seasonal fylde of lancashire 1957-1981

Journal of Epidemiology and Community Health, 1989, 43, 330-342

Seasonal prevalence ofmajor congenital malformationsin the Fylde of Lancashire 1957-1981J P BOUND,' P W HARVEY,2 AND B J FRANCIS3From "the Department of Paediatrics, Victoria Hospital, Blackpool; 2the Department of Pathology, RoyalLancaster Infirmary, Lancaster; and 3the Centre for Applied Statistics, University of Lancaster, Lancaster.

ABSTRACT The seasonal prevalence of major congenital malformations was studied in a prospectivesurvey of 88 449 children born in the circumscribed Fylde of Lancashire to residents there over 25years. Ascertainment was thought to be as complete as was practically possible because cases wererecorded daily by one, and for 17 years the only, paediatrician and a very high rate of necropsies wasmaintained. The number of malformations were classified by month of maternal last menstrualperiod and seasonal variation was assessed by three statistical models. Neural tube defects showed asignificant seasonal variation in month of last mentrual period but not in month of birth. From May1956 to April 1968, when the prevalence of neural tube defects was high (5 5 per 1000 total births),conceptions were significantly more common in December to May. For anencephaly alone thefigures were not significant, but spina bifida and cranium bifidum were more common in March toMay. From May 1968 to April 1981, when the prevalence of neural tube defects fell below thenational average, the significant variations disappeared. Seasonality for spina bifida and craniumbifidum was seen only in "singles" (cases with no other major lesion), but for anencephaly it was seenonly in "multiples" (cases with other lesions). The three types of cardiac septal defect and persistentductus each showed a higher prevalence of conceptions at some time during May to October. Incontrast the commonest group ofcyanotic cases showed no such pattern but with greater numbers inwinter. There was evidence of a seasonal variation for bilateral renal agenesis and for vesicouretericreflux as ascertained.

Seasonal prevalence in an aetiological factor for certain malformations of the central nervoussystem, cardiac and urinary systems.

This paper examines the seasonal prevalence, usingmonth of maternal last menstrual period, ofbabies with major congenital malformations bornover 25 years to residents in the Fylde ofLancashire.

Seasonal changes in prevalence are one pointer toenvironmental factors in causation, and interest inthem has continued since the demonstration byMcKeown and Record' in 1951 that more cases ofanencephaly were born in autumn and winter.However, a modern paediatric textbook2 discussingseasonal variations states: "conflicting reports havecome from different regions in one country and fromdifferent countries, so that much more information isrequired".

Contradictions may arise for several reasons.Firstly, the inevitable selection of cases in hospitalseries on which some reports have been based.3Secondly, inadequate ascertainment may occur inpopulation studies from reliance on notification ofcases. The resultant problems of under-reporting and

inconsistencies in diagnosis have been discussed forcentral nervous system malformations.4 Theparticular difficulties with cardiac malformationswhere many cases are not recognisable at birth wereoutlined by Hay.5 Another source of inadequateascertainment is a low rate or poor quality ofnecropsies, especially for perinatal deaths wheremalformations may be otherwise unsuspected ormisdiagnosed. It has been shown that the prevalenceof cardiac malformations fell from 6-8 to 5 8 per 1000total births if lesions in stillbirths, which wereunrecognisable without necropsy, were excluded.6 Athird reason for contradictions is the use of date ofbirth for studying seasonal prevalence,3 7 whenvariable lengths ofgestation may cause inaccuracies. Ithas been shown that significantly more babies withcardiac malformations than controls have shortgestations.8 Fourthly, in conditions withmultifactorial aetiology, one factor which could beseasonal may assume particular importance overanother in a given area.

330

Protected by copyright.

on January 2, 2022 by guest.http://jech.bm

j.com/

J Epidem

iol Com

munity H

ealth: first published as 10.1136/jech.43.4.330 on 1 Decem

ber 1989. Dow

nloaded from

http://jech.bmj.com/

Seasonal prevalence of major congenital malformations in the Fylde of Lancashire 1957-1981

In this investigation we attempted to get ascomplete an ascertainment of major malformationsin a population as is possible in practice and toeliminate problems caused by varying lengths ofgestation.

Methods

Major malformations were studied among babiesborn to mothers who lived in the most populous partof the Fylde of Lancashire. This is a well demarcatedarea approximately 24 km by 13 km, bounded on threesides by water barriers. The fourth, landward, borderis in sparsely populated country (fig. 1). Thepopulation is concentrated around the coast and hasincreased from about 250 000 to over 300 000 duringthe period of study. Blackpool is situated centrally onthe west coast. In the north the main urban areas areFleetwood, Thornton-Cleveleys and Poulton-le-Fylde, and in the south, Lytham St Annes andKirkham. The remainder ofthe district consists ofrichfarming land.

Fig I The Fylde of Lancashire

The district is 50 miles from the nearest Universityreferral centre and medical services are based on asingle district general hospital (Victoria), so that onedoctor was able personally to collect datasystematically on a day to day basis. The numbers oflive births, stillbirths and neonatal deaths were

notified to us by the medical officers of health andafter 1974 by the community physician. There were88 449 births in the 25 years.

Babies studied were those born in the Fylde toresidents there between 1 January 1957 and 31December 1981. Cases of malformations in babiesborn during this period were collected until 31December 1986, so that there was a minimumfollow up of 5 years. Major malformations wereascertained from: (1) necropsies, (2) the dailypractice of one paediatrician, and (3) in later years,the local obstetricians' records of pregnancyterminations.A very high rate of necropsies, all by consultant

pathologists, of stillbirths and all children dying up to15 years of age was maintained throughout the 25years. After 1957 arrangements for necropsies ofbabies stillborn at home were facilitated by thecontinuation of procedures introduced for localparticipation in the British Perinatal MortalitySurvey. Ninety three per cent of all perinatal deaths inthe area were examined, wherever they were born. Thenumber of necropsies was checked against the notifieddeaths and information obtained from midwives anddoctors about any visible malformation in the verysmall number of cases not examined by thepathologists. Necropsy reports of all Coroners' caseswere received.

Terminations of pregnancy because of antenataldiagnosis of a serious malformation were carried outfrom 1979. There were only four such cases, all neuraltube defects.For 17 years the great majority of cases of major

malformations were seen by one paediatrician (JPB) inhis practice, which included full charge of neonatesfrom the outset. He attended necropsies wheneverpossible and always reviewed cases with thepathologists. Members of the medical and nursingprofessions were aware of the study and wereencouraged to inform him of children with lesionswhich might have escaped his attention. He visited thegeneral practitioner maternity units at least weekly. Inaddition outpatient clinics were held not only atVictoria Hospital but also in the north (Fleetwood)and south (Lytham) of the district enabling goodliaison with general practitioners. Further, regularconsultative clinics were held with the local authoritydoctor responsible for the infant welfare and schoolmedical services in Blackpool.

Victoria Hospital has a department of cardiologyand a children's clinic was held by the cardiologist andpaediatrician together. Complete recording of cardiacmalformations presenting after the neonatal periodwas thus facilitated and diagnoses were constantlyupdated where necessary.

331



j.com/

J Epidem

iol Com

munity H


ber 1989. Dow

nloaded from


332In the last 8 years ofthe study a second paediatrician

cooperated fully in reporting cases seen in his practiceonly. These were few in number as neonatal work wascompletely shared and JPB continued to organise thecombined cardiac clinic.

Babies with multiple malformations were countedunder each major lesion present. The followinglesions were not recorded: inguinal and umbilicalhernias, hypospadias, undescended testes, hydrocele,talipes, polydactyly, syndactyly, accessory auricle,naevi.For some analyses cases of neural tube defect were

divided, as in a recent American study,' into"multiples" with one or more additional majorlesions, and "singles" with no other lesions or onlythose considered secondary to the neural tube defect.Secondary lesions included other central nervoussystem defects, talipes, dislocation of the hip,contractures, spine and rib defects and lung andadrenal hypoplasia.

Classification of cardiac lesions was based on thescheme of the International Society of Cardiology.'0In addition cases of uncomplicated ventricular septaldefect were divided into clinical and proven. Clinicalventricular septal defect was diagnosed from thepresence of a typical murmur with normal chestx ray and electrocardiogram. Fewer cases fell intothis category after the introduction ofechocardiography in the early seventies, whichenabled many small defects to be visualised withoutdisturbing the infant.

Uncomplicated persistent ductus arteriosus wasdistinguished from delayed closure of the ductus inpremature and small for dates babies. " Occasionally apatent ductus in one of these two groups of low birthweight babies required surgical treatment before thedistinction could be made clinically. Therefore ourcategory of persistent ductus was restricted to babieswith a gestation of 37 weeks or more and a birth weightover 2500 g.The antenatal notes of mothers giving birth to

babies with malformations were studied and, inparticular, the month of the last menstrual periodwas noted. The month was almost always recordedeven when sometimes the day of onset was notrecollected.Data on the month of last menstrual period and

month ofbirth was also collected for all hospital birthsfor a selection of years over the period of the study.For two early years of the study, data were alsocollected on month of birth for all births occurring inthe Fylde. These data sets provided information onany potential seasonal variability in the Fyldepopulation of births. Hospital births represented 33%of all births in 1959, increasing to 93% of all births in1981.

J P Bound, P W Harvey, and B J Francis

STATISTICAL METHODSEarly methods for detecting seasonality inepidemiological data have relied almost exclusively ontwo methods: the simple X2 test for heterogeneity and atest devised by Edwards.'2 Both methods have beencriticised; the former for its failure to utiliseinformation in neighbouring months'2 and the latterfor the poor performance of the test with smallsamples.3 Edwards' method has been extensivelystudied, and many modifications to the test have beensuggested, both to extend its generalityl4 and toimprove its small sample properties. 15 16 Many of thelatter modifications have assumed a multinomialdistribution, modelling the probability of monthlyoccurrence by a linear function. Recent work'7 hassuggested a linear model for the rate of the underlyingprocess; we extend this idea by assuming an underlyingPoisson distribution and a log-linear model for themonthly rates. These assumptions are natural for thedata, and provide (through the theory of generalisedlinear models)'8 a well established inferentialframework for the examination of hypotheses. Wealso explored a variety of seasonality models, allowingfor cyclical effects other than six months.The observed malformations were first classified

either by month of last menstrual period or bymonth of birth. We assume for the time being that thetotal number of births in the population shows noevidence of seasonality. If no seasonality is presentin the malformation under consideration, then thetrue underlying rate of occurrence per month isconstant, and the observed counts will have aPoisson distribution. If seasonality is present,then the underlying monthly rates will vary, andcan be modelled. The standard method of analysingPoisson counts is through the log-linear model.Let Yi, i = 1....,12 be the monthly counts fora malformation. Then the model of no seasonalityis

log(pi) = awith E(Yi) = pi

and Y, - Poisson (pi)The pi is the expected number of counts in month i.

We consider three models of seasonality:

(1) The two rate modellog(pi) = al i=j, j+ l,....,j+k- I

=a2 i=j+k,....,12,1....,j-IThis model assumes two underlying monthly rates ofoccurrence, one high and one low, and each for aconsecutive period of calendar months. The firstperiod starts in month j and is of length k months; thesecond period starts in month j + k and lasts for 12-kmonths. The parameters al, a2, j and k are estimatedfrom the data.



j.com/

J Epidem

iol Com

munity H


ber 1989. Dow

nloaded from



(2) The Sin modellog(p;) = Xo + XI sin (xi+m)

where xi is the month number i multiplied by 300, X isthe start of the cycle in degrees, and Xo and XI areunknown parameters to be estimated. Here, theunderlying monthly rate of occurrence variessmoothly, showing a six month period of increase,followed by a six month period of decrease. If X1 iszero, then the model reverts to that of no seasonality.

(3) The Sin squared modellog(ji) = Xo + XI sin(xi + m) + X2 sin2(X, + m)

where xi and X are defined as before. For this model,the underlying monthly rate of occurrence still variessmoothly, but follows a more complex pattern ofincreases and decreases. A version of this model withthe counts transformed to approximate normality wassuggested by Sandahl,'9 who recommended it for thedetection of asymmetrical periodic changes.

Each method has its advantages and disadvantages.The two rate model has the advantage that the modelis easily interpreted, but it may not be a realisticrepresentation of the underlying process. The sinmodel has a smooth variation in rate over the 12months, but the assumption of six months of increasefollowed by six months of decrease may not beappropriate; important cyclical effects other than sixmonths may well be missed. The sin squared modelavoids this problem by adding a squared sin term tothe model allowing for asymmetrical changes over theyear. The estimated function produced by thismethod, however, is often difficult to interpret, andcan have multiple maxima over the 12 monthperiod.For a given set of data, it is possible that certain

models may show significant results and others non-significant results. This indicates that certain of themodels are more effective at detecting certain types ofperiodicity compared with others.The parameters for all the above models were

estimated using the Poisson regression facility of theGLIM1 statistical package.20 Maximum likelihoodestimates of the parameters were found by acombination of GLIM model fitting and grid searchtechniques. For the two rate model, this procedureinvolved fitting a separate model to estimate ac ,aC2 andthe log likelihood for every combination of(j,k) over agrid of all possible values of j and k. The values(imax, kmax) ofj and k which gave the maximum valueof the log likelihood over the grid were taken as themaximum likelihood estimates of j and k. For the sinand sin squared models, a one dimensional grid with mtaking the values 00, 30°,....,3300 was chosen and asimilar procedure was used.

The deviance (minus twice the log likelihood) of theestimated model of seasonality was then subtractedfrom the deviance produced by the model of noseasonality to give a likelihood ratio test statistic. Forthe sin (and the sin squared) models, we can fix X andX1 (and X2) at zero to obtain the model of noseasonality; the test statistic is thereforeapproximately distributed as %2 on two (and three)degrees of freedom if the model of no seasonalityholds. For the two rate model, we can fix j and k at 0 toobtain the model ofno seasonality, but the parametersj and k then lie on the boundary of the parameterspace, and regularity conditions for the likelihoodratio test are violated. For this model, the test statisticunder the hypothesis of no seasonality is therefore notx2 distributed, and simulations of the distribution arenecessary. The results of a series ofsimulations of 1000trials based on constant monthly Poisson rates of 30cases every 25 years increasing in steps of 30 up to amaximum of 240 cases every 25 years, produced noevidence of the distribution of the test statistic beingrelated to the underlying Poisson mean. The meanvalue produced for the 95th percentile was 10-34; thatfor the 99th percentile was 13-89 and that for the99-5th percentile was 15-36. These values were used insubsequent testing.

All three models were applied both with andwithout allowing for seasonal variation in the totalnumbers of births in the study population. Theprocedure when allowing for such variations involveddetermining PBi (i = 1....,12), a suitable monthlyclassification of all births in the population underconsideration. In this analysis, we used hospital birthsfrom 1979 to 1981 to represent the overall pattern ofbirths in the Fylde. These data, which represent 92%of all births, were classified both by month of lastmenstrual period and by month of birth to providesuitable values for the PBi (depending on whether theanalysis was on malformation classified by month oflast menstrual period or birth month). For each of themodels, an extra term log(PBJ) is included, with a fixedregression parameter of unity. For example, the sinmodel becomes

log(gi) = log(PB;) + X. + XI sin (xi+ t) Yi -

Poisson (tLi)This model is compared to a modified "noseasonality" model

log(pi) = log(PBi) + X..

Results

The total prevalence of major malformations over 25years is shown in table 1. The prevalence of casesdiagnosed up to the age of 5 years, the minimumfollow up for children born at the end of the survey, is

333



j.com/

J Epidem

iol Com

munity H


ber 1989. Dow

nloaded from


Table 1 Prevalence ofsome common major malformations in the Fylde 1957-81

Ascertained up toage S years

TotalPrevalence prevalenceper 1000 Ascertained per 1000

Malformation n births later (n) births

Anencephaly1957-68 129 2-74 - 2741969-81 40 0 97 - 097

Spina bifida cystica and cranium bifidum1957-68 131 2 78 - 2781969-81 69 1 67 - 167

Uncomplicated congenital hydrocephalus 34 0 38 - 0-38Down's syndrome 124 1 40 - 1-40Congenital microcephaly 16 0-18 - 0-18

Ventricular septal defect uncomplicated 165 1-87 4 1-91Ostium secundum atrial septal defect 59 0-67 10 0-78Ventricular septal defect with right ventricular outflow obstruction 50 0 57 0 0-57Endocardial cushion defect 42 047 0 0-47Persistent ductus arteriosus 30 0 34 9 044Coarctation of aorta 36 0-41 2 0-43Transposition of great arteries 31 0 35 0 0-35Aortic stenosis 22 0 25 5 0-31Pulmonary stenosis, isolated 20 0-23 1 0-24All congenital heart disease 579 6-55 44 7-04

Diaphragmatic hernia 36 0 41 - 0-41Exomphalos 39 0 44 - 0-44Oesophageal atresia 41 0 46 - 0-46Intestinal atresia 25 0 28 - 0-28Anorectal anomalies 57 064 - 0-64

Bilateral renal agenesis 24 0 27 0 0-27Unilateral renal agenesis 28 0 32 2 0 34Renal hypoplasia 23 0-26 1 0 27Polycystic kidneys 23 0-26 1 0 27Hydronephrosis (excluding cases secondary to spina bifida) 69 0 78 25 1-06Primary vesicoureteric reflux 62 0 70 34 1-09Duplex kidneys 23 026 21 050

Cleft lip with or without cleft palate, 1962-81 67 0-96 - 0-96Isolated cleft palate, 1962-81 48 068 - 068Congenital dislocation of hip 99 1-12 - 1-12Reduction deformities of limbs 55 0-62 - 0-62

also recorded. The difference between the two rates is afeature of cardiac and particularly of urinary systemmalformations. Table 2 gives the observed number ofcases for each malformation classified by month oflastmenstrual period. Table 3 shows the same casesclassified by month of birth for malformations whichshowed a significant variation in month of lastmenstrual period. Congenital dislocation of the hipwas also included in table 3 because some possibleaetiological factors operate in late pregnancy. Forcertain malformations, the data on last menstrualperiod have been further subdivided into two periodsof 12 and 13 years because a sudden significantdrop in the prevalence of anencephaly in the Fyldeamong conceptions after 1967 had been shownpreviously.2'

Table 4 shows the pattern of hospital births for aselection of years over the study period, classified bymonth of last menstrual period and month of birth

respectively. Table 4 also shows the pattern of allbirths for two years early in the study period, classifiedby month of birth. We used these data to determinewhether any seasonality in all births in the Fylde waspresent. To this end, we analysed the hospital birthsusing the three seasonality tests described earlier. Forthe hospital births classified by last menstrual period(fig 2), only one year (1976) showed a significantseasonal variation on the sin test, with a peak month ofSeptember. No other years showed a significantseasonal effect. For hospital births classified by monthof birth, 1979 was the only year to show a significantseasonality effect, with a peak month of October. Noconsistent seasonal pattern is therefore observable inthe births of Fylde residents whether last menstrualperiod or birth month is analysed. Finally, weexamined the representativeness of hospital birthscompared to all births in the early years of the study.Using the three seasonality tests, data on hospital

334 J P Bound, P W Harvey, and B J Francis



j.com/

J Epidem

iol Com

munity H


ber 1989. Dow

nloaded from



1 >0>C b

00Mc 'IO NNco00x4 Om cCt>W0; r~O~ cirCLi-0% " C

C C

OCrf(N N0000'IT

000otNN0O_.u en 09X d~C£as o s

c;f r-o br O o-0 -Xt -O -

en 0(0 d XI..T

ONO(N _ (N(N r

0o \0NNr

la t %n %eC t

0o o - 0x

ON (N= = NI

O0 Wv -O XO en

- X

e0 Oo e00o(ol-

-fi _ _ 0, _

ecd 00-

m 00(0-

CA C IC10 Ci4 ~o

L00 en

(N-_N en

'o N7-tm-u- 00 (N ^

-) CD

\000 00(N (N

0CX N

x ON N °

d -0t_

--N -

I-, 0-

0 00

r..

0 0

N n

;z as

,It rW

t- N

X

Ch lt t-

o 00

tnC

00N00 Nm r

o oo q =oo

-

bo

- c- c 'I r-

~ 00 >C

>0

00 WIo) c7s

00000 WIt

11 CD cten (

__+F**'~Q= lq s-X- n

'I o7cz-T cra

335

t3 0

0b

:5

COcd

0

0o

00 C(4 W)w

en CUCO

z

Z

r OP Q~oo- r

00en I 00

10 en er-C4-l

0

mC>-

N- N00

_~r 00_ (N

-^rN \

_ _ ;e ;O

NN 0 Ycd

-> (

0_ _* -

- 'R 'N-ONCt ti ri

e0 00 (N eON00 00 ON0 00

0d 00< I

4

- X0 W(N-

WI) el) C7. (N4 en 0

'IO 0 (N1 (NO1C

W) (N 00 (n

en eM 00 W -

a, t cn __

.00

0.

0

CO

CO0

0

0

0

0

0

000CO

0CO

0

CO

.0

0

0CO

0

0

0

00

0

0

0CO

-0

(N

CU

W ON 0000

C> (N=C4e

- ~.

06v0

O 0 O (N en

en en M en M

"t "t C4en 'IT

CU4

"t M ltdb-o

Cq q

enNtt

C) en4

c

(0.

co c o

0



j.com/

J Epidem

iol Com

munity H


ber 1989. Dow

nloaded from



0.

000N NW)

00_14It %n 't r-00

*at%Or'% -al e

en>'ITen - C'

-mm_N"Ituo

o-46 -oo0

e) tNO ene

t- 0" ooo

eit ar N 00^)t-

0 st0- NcryN

O db-Nd--t--09

.0 4

O O o CCZZZ"=<II~II,(

C_ Ct00 el

-NW)'tal

Cl4 Cl 00 Cl4

.O NO

00 N- C

tjo-

-_ WI 00 N N

R 00-

N 00 en

t% Cm 00000

N NE o0en

00 110 C4 'IO N

0)~ ~ 0No r.

U00D E> < .0L*5 j; 3-

* a <>

0

00

9 D.0

C >

ow ,

o0-

~0 0)

00

V 00'-

V >

0.(0-

'-0

oCd

00

1-.0

aL

0.- cO

0 W

0 C

(00 w

0o O

id 0).

0e <0Q *0.

(0.0co0

II O

-0S-0

0

te

-Z0

.@

0

S

-0

0s0

0

0

0

03S0Q0iS

C.)CA)

0L)-0

Cad

k;0j0

Q

F-a

0.j

-0La,

0'I'I

1--

W0

en) C4 00en19 O C? I?

6,t o_br,

-_

aoooX\000mt 1NIat% 0000

I%t - It %00% -

N0 00 4 WINr)4=C 0N1C14 - -

Ioot ro> >£

CD0ooN 00oen

e-N Ne%00N-'_

000 at aen -(*e)

a-- -')'o -

C> r- "o en en^.m o

it

CS-O0%0 .° Nr0e^C'x0

WI 0 " Cfo 0

-%00%0 4~~~0t' n 'I

IC ONCX est o C4 x- -- 4 - -

_ _00 _r en "iten

- 0%00--C

---<, - -~~~5

*0

c0 5̂wo * 0C

0iB ig0<<> ;

336

k

02

0i

00j

--

Q

-a

,0Z

r0)

-0Is

0

0:

-0z0zQ

Q-

-0~S

00 W)

C6

(0

-40 00

00)

C> 00 (7

__ t3

too

0) 0)

0C

cnoC

Cl

-0 N

04*.0. 0

0o

_0,. o0

~0 0

>> U

.0.000

00

00

4.00

0.

0.

w

9

0

64K(DvCkw

C-o



j.com/

J Epidem

iol Com

munity H


ber 1989. Dow

nloaded from



Table 4 Selectedyears ofhospital births to Fylde residents by month oflast menstrualperiod (LMP) andmonth ofbirth, and alIbirths 1959 and 1960, by month of birth

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Month ofLMP (hospital births)1959 97 100 91 105 95 107 97 101 109 97 103 1051960 116 112 111 91 110 112 113 120 102 115 83 1231965 147 125 160 147 137 154 152 134 138 163 138 1511969 116 149 139 142 137 138 142 160 142 142 132 1611976 175 160 166 176 184 192 187 195 203 191 208 1961979 291 225 228 198 214 225 237 249 228 246 196 2501980 250 221 254 257 251 258 249 244 245 249 229 2681981 253 237 242 230 257 257 249 238 237 235 233 227

Month of birth (hospital births)1959 101 89 113 95 110 94 119 91 113 94 93 951960 80 98 118 125 116 110 109 88 119 120 117 1081965 154 128 157 147 127 148 163 151 138 145 138 1501969 140 131 150 129 165 138 129 151 160 133 143 1311976 171 179 197 178 206 195 204 203 189 177 157 1771979 208 196 241 216 255 234 235 212 239 292 246 2131980 274 268 257 249 245 239 261 230 242 281 211 2181981 246 216 255 259 261 224 238 218 238 251 243 246

Month of birth (all births)1959 308 292 297 299 347 298 344 290 298 285 260 2761960 266 309 322 314 337 306 317 271 329 356 330 314

births in 1959 and 1960 were now examined using the ANENCEPHALYmonthly variation in all births as a correction factor. The only anencephaly data to show a significant effectNo significant effects were found. on all three tests were the anencephaly "multiples"Although no consistent pattern of seasonal based on month of last menstrual period. The two rate

variation was found, we considered it important to and sin squared models show that the high period wasanalyse the data both correcting for and not correcting November to May.for monthly last menstrual period or birth variation,and so the number of hospital births in the Fylde over SPINA BIFIDA AND CRANIUM BIFIDUMthe period 1979 to 1981 was used as a correction For data based on month of last menstrual period,factor. These years were chosen to represent all births, significant seasonal effects were found for spina bifidasince hospital births in this period constituted 92% of and spina bifida plus cranium bifidum, both in theall births in the area. period May 1956-April 1968 and in the whole of the

Tables 2 and 3 also give the values for the test survey period. Spina bifida and cranium bifidumstatistics for the three methods described earlier. The "singles" also showed a significant seasonaltests were carried out both on uncorrected data and on prevalence. All conditions and combinations ofdata corrected for monthly variations in all conditions showed a high rate during the period ofconceptions or births. In general, corrected and March to May on both the two rate and the sinuncorrected values of the test statistics were similar squared test. The lack of significance of the sin test forand we therefore quote only the uncorrected values. some of the above conditions shows the poorFor each malformation showing a statistically performance of this test in detecting short seasonalsignificant seasonal effect, the pattern of the effect effects. Figure 3 shows the observed data and the threefound is also given. For the two rate model we give fitted models (uncorrected) for spina bifida andmonths of the high rate period; for the sin model the cranium bifidum "singles".month corresponding to the peak of the sin curve, and For data based on month of birth, only the sinfor the sin squared test the months in which the fitted squared test for spina bifida shows a significant result,curve lies above the line of no seasonality. As an with a peak between the months of December toindication to other researchers, we also give the February and a second smaller peak in June. The factmonths (enclosed in parentheses) of high rate period that the fitted sin squared function has two high peaksfor the two rate model even where there is no makes this difficult to interpret, and the significantsignificant effect. values should be treated with caution.

337



j.com/

J Epidem

iol Com

munity H


ber 1989. Dow

nloaded from



A .". /I

I \\

Jan Feb Mar Apr May Jun Jul Aug Sep

Month of last menstrual period

Fig 2 Monthly ratio of actual to expected conceptions of babies born in hospital in selected years

Oct Nov Dec

ALL NEURAL TUBE DEFECTSBased on month of last menstrual period, the data forall neural tube defects, both for May 1956-April 1968and for the complete study period, show a seasonaleffect on all three tests. Neural tube defect "singles"showed a similar seasonal effect. The two rate and sinsquared models give a high period of December-Mayfor the May 1956-April 1968 period, and February-May for all neural tube defects and neural tube defect"singles". The December to May period was alsosignificant for the latter two categories. There were no

significant effects when the data were based on monthof birth.

CARDIOVASCULAR SYSTEM

Both ostium secundum atrial septal defect andendocardial cushion defect showed stronglysignificant seasonal effects on data based both on lastmenstrual period and on date of birth, with a high lastmenstrual period rate in May to July for atrial septaldefect, and in April or May to November forendocardial cushion defect. Significant summer

seasonal effects, though not on all three tests, were alsofound for proven ventricular septal defects, allventricular septal defects, and persistent ductusarteriosus. There is also some evidence in our data tosuggest that all abnormalities of bulboventricular

1-31

1.21

1959 .....................1976

-- 1960 ~ ~ --- 1979

------- 1965 --- 1 980

- --1 969 - 800.

CL4)

8 1-1-a,UQxw

06-i

, 1.0-U

0

0.Ucr:

0.9-

0.8-

n.7

338



j.com/

J Epidem

iol Com

munity H


ber 1989. Dow

nloaded from



For vesicoureteric reflux, the sin and sin squaredtests were both significant, with a seven monthperiod from April to October being identified bythe sin squared model and a peak ofAugust identifiedby the sin model. The two rate model fits the data

- -----------------u----------- less adquately than a smooth function for thiscondition.

.

z

z

25- Sin model

20-

15- ~

10- *

5

25- Sin squared model

20-

15- /

10 F--------------------

m* m10-

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov DecMonth of last menstrual period

Fig 3 The threefittedmodelsfor nwnbers ofspina bifida andcranium bifidum "singles" (no other major malformations) bymonth of last menstrual period.

growth have a higher rate in winter when classified bymonth of last menstrual period, with a peak inDecember on the sin test just failing to reach standardlevels of significance (p = 0.052). When month ofbirth was considered, no significant seasonal effectwas shown for proven ventricular septal defect, andseasonal effects, although present for all ventricularseptal defects and persistent ductus, were significanton only one ofthe three tests, making the evidence lessstrong.

URINARY SYSTEMTwo conditions produced significant results (thoughneither on all three tests) when the data was classifiedby month of last menstrual period.For bilateral renal agenesis, the two rate model

produced a high period from March to October.Examination ofthe data for this condition reveals thatthe low winter period of four months produced onlyone case, compared with 28 for the 8 month summer

period. The two rate model is better able to fit this typeof data than the smooth sin and sin squared models,explaining the lack of significance of the latter twotests.

Discussion

ASCERTAINMENTThe greatest problem in epidemiological studies ofcongenital malformations is adequate ascertainmentof cases. The success of a malformations registry hasbeen stated to depend in part on the constant activityof one enthusiastic individual who runs it.5Ascertainment is facilitated if this person is a singlehanded paediatrician working in a well circumscribedarea served by one district general hospital and 50miles from the nearest university referral centre. Thenthe vast majority of major malformations will be seen

during his routine daily clinical practice.Collaboration with a pathologist interested in

paediatric and especially perinatal pathology isessential, so that lesions are not missed, particularly instillbirths, and for accurate diagnosis. A very highnecropsy rate is mandatory. Stillbirth and deathcertificates, with their well known inaccuracies indiagnoses,22 are used only to identify cases notexamined by the pathologists, so that any availableinformation about these babies can be obtained fromtheir medical attendants or midwives.

This was the situation in the Fylde for the first 17years of this study and led us to conclude thatascertainment was as complete as it is possible to get inpractice. This conclusion was not altered in the lasteight years when a second paediatrician cooperatedfully with the study. Support for our claim wasobtained where it was possible to compare our

prevalence rates with others' findings, as wewere able to do for example for congenital heartdisease.6

SEASONAL VARIATIONSA seasonal prevalence for a malformation seems toindicate an environmental factor in aetiology.Certainly for anencephaly a study in Birminghamsuggested that the seasonal prevalence couldnot be explained by fluctuation either in theconception rate of mothers or in abortion of affectedembryos.Time of conception is crucial to the search for

environmental aetiological factors. For this reasonmonth of maternal last menstrual period was used inthis study. With date of birth, errors may occur

because of varying gestations.

25, Two rate model

U>U

20

o 15-z

10.

*U

339

r,



j.com/

J Epidem

iol Com

munity H


ber 1989. Dow

nloaded from


340

We have shown that a seasonal variation for neuraltube defects, which was significant when month of lastmenstrual period was used, was not significant whendate of birth for the same cases was substituted. Casesof spina bifida alone showed a seasonal prevalencewith month of last menstrual period and also withmonth of birth in one of the three tests used. Themasking effect from using date of birth in the wholegroup appeared to be due mainly to the anencephalycomponent. This resulted from a greater variation ingestation in anencephaly, with 74% under 37 weekscompared with 13% for spina bifida.The study of anencephaly and spina bifida by the

Office of Population Censuses and Surveys24 tried toovercome this difficulty by using an estimated lastmenstrual period based on the percentage distributionofpregnancy duration for stillborn cases. However, inour series slightly less than a quarter of cases of spinabifida were stillborn and 71% of these had gestationsof 37 weeks or over compared with 91% in live births.Therefore use of the distribution of gestations instillbirths for estimating month of last menstrualperiod for all cases may have distorted the OPCSresults.Concealment of a significant seasonal prevalence

by not using date of last menstrual period couldbe seen with other system lesions, for example incases of proven ventricular septal defect andbilateral renal agenesis. However for most cardiaclesions a significant seasonal variation shown bymonth of last menstrual period was also revealed bydate of birth.

In this paper we have confined ourselves to showingwhich major malformations have evidence of seasonalvariations and have not tried to identify possibleenvironmental factors.

CENTRAL NERVOUS SYSTEM LESIONSFrom May 1956 to April 1968, when the prevalence ofneural tube defects was high in the Fylde, conceptionsof anencephaly were more common although notsignificantly so, in the six months December to May.This was a wider spread than the peak months ofFebruary to April reported for England and Wales24using estimated last menstrual period. In the sameperiod in the Fylde conceptions of spina bifida andcranium bifidum were significantly more common inMarch to May. This was a month earlier than, but ofthe same duration as, the peak national figures ofApril to June. Thus our findings for this period wereessentially similar to the national figures for 1964 to1972.After 1967 the fall in prevalence of neural tube

defects in the Fylde was associated with increasingsummer hardness of the drinking water.21 This falloccurred particularly in the months December to


August so that the significant seasonal variationsdisappeared. The appearance ofa low number ofcasesin June to August was not statistically significant,perhaps due to the fall in total numbers.

In an American study of neural tube defects9 only"6singles" showed the well known epidemiologicalcharacteristics. No seasonal variations were found ineither "singles" or "multiples", but only date of birthwas analysed. Our study, using maternal lastmenstrual period, showed a significant seasonalvariation in all neural tube defects considered togetherfor "singles", but not for "multiples". Spina bifida andcranium bifidum gave a similar picture, but withanencephaly it was "multiples" which showed thesignificant variation. A seasonal prevalence is one ofthe epidemiological characteristics of neural tubedefects, but our findings show that it is not invariablyconfined to "singles".A seasonal prevalence for Down's syndrome has

been shown by some investigations25 26 but notothers.27 Although we found fewer cases withmaternal last menstrual period in November toJanuary the figures were not significant, either whenthe mother was aged 35 years and over, or withyounger mothers. We offer no support for those whohave suggested that seasonality plays a part in theaetiology of Down's syndrome.

Uncomplicated congenital hydrocephalus was farless common than hydrocephalus with spina bifidaand our low numbers showed no significant seasonalvariation.

CARDIOVASCULAR SYSTEM LESIONSThe prevalence of congenital heart disease has beenreported to be highest in winter births7 or summerconceptions.6 Now we have shown that the excess ofconceptions in summer was due mainly to three of thecommonest types of heart lesion, uncomplicatedproven ventricular septal defect, atrial septal defectand endocardial cushion defect. It may be relevantthat these lesions share a common basis of incompleteseptation of the heart at some stage of development.These findings for septal defects receive some supportfrom a population survey in South Wales.28

Clinical ventricular septal defect also showed atendency to more conceptions in summer, althoughnot significantly so. The less strong seasonal variationcould suggest that a different aetiological factor wasmore important for smaller defects. Alternatively andmore likely, some clinical diagnoses could be wrong,although care was taken in the differential diagnosis,particularly from non-specific, transient systolicmurmurs.An additional contribution to the excess of summer

conceptions was provided by persistent ductusarteriosus, another common lesion. This condition



j.com/

J Epidem

iol Com

munity H


ber 1989. Dow

nloaded from


Seasonal prevalence of major congenital malformations in the Fylde of Lancashire 1957-1981showed a significant increase of conceptions in latesummer, boys and girls contributing almost equally tothe total. There was a corresponding peak in birthsfrom May to July, which was earlier than a previouslyreported peak in affected girls from August toNovember.3 Four infants in our series had persistentductus arteriosus due to maternal rubella, three beingconceived in late winter or spring. The late summerpeak may result from infections of a different virus.

In contrast to the findings for septal defects werethose for another group of lesions which have beenclassified as showing abnormalities of bulbo-ventricular growth.29 These are ventricular septaldefect with right ventricular outflow obstruction,including Fallot's tetralogy, double outlet rightventricle and transposition of the great arteries. Thisgroup, the commonest cause of cyanotic congenitalheart disease, showed its lowest prevalence from Mayto August. The figures did not reach statisticalsignificance but contrasted sharply with the findingsfor septal defects. An excess ofwinter conceptions wasshown by both transposition of the great arteries andventricular septal defect with right ventricular outflowobstruction. The latter finding is supported by thereport from Liverpool8 of a significant excess ofinfants with Fallot's tetralogy conceived from Januaryto June.A seasonal prevalence for coarctation of the aorta

has been reported, with an excess of boys born inMarch and April,3 while another hospital seriesshowed two peaks.30 Our menstrual period datashowed no seasonal variations, even for boysalone.An excess ofboys with pulmonary valvular stenosis

born from July to September has also been noted.3 TheFylde had a low prevalence of isolated pulmonarystenosis6 and our small numbers did not show asignificant seasonal variation.

ALIMENTARY SYSTEM LESIONSA summer excess of births for abnormalities of thelower gut and a winter excess for oesophageal atresia,but not tracheo-oesophageal fistula, has beenreported.3' We did not show a significant seasonalprevalence in month of last menstrual period foranorectal anomalies (rectal agenesis, ectopic anus andcovered anus), oesophageal atresia including tracheo-oesophageal fistula or any other major alimentarymalformation.

URINARY SYSTEM LESIONSIn published reports a seasonal variation in prevalencehas not been a noted feature of urinary tract lesions.

In the Fylde bilateral renal agenesis wassignificantly decreased in babies conceived inNovember to February. This finding was strengthened

by the fact that ascertainment should have beencomplete. Cases are stillborn or fatal within two daysof birth and very unlikely to have been missed with a93% necropsy rate for perinatal deaths.

Primary vesicoureteric reflux was revealed bycystourethrography but was also diagnosed inchildren with renal scarring in whom reflux could nolonger be demonstrated. In the Fylde, cases ofprimaryvesicoureteric reflux showed a significant excess ofconceptions from April to October. This finding mustbe treated with reserve as the degree of ascertainmentis unknown. Cases will be recognised only if a urinarytract infection occurs and is managed correctly,32 or ifa baby is investigated because of a family history ofreflux.No other common urinary tract malformation

showed a significant seasonal variation.

SKELETAL SYSTEM LESIONSWe did not keep records of some common limbmalformations such as talipes because we could notensure adequate ascertainment.

Facial clefts were analysed from 1961, after whichthe vast majority of cases were seen before referral tosurgeons. Our recorded prevalence was similar to thatreported from nearby Liverpool over a comparableperiod.33 In Birmingham34 a significant seasonality forcleft lip was shown using month of birth, but was notconfirmed at Oxford.35 In Sweden36 a seasonalvariation was found for cleft palate and cleft lip withor without cleft palate using calculated last menstrualperiod, but not with birth date. In Liverpool33 therewas a trend towards conception of babies with cleftpalate in the second half of the year, but this wassignificant for girls only.We found no significant seasonal prevalence in

month of last menstrual period for cleft lipwith or without cleft palate, or for isolated cleftpalate.

In the 1950s an excess of births of babies withcongenital dislocation of the hip in winter wasreported.3' 34 However, we found no suggestion of aseasonal variation whether using date of birth or oflast menstrual period.

We thank Drs D S Harry, S Murray and K S Vasudev,Consultant Pathologists, for their interest inpaediatric necropsies; Drs W F W E Logan and J SWright, Consultant Cardiologists, for their help,especially in classifying cardiac cases; and Dr C JWoods, Consultant Paediatrician, for his co-operation after 1973. We are grateful to Profs AMercer and M Aitkin of the University of Lancasterfor advice and encouragement, and also to Prof J ADavis of the University of Cambridge for hissupport.

341



j.com/

J Epidem

iol Com

munity H


ber 1989. Dow

nloaded from


342

Address for correspondence and reprints: Dr J PBound, Department of Paediatrics, Victoria Hospital,Blackpool FY3 8NR, UK.

References

1 McKeown T, Record RG. Seasonal incidence ofcongenital malformations of the central nervous system.Lancet 1951; i: 192-6.

2Jolly H, Levene MI. Diseases of children. 5th ed. Oxford:Blackwell Scientific Publications, 1985: 133.Campbell M. Causes of malformations of the heart. BrMed J 1965; ii: 895-904.

4Laurence KM, Carter CO, David PA. Major centralnervous system malformations in South Wales. Br J PrevSoc Med 1967; 21: 146-60.

5 Hay JD. Population and clinic studies of congenital heartdisease in Liverpool. Br Med J 1966; ii: 661-7.

6 Bound JP, Logan WFWE. Incidence of congenital heartdisease in Blackpool 1957-1971. Br Heart J 1977; 39:445-50.

7 Feldt RH, Avasthey P, Yoshimasu F, Kurland LT, TitusJL. Incidence of congenital heart disease in children bornto residents of Olmsted County, Minnesota, 1950-1969.Mayo Clin Proc 1971; 46: 794-9.

8 Kenna AP, Smithells RW, Fielding DW. Congenital heartdisease in Liverpool: 1960-69. Q J Med 1975; 43: 17-44.

9 Khoury MJ, Erickson JD, James LM. Etiologicheterogeneity of neural tube defects: clues fromepidemiology. Am J Epidemiol 1982; 115: 538-48.

International Society ofCardiology. Classification ofheartdisease in childhood. Groningen: VRB Offsetdrukkerij,1970.

1 Becker AE, Anderson RH. Pathology of congenital heartdisease. London: Butterworths, 1981: 353.Edwards JH. The recognition and estimation of cyclictrends. Ann Hum Genet 1961; 25: 83-7.

3 Hewitt D, Milner J, Csima A, Pakula A. On Edwards'criterion of seasonality and non-parametric alternative.Br J Prev Soc Med 1971; 25: 174-6.

4 Walter SD, Elwood JM. A test for seasonality of eventswith a variable population at risk. Br J Prev Soc Med1975; 29: 18-21.

5 St Leger AS. Comparison of two tests of seasonality inepidemiological data. Appl Statist 1976; 25: 280-6.

6 Roger JH. A significance test for cyclic trends in incidencedata. Biometrika 1977; 64: 152-5.

7 Nam J. Efficient method for identification of cyclic trendsin incidence. Commun Statist Theor Meth 1983; 12:1053-68.

18 McCullagh P, Nelder JA. Generalised linear models.London: John Wiley, 1983.

J P Bound, P W Harvey, and B J Francis19 Sandahl B. Seasonal Incidence of some congenital

malformations in the central nervous system in Sweden1965-1972. Acta Paediatr Scand 1977; 66: 65-72.

20 Baker R, Clarke MRB, Francis B, et al. The GLIM systemrelease 3.77 Users' Guide. 2nd ed. Oxford: NumericalAlgorithms Group, 1986.

21 Bound JP, Harvey PW, Brookes DM, Sayers BMcA. Theincidence ofanencephalus in the Fylde peninsula 1956-76and changes in water hardness. J Epidemiol CommunityHealth 1981; 35: 102-5.

22 Emery JL, Irvine KA. Death certification of children. BrMed J 1958; ii: 1510-2.

23 Leck I, Record RG. Seasonal incidence of anencephalus.Br J Prev Soc Med 1966; 20: 67-75.

24 Rogers SC, Weatherall JAC. Anencephalus, spina bifidaand congenital hydrocephalus. England and Wales 1964-1972. Office ofPopulation Censuses and Surveys. Studieson Medical and Population Subjects No. 32. London:HMSO, 1976.

25 Leck I. Incidence and epidemicity of Down's syndrome.Lancet 1966; ii: 457-60.

26 Harlap S. A time-series analysis ofthe incidence of Down'ssyndrome in West Jerusalem. Am J Epidemiol 1974; 99:210-7.

27 Mikkelsen M. Epidemiology of trisomy 21: population,peri- and antenatal data. Hum Genet 1981; 2: 211-26.

28 Roberts CJ, Lowe CR, Lloyd S. Cyclic variations in date oflast menstrual period of mothers of infants withcongenital malformations in South Wales, 1964-66. Br JPrev Soc Med 1972; 26: 212-8.

29 Anderson RH, Ashley GT. Growth and development ofthe cardiovascular system: anatomical development. In:Davis JA, Dobbing J, eds. Scientifc foundations ofpaediatrfcs. London: William Heineman, 1974: 165-98.

30 Miettinen OS, Reiner ML, Nadas AS. Seasonal incidenceof coarctation of the aorta. Br Heart J 1970; 32: 103-7.

31 Slater BCS, Watson GI, McDonald JC. Seasonal variationin congenital abnormalities. Br J Prev Soc Med 1964; 18:1-7.

32 White RHR. Management of urinary tract infection. ArchDis Child 1987; 62: 421-7.

3 Owens JR, Jones JW, Harris F. Epidemiology of facialclefting. Arch Dis Child 1985; 60: 521-4.

3 Edwards JH. Seasonal incidence of congenital disease inBirmingham. Ann Hum Genet 1961; 25: 89-93.

5 Fraser GR, Calnan JS. Cleft lip and palate: seasonalincidence, birth weight, birth rank, sex, site, associatedmalformations and parental age. Arch Dis Child 1961; 36:420-3.

36 Sandahl B. Seasonal incidence ofcleft lips and cleft palatesin Sweden, 1965-1974. ScandJ Plast Reconstr Surg 1977;11: 39-43.

Acceptedfor publication April 1989.



j.com/

J Epidem

iol Com

munity H


ber 1989. Dow

nloaded from


seasonal fylde of lancashire 1957-1981

Documents