshown that recovery starts with weight gain andcontinues with linear growth. In fact, there are someindications that height gain amongst recovering chil-dren would begin only when their weight for height hadreached a defined cut-off level (80% of the referencemedian) (18, 20). Whereas a similar cut-off value couldnot be determined in our study, the results suggest thatweight for height was associated with height gainamong an unselected but poorly growing populationof children under 3 y of age in rural Malawi. Thus,interventions improving weight for height values insuch a target population would be expected to have asubsequent influence on the children’s heights, as well.Importantly, however, children’s weight for heightexplained only a small fraction of the variation inheight gains, suggesting that the seasonality of lineargrowth is affected more by other factors. Issues thathave been suggested to affect height gain includephysical activity and social stimulation (8, 21). Theoccurrence of these conditions may also follow aseasonal pattern and it might prove useful to studytheir association with growth in Malawi.

Two issues need to be considered when interpretingthe current findings on the interaction between weightand height gain. Theoretically, the results could beaffected by a tendency of regression towards the mean,i.e. the smallest children in a population at any one pointgrowing faster than average, and vice versa. This trendcould be aggravated by the fact that growth is often nota linear process, but occurs in spurts (8). However,when controlled for other variables, height gains in anytwo consecutive age intervals showed very poor

correlation, contrary to what would be expected ifregression to the mean were common. This, togetherwith the consistency of the weight–height relationshipthroughout different seasons and age groups givescredibility to the findings.

In the current population-based cohort of under 3-y-old children from rural Malawi, growth velocity and theprevalence of malnutrition were dependent on season.After the first year of life, periods of best weight andheight gains succeeded each other with an approxi-mately 3-mo lag period. On an individual level, therewas a correlation between children’s weight for heightand their subsequent height gain, but this explained onlya small proportion of linear growth. This is compatiblewith a hypothesis that seasonality affects weight andheight gain through different mechanisms. Further dataare, however, needed to clarify the issue.

Acknowledgements.—We are grateful to the people of Lungwena, thestaff at the Lungwena Training Health Centre and our researchassistants for their positive attitude, support and help at all stages ofthe study. The study was funded by the Academy of Finland, the EmilAaltonen Foundation, the Foundation for Paediatric Research, theMedical Research Fund of Tampere University Hospital, theResearch Foundation of Mannerheim League for Child Welfareand the Research Foundation of the University of Tampere.

References1. Brown KH, Black RE, Becker S. Seasonal changes in nutritional

status and the prevalence of malnutrition in a longitudinal studyof young children in rural Bangladesh. Am J Clin Nutr 1982; 36:303–13

2. Tomkins AM, Dunn DT, Hayes RJ, Bradley AK. Seasonalvariation in the nutritional status of urban Gambian children. Br JNutr 1986; 56: 333–43

3. Nabarro D, Howard P, Cassels C, Pant M, Wijga A, Padfield N.The importance of infections and environmental factors aspossible determinants of growth retardation in children. In:Waterlow JC, editor. Linear growth retardation in less developedcountries. New York: Raven Press; 1988: 165–84

4. Pollit E, Arthur J. Seasonality and weight gain during the firstyear of life. Am J Hum Biol 1989; 1: 747–56

5. Marshall WA. Evaluation of growth rate in height over a periodof less than one year. Arch Dis Child 1971; 46: 414–20

6. Gelander L, Kalberg J, Albertsson-Wikland K. Seasonality inlower leg length velocity in prepubertal children. Acta Pædiatr1994; 83: 1249–54

7. Eveleth PB, Tanner JM. Worldwide variation in human growth,2nd ed. Cambridge: Cambridge University Press; 1976; 190

8. Waterlow JC. Relationship of gain in height to gain in weight.Eur J Clin Nutr 1994; 48: Suppl: S72–S74

9. McGregor IA, Rahman AK, Thompson B, Billewicz WZ,Thomson AM. The growth of young children in a Gambianvillage. Trans R Soc Trop Med Hygiene 1968; 62: 341–52

10. Xu X, Wang WP, Guo ZP, Cheung YB, Kalberg J. Seasonality ofgrowth in Shanghai infants born in 11 consecutive years. Eur JClin Nutr 2001; 55: 714–25

11. Kulmala T, Vaahtera M, Ndekha MD, Koivisto A-M, Cullinan T,Salin M-L, et al. Inadequate socio-economic support for goodhealth in rural Malawi. East Afr Med J 2000; 77: 168–71

12. Kulmala T, Vaahtera M, Ndekha M, Cullinan T, Salin M-L,Ashorn P. The relationship between antenatal risk characteristics,place of delivery and adverse delivery outcome in rural Malawi.Acta Obstet Gynecol Scand 2000; 79: 984–90

496 K Maleta et al. ACTA PÆDIATR 92 (2003)

13. Vaahtera M, Kulmala T, Hietanen A, Ndekha M, Cullinan T,Salin M-L, et al. Breastfeeding and complementary feedingpractices in rural Malawi. Acta Paediatr 2001; 3: 328–32

14. Vaahtera M, Kulmala T, Maleta K, Cullinan T, Salin M-L,Ashorn P. Epidemiology and predictors of infant morbidity inrural Malawi. Paediatr Perinat. Epidemiol. 2000; 14: 363–71

15. Vaahtera M, Kulmala T, Ndekha M, Koivisto A-M, Cullinan T,Salin M-L, et al. Antenatal and perinatal predictors of infantmortality in rural Malawi. Arch Dis Child 2000; 82: 200–4

16. World Health Organisation.. Measuring Change in NutritionalStatus: guidelines for assessing nutritional impact of Supple-mentary feeding programmes for vulnerable groups. Geneva:WHO; 1983

17. Jalil F, Kalberg J, Hanson LA, Lindbland BS. Growth dis-turbance in an urban area of Lahore, Pakistan related to feeding

patterns, infections and age, sex, socio-economic factors andseasons. Acta Pædiatr 1989; 350 Suppl: 44–54

18. Costello AM de L. Growth velocity and stunting in rural Nepal.Arch Dis Child 1989; 64: 1478–82

19. Kalberg J, Jalil F, Lam B, Low L, Yeung CY. Linear growthretardation in relation to the three phases of growth. Eur J ClinNutr 1994; 48: Suppl 1: S25–44

20. Walker SP, Golden MHN. Growth in length of childrenrecovering from severe malnutrition. Eur J Clin Nutr 1988; 42:395–404

21. Golden MHN. Is complete catch-up possible for stunted mal-nourished children? Eur J Clin Nutr 1994; 48 Suppl 1: S58–71

Received July 2, 2002; revisions received Dec. 17, 2002; acceptedDec. 18, 2002

ACTA PÆDIATR 92 (2003) Seasonality of growth in rural Malawi 497