Comparison of the Seasonal Pattern in the Clinical - Diabetes Care

E p i d e m i o I o g y / H e a 11 h S e r v i c e s / P s y c h o s o c i a I R e s e a r c hN A L A R T I C L E

Comparison of the Seasonal Pattern inthe Clinical Onset of IDDM in Finland andSardiniaMARJATTA KARVONEN, PHDVIRVAJANTTI, BSCSANDRO MUNT0N1, MDMARCO STABILINI, MD

LILIANA STABILINI, MDSERGIO MUNTONI, MDJAAKKO TUOM1LEHTO, MD, PHD

OBJECTIVE — To examine the seasonal pattern for the clinical onset of IDDM in Finlandand Sardinia, two areas where the incidence of IDDM is the highest in the world, and to deter-mine the effect of climate and temperature on the clinical onset of IDDM.

RESEARCH DESIGN AND METHODS— Analysis of seasonality for the diagnosis ofIDDM was based on 1,405 cases in Finland and 425 cases in Sardinia diagnosed at ^14 yearsof age from 1989 to 1992. The average annual incidence of IDDM was 36.4/100,000 in Finlandand 34.4/100,000 in Sardinia. Seasonal patterns were estimated presenting the data as shortFourier series up to three harmonics together with a possible linear trend. Likelihood ratio testsand Akaike's information criterion were used to determine the number of harmonics necessaryto model the seasonal pattern. Seasonal patterns in both countries were compared betweensexes and between the three 5-year age-groups, each controlling for the other's effect.

RESULTS — In both countries, a significant seasonal pattern during a calendar year wasfound for the sexes combined and for two age-groups (0-9 and 10-14 years). In Sardinia, twodistinct cycles were found in the younger age-group, with a decreased incidence during Maythrough August and an increased incidence during the autumn months. Two cycles were appar-ent in the older age-group, with the nadir occurring during June through September. In Fin-land, one cycle was found in the younger age-group, with a decreased incidence in June. In theolder age-group, there were two distinct cycles, with a decreased incidence in June and in theSeptember through December period.

CONCLUSIONS — Differences between Finland and Sardinia in the seasonal pattern forthe incidence of newly diagnosed IDDM cannot be explained by differences in climate, tem-perature, a longer warm period in Sardinia, or other climatic phenomena. The results do notprovide evidence in favor of a specific viral etiology of IDDM. It may be suggested that thereare triggering events at certain times, but they are likely to be unspecinc. Nevertheless, why theincidence of IDDM in these two populations is equally high despite differences in climate, envi-ronment, and genetic background remains an unsolved question.

Several epidemiological studies havedescribed seasonal patterns in newlydiagnosed cases of IDDM in children.

Most studies have reported a higher occur-rence of IDDM during the cold autumn andwinter months than during the warmerspring and summer months (1-16). Sea-

sonal variation in the diagnosis of IDDM hasbeen considered as indirect evidence forenvironmental exposure in the developmentof IDDM. Recent studies have providedmore indirect evidence for an associationbetween viral infections and the pathogene-sis of IDDM, but definitive evidence for viral

From the Department of Epidemiology and Health Promotion (M.K., VJ.J.T.), the Diabetes and Genetic Epi-demiology Unit, National Public Health Institute, Helsinki, Finland; and the Centre for Metabolic Diseasesand Atherosclerosis (S.M., M.S., L.S., S.M.), The ME.D1.CO. Association, Cagliari, Italy.

Address correspondence and reprint requests to Marjatta Karvonen, PhD, Diabetes and Genetic Epi-demiology Unit, Department of Epidemiology and Health Promotion, National Public Health Institute, Man-nerheimintie 166, FIN-00300 Helsinki, Finland. E-mail: [email protected].

Received for publication 16 October 1997 and accepted in revised form 20 March 1998.Abbreviations: AIC, Akaike's information criterion.

causes of IDDM is still missing (17,18).Worldwide variation in the incidence of

IDDM, both between and within conti-nents, is prominent. Data on incidencereported during the 1980s and 1990s sug-gested a correlation between the incidenceand northern latitude or cooler averageyearly temperature (19). A striking excep-tion to this rule is the Mediterranean islandof Sardinia (20). The highest incidence inthe world is in Finland (21), but in Sardinia,the incidence is similarly high—approxi-mately three times higher than the averageincidence in Europe (22). Therefore, it isinteresting to compare the seasonal patternsof newly diagnosed IDDM between thesetwo high-risk populations.

This study compared the seasonal varia-tion in the onset of IDDM among childrenaged ^ 1 4 years in Finland and Sardinia,two areas that have different climates but anequally high incidence of IDDM, and exam-ined the suggested association between acool environment and an increased incidenceof IDDM. Limited data exist on seasonality ofthe IDDM incidence in Finland. During the1950s, the onset of IDDM seemed to be dis-tributed rather evenly throughout the year,although the number of new cases was high-est in September (23). During the periodfrom 1968 to 1979, seasonal variation wasnoted in the incidence of IDDM, with peaksin April and September reported in childrenaged ^ 1 4 years in northern Finland (24).The most recent report on the seasonal pat-tern in Finland showed a significantlydecreased incidence of IDDM in June andduring the period from November toDecember (25). In Sardinia, the high inci-dence was reported during 1989-1992, withpeaks in the fall and winter and a lower inci-dence in the summer (26).

RESEARCH DESIGN ANDMETHODS

FinlandPopulation data were obtained from theNational Population Registry, which isupdated continuously. The Finnish popu-lation aged ^14 years varied from 952,943to 968,280 during 1989-1992.

DIABETES CARE, VOLUME 21, NUMBER 7, JULY 1998 1101

Dow

nloaded from http://diabetesjournals.org/care/article-pdf/21/7/1101/586077/21-7-1101.pdf by guest on 27 D

ecember 2021

IDDM in Finland and Sardinia

Table 1—Number of newly diagnosed cases and the incidence oflDDM in children <15 years of age in Finland during 1987-1992 accord-ing to age, sex, and year

Year1989

IncidenceNumber

1990IncidenceNumber

1991IncidenceNumber

1992IncidenceNumber

All yearsIncidenceNumber

0-4

28.445

28.946

32.252

23.839

28.3182

Boys (age in years)5-9

44.374

34.758

41.569

37.261

39.4262

10-14

54.290

36.761

52.687

45.976

47.4314

All

42.3 (36.6-48.209

33.4 (28.3-38.165

42.1 (36.4-47.208

35.6 (30.3-40.176

38.4(35.6-41.758

1)

5)

8)

8)

1)

0-4

27.041

34.252

27.242

27.343

28.9178

Girls (age in years)5-9

27.043

38.251

41.566

39.562

36.6232

10-14

37.860

37.159

44.871

29.647

37.3237

All

30.6 (26.4-34.8)144

36.5(32.1-40.1)172

37.8 (33.4-42.3)179

32.1(28.2-36.1)152

34.3 (31.6-36.9)647

36.5

35.0

40.0

33.9

36.4

All

(32.9-40.0)353

(31.6-38.3)337

(36.3-43.6)387

(30.6-37.1)328

(34.4-38.3)1,405

Age-adjusted 95% CI is given in parentheses.

Since 1987, all hospitals in Finlandtreating diabetic children have participatedin the prospective nationwide registrationof childhood IDDM (27). The case ascer-tainment is virtually 100% complete.Details of the procedures in case-ascertain-ment have previously been described else-where (27). During 1989-1992, IDDMwas diagnosed in 1,405 Finnish children

< 14 years.

Finland is one of the northernmostcountries in the world. Although othercountries extend further north, the focus oftheir regional structure and distribution ofpopulation lies further south. Finland is inan area with a snowy forest climate anddamp cold winters. The mean temperaturein the warmest month of the year is > 10°Cand in the coldest month it is below — 3°C.Precipitation occurs in all months, and the

area has at least a short period of distinctsummer and winter (28). However, due tothe Gulf Stream, Finland's climate is morefavorable than that of most other areas atthe same latitude. The basic features of theFinnish climate and its regional variationsare mainly products of location in the mid-dle latitudes, with the Gulf Stream in theWest and the vast continent of Eurasia inthe East. Finland's location means that it

Table 2—Number of newly diagnosed cases and the incidence of IDDM in children <15 years of age in Sardinia during 1987-1992 accord-ing to age, sex and year

Year1989

IncidenceNumber

1990IncidenceNumber

1991IncidenceNumber

1992IncidenceNumber

All yearsIncidenceNumber

0-4

22.710

27.212

24.911

31.814

26.747

Boys (age in years)5-9

42.422

50.126

44.323

52.027

47.298

10-14

56.936

28.418

42.727

45.829

43.5110

Age-adjusted 95% CI is given in parentheses.

40

35

37

43

39

.7

.2

.3

.2

.1

All

(30.9-50.4)68

(25.9-44.6)56

(33.0-53.4)61

(33.0-54.3)70

(34.3-43.9)255

0-4

24.010

24.010

21.69

24.010

23.439

Girls (age in years)5-9

43.521

29.014

37.318

27.013

34.266

10-14

35.321

33.620

16.810

23.514

27.365

All

34.3(27.2-41.3)52

28.9(21.9-35.9)44

25.2(19.4-31.1)37

24.8(18.2-31.3)37

28.3 (24.0-32.6)170

All

37.5 (31.4-43.5)120

32.1 (26.2-37.9)100

31.3 (25.7-36.8)98

34.0 (28.0-40.0)107

34.4(30.6-37.1)425

1102 DIABETES CARE, VOLUME 21, NUMBER 7, JULY 1998

Dow


ecember 2021

Karvonen and Associates

lies at the polar front, at the boundary sep-arating the cold air masses of the arcticarea from the warmer subtropical masses.

SardiniaThe population denominator was obtainedfrom the 1991 population census for therelevant age-groups (29). The Sardinianpopulation aged ^14 years in 1991 was303,958. The Sardinian 1DDM IncidenceRegistry was established in 1993. The crite-ria for the diagnosis of IDDM were thoseadopted by the EUROD1AB ACE Study(22). Since 1 January 1989, all newly diag-nosed cases aged ^29 years as notified byall Sardinian hospitals, outpatient clinics,family doctors and pediatricians have beenregistered and reported in the newsletterissued by the Epidemiological Observatory(30), which served as a primary data source.The secondary, independent source was thelocal IDDM patient association, Associ-azione Diabete Infantile Giovanile Sardegna.The case ascertainment was 91%. During1989-1992, IDDM was diagnosed in 425Sardinian children aged ^14 years.

The climate in Sardinia is Mediter-ranean, with mild temperatures and a pro-nounced summer aridity. The fairly highaltitude (half of the island is above 500 m)modifies the temperature and the distribu-tion and total amount of rainfall. Averagetemperature during the summer is ~28°C,and during the cold season, it is ~10°C.

Statistical methodsThe average annual incidence rates werecalculated per 100,000 population peryear. The mid-year populations aged ^14years were used as the denominator. The95% CIs were estimated assuming the Pois-son distribution of the cases. Age adjust-ment of the rates was done using 5-yearintervals (0-4, 5-9, and 10-14 years) withthe proportion being one-third for eachgroup as the standard according to the pre-vious approach by the Diabetes Epidemiol-ogy Research International Study Group(32).

Seasonal patterns in the incidence ofIDDM were evaluated using the method asdescribed by Jones et al. (33). The methodfits sine waves to the incidence data with afundamental period of one cycle per year,and when necessary, higher harmonicscould be used. It also allows for differentlengths of time intervals and different sizesof populations at risk.

The incidence model for group i attime t can be written as follows:

CRUDE INCIDENCE IN MALES AND FEMALES, 1989-19923-MONTH MOVING AVERAGE

A 9°V

E 80RAG 70E

M 600NT 5 0

H

Y<°

1 30NC1 20DEN 10

A

i W• ij')_;.

: /Y°-'.

>\*ir*/' '• • , •* \ . ' \ ' II'•X, \ ; '.,/'• j \^y :'•• "~:'..; >'» '" v \ /

i : \y .

K i: v.s ;• • / '

\

• X\

• / \

\ r V^

' • • / '•'

1991

YEAR

CRUDE INCIDENCE IN MALES AND FEMALES, 1989-19923-MONTH MOVING AVERAGE OF THE 4 YEARS

Figure 1—A: The incidence of IDDM each month in Finland (—) and Sardinia (- - -), 1989-1992.Data are presented as 3-month moving averages for males (*) and females (O). B: The monthly inci-dence of IDDM during 1989-1992 in Finland (—) and Sardinia (- - -). Data are presented as 3-monthmoving averages of the 4-year data for males (O) and females (O).

HO =

all+fclt-T

where aih equals cihcos(<5ih) and 3i h equals—CihSin(Oih), tfj equals a parameter thatallows the size of each group to be arbitrary,b, equals (t — T/2) the time trend, cih equalsthe amplitude of the cyclic curve, O^equals the phase angle, P equals the funda-mental period of the seasonal model, hequals the harmonic of the fundamentalfrequency and p equals the number ofterms in the sum.

In this case, the 4-year data starting at 1January 1989 has been divided into 48 inter-vals, at the time points tl = 31, Q. = 31 + 28,t3 = 31 + 28 + 31 a day as the unit oftime and the fundamental period of seasonal

model P = 365. This method allows an arbi-trary shape for the seasonal effect by pre-senting the data as a short Fourier series upto three harmonics together with a possiblelinear trend. Likelihood ratio tests andAkaikes information criterion (AIC) (34)were used to determine the number of har-monics necessary to model the seasonal pat-tern adequately and to test differences amongsubgroups in the population. Maximum like-lihood estimation based on Poisson distribu-tion was used to estimate the parameters ofthe model. Seasonal pattern in incidence wasalso compared between males and femalesand between the three 5-year age-groups,each controlling for the others effect.

RESULTS— During 1989-1992, theaverage annual incidence of IDDM was


Dow


ecember 2021


Table 3—Number of newly diagnosed cases of IDDM per month in children <15 years of age during 1989-1992

FinlandBoys

1989199019911992

Girls1989199019911992

SardiniaBoys

1989199019911992

Girls1989199019911992

January

21161716

10151616

1131

19

3214

February

14141213

15151815

326

10

5744

March

309

1911

12151214

12455

5222

April

14111912

15121711

6676

7542

May

15112219

1212127

4382

5222

June

1046

11

1069

10

4463

0141

July

19151721

10131711

2112

2324

August

25171718

1313169

3411

3251

September

12101823

1217179

3742

7332

October

20192012

12231312

61379

5545

November

17251711

9171518

10296

2317

December

121424

9

14141720

4765

4953

36.4/100,000 (95% CI 34.4-38.3) in Fin-land (Table 1) and 34.4/100,000 (95% CI31.3-37.9) in Sardinia (Table 2). In bothcountries, the average incidence of IDDMwas higher in males than in females. Theincidence among Finnish females washigher than that among Sardinian females.

The 3-month moving averages ofIDDM incidence are shown in Fig. LA andthe number of newly diagnosed cases foreach month are shown in Table 3. Therewere six monthly peaks in incidence inmales in Sardinia when the incidenceexceeded 40/100,000 per month (early1989, late 1989, late 1990, early 1991, andlate 1991). Although the monthly variationin incidence was less in Finnish males thanSardinian males, high peaks in incidencewere also observed during these periods.The month-to-month variation in incidencewas more pronounced in Sardinia than inFinland but without any obvious pattern inthe incidence (Fig. 1A). The low and highpoints in monthly incidence did not occur inthe same months in these two populations.

The average monthly incidence ofIDDM calculated for the 4 years (Fig. IB)was significantly lower in June than theyearly average among both sexes (P = 0.000in males and P = 0.0064 in females) in Fin-land. In Sardinia, the incidence in June

was significantly lower than the yearly aver-age only among females (P = 0.0005).Among Sardinian males, the incidence inJuly (P = 0.0001) and August (P = 0.0208)was significantly lower than the yearly aver-age incidence, whereas in October (P =0.041) the incidence was significantlyhigher than the yearly average incidence.

To find out whether the variation in theIDDM incidence within the calendar yearwas random or whether a seasonal patternexists, the incidence was modeled using themethod by Jones et al. (33). The selection ofbest models for seasonal variation in theincidence was based on the AIC, with theminimum AIC being the best. The changein —2 (log-likelihood) for models was testedusing a x2 distribution with degrees of free-dom equal to the difference in the numberof parameters. Tables 4 and 5 show the val-ues of —2(log-likelihood), the number ofestimated parameters, and AIC from theanalysis of seasonal variation in incidence,fitting different coefficients to each sex (two)and age-groups (three) with and withouttime trend for selected poolings.

In the first analysis, different coefficientswere fitted to each of the three age-groups(0-4, 5-9, 10-14 years) and to both sexesfor up to three harmonics with and withouttrend. The best fit in this test for the Sardin-

ian data was the model with one harmonicwithout trend and with the minimum AICof —20.40 (Table 4). For the Finnish data,the best-fit model was with null harmonics,no trend, and with the minimum AIC of 0.0(Table 5). Seasonal patterns in incidencewere also compared between males andfemales and among the three 5-year age-groups, each controlling for the others effect.After several poolings of the incidence data,the comparison of the AICs showed thatcombining the two youngest age-groups andboth sexes was acceptable for both the Sar-dinian (Table 4) and the Finnish data (Table5). For the Sardinian data, the best fit wasthe model with two harmonics withouttrend and with the minimum AIC of—31.61 (Table 4). The summary of stepwisefitting of these models from zero to threeharmonics is shown in Table 6. In Sardinia,there was a significant seasonal pattern in theIDDM incidence with two harmonics with-out trend in both age-groups and the mini-mum AIC of —20.00 in the age-group 0-9years, and minimum AIC of —11.61 in theage-group 10-14 years. For the Finnishdata, the best fit was also the model with twoharmonics without trend and with the min-imum AIC of —17.34 (Table 5). The age-group 0-9 years showed a significantseasonal pattern with one harmonic (mini-


Dow


ecember 2021


Table 4—Summary table for model selection, Sardinia

For six age-sex groupswithout pooling*

For four age-sex groupsafter pooling males and females

For three age-sex groups after poolingtwo lower ages and males and females t

Number ofharmonics

0123

012

3

0123

- 2 log-likelihood

0.00-44.40-62.59-70.82

0.00-38.61-49.50-54.00

0.00-36.85-47.61-50.44

No trend

Number of estimatedparameters

0122436

06

12

18

048

12

- 2 log-likelihood

-3.71-48.56-67.51-74.37

-2.01-40.90-53.06-57.15

-1.95-38.56-50.47-53.11

Trend


6183042

39

1521

26

1014

AICNo trend

0.00-20.40-14.59

1.18

0.00-26.61-25.50-18.00

0.00-28.85-31.61-26.44

Trend

8.29-12.56

-7.519.63

3.99-22.90-23.06-15.15

2.05-26.56-30.47-25.11

"Age-groups: 0-4, 5-9, and 10-14 years; tage-groups: 0-9 and 10-14 years.

mum AIC of —13.93), but in the age-group0—14 years, there were two harmonics (min-imum AIC of-7.07).

The observed (pointwise) and esti-mated seasonal patterns in Finland andSardinia for sexes pooled in two age-groups

(0-9 and 10-14 years) with an overall 95%CI for the curves are shown in Figs. 2 and3. There are some points that fall outsidethe CI, indicating that pointwise month-to-month variation is great and the model isnot perfect, although it is the best fit. In

Sardinia, there were two distinct cycles dur-ing the calendar year, with a decreased inci-dence during warm months in bothage-groups (0-9 and 10-14 years). In theage-group 0-9 years, the incidence waslow during the period from April to August

Table 5—Summary table for model selection, Finland

For six age-sex groupswithout pooling*

For four age-sex groupsafter pooling males and females

For three age-sex groups after poolingtwo lower ages and males and females!

Number ofharmonics

0123

0123

0123

- 2 log-likelihood

0.00-23.84-40.33-50.16

0.00-19.96-34.18-39.62

0.00-19.27-33.34-37.75

No trend


0122436

06

1218

048

12

* Age-groups: 0-4, 5-9, and 10-14 years; tage-groups: 0-9 and 10-14 years.

- 2 log-likelihood

-7.07-30.46-46.42-56.48

-3.57-21.96-35.61-41.16

-2.45-20.64-34.03-38.60

Trend


6183042

39

1521

26

1014

AICNo trend

0.000.167.67

21.84

0.00-7.96

-10.18-3.62

0.00-11.27-17.34-13.75

Trend

4.935.54

13.5827.52

2.43-3.96-5.61

0.84

1.55-8.64

-14.03-10.60


Dow


ecember 2021


Table 6—Summary for stepwise fitting of the model for age-groups 0-9 and 10-14 years withpooled boys and girls in Sardinia and Finland, 1 fanuary 1989 to 31 December 1992

Sardinia

Finland

Number ofharmonics

0123

0123

Age-group 0-9 years

- 2 log-likelihood

0.00-23.09-28.00-30.54

0.00-17.93-18.28-19.14

MALE & FEMALENO TREND / 1

AIC

0.00-19.09-20.00-18.54

0.00-13.93-10.28

-7.14

/ 0-9 YEARSHARMONIC

Age-group

- 2 log-likelihood

0.00-13.76-19.61-19.90

0.00-1 .34

-15.07-18.61

10-14 years

AIC

0.00-9 .76

-11.61-7 .90

0.002.66

-7 .07-6 .61

DE 1.50

N 1.35

cE 1.20

V '.os

A 0 90RI 0.75

A

" 060

I 045

MALE & FEMALE / 10-14 YEARSNO TREND / 2 HARMONICS

2 3 4 5 6 7

MONTH

Figure 2—Estimated and observed (—) seasonal pattern in the diagnosis of IDDM in both sexes inSardinia, 1989-1992. The value 1 on y-axis denotes the mean of the incidence during the entire studyperiod for the particular age-group. The inner interval of the two CIs shown is the pointwise 95% Cl,and the outer inteiyal is a 95% Clfor the model. The equations for the seasonal patterns are as follows:males and females aged 0-9 years: Aft) = 1 + 0.3876cos(2-nt/365) - 0.0635sin(2in/365) -0.1945COS(4TTI/365) + 0.0309sin(4irt/365); males and females aged 10-14 years: Aft) = J +0.3316COS(2TH/365) + 0.1737sin(2>nti365) - 0.0877cos(4in/365) ~ 0.2418sin(4Trt/365).

and higher during early spring and fall.Among older children (aged 10-14 years),the decrease in incidence was seen about 1month later than it was seen in youngerchildren during the period from May toSeptember, while during the rest of theyear, the incidence in older children fol-lowed the pattern of younger children.Among the younger children (aged ^ 9years) in Finland, there was only one dis-tinct cycle within the calendar year, withlower rates during the early spring andsummer (from March to July) and higherrates during the rest of the year. Amongchildren aged 10-14 years, there were twodistinct cycles during the calendar year,with nadirs in incidence during the periodsfrom April to July and from October toDecember through January

CONCLUSIONS— During 1989-1992, the incidence of IDDM in childrenaged <14 years fluctuated to some extentwithin a calendar year in Sardinia and inFinland, countries with the highest rates ofIDDM in the world but with different cli-mates and genetic backgrounds of theirpopulations (25,26,35). The incidencerates were similarly high and higher inmales than in females in both countries. Aseasonal pattern in the incidence wasexamined using statistical methods thathave been developed and applied to ana-lyze the seasonal pattern in incidence in theDenver IDDM registry (33). This methodallows to test whether seasonal patternsexist among all study subjects, in males andfemales, and in age-groups. It can also beused to test whether two or more groupshave the same seasonal pattern in inci-dence. Although the number of cases inSardinia was only one-third of that in Fin-land, there were enough cases for modelingthe IDDM incidence. However, the resultsmust be interpreted cautiously, taking intoaccount that the model of seasonal patternin incidence is based on a low number ofcases in Sardinia and thus has less reliabil-ity and precision than the model based onFinnish IDDM data.

Although a clear seasonal pattern inIDDM incidence for sexes pooled and fortwo age-groups (0-9 and 10-14 years) wasfound in both countries, the pattern of sea-sonal variation in incidence deviated. InSardinia, there were two distinct cycles dur-ing the calendar year, with a decreased inci-dence during warm months in bothage-groups (0-9 and 10-14 years). Amongyounger children (aged ^ 9 years), a low


Dow


ecember 2021


MALE & FEMALE / 0 - 9 YEARSNO TREND / 1 HARMONIC

A 0.90

MALE & FEMALE / 10-14 YEARS

NO TREND / 2 HARMONICS

Figure 3—Estimated and observed (—) seasonal pattern in the diagnosis of IDDM in both sexes inFinland, 1989-1992. The value 1 on y-axis denotes the mean of the incidence during the entire studyperiod for the particular age-group. The inner interval of the two CIs shown is the pointwise 95% Cl,and the outer interval is a 95% Clfor the model. The equations for the seasonal patterns are as follows:males and females aged 0-9 years: Aft) = 1 + 0.1457cos(2im/365) - 0.1451sin(2Tt\J365); males andfemales aged 10-14 years: Aft) = 1 + 0.059cos(2-ai/365) + 0.0501sin(2irt/365) -0.1293cos(4-nt/365) + 0.2027sin(4v\J365).

incidence from April to August and ahigher number of cases diagnosed duringthe fall, winter, and early spring was themost visible seasonal pattern. Although theseasonal pattern from May to Septemberamong older children (aged 10-14 years)was nearly identical to that among youngerchildren, the decrease in incidence was ~ 1month later in older children as comparedwith younger children.

In Finland, the seasonal pattern in theIDDM incidence deviated from that in Sar-dinia among the younger children (aged^ 9 years). In Finland, there was only onedistinct cycle within the calendar year, with

lower rates during the spring and earlysummer and higher rates during the rest ofthe year. The nadir in the IDDM incidencein the spring or early summer amongyounger children was seen 1-2 monthsearlier than in Sardinia. Among childrenaged 10-14 years, two distinct cycles werefound during the calendar year, as was alsoseen in Sardinia, but the seasonal pattern inincidence was almost the opposite of thatamong children in Sardinia in the sameage-group. In Finland, the first nadir inthe incidence during the calendar year was~1 month earlier than in Sardinia. DuringAugust through October, the incidence in

Finland was increased, but this was theperiod of low incidence in Sardinia. Coin-cidentally with the nadir in incidence inFinland, around November was a period ofhigh incidence in Sardinia.

A seasonal variation in the onset ofIDDM, with higher rates during cooler sea-sons (autumn and winter) and lower ratesduring the summer, has previously beendescribed by other investigators (2-16,28,36-40). Our results showed that in bothcountries there were no high peaks in inci-dence during the cool months that signifi-cantly deviated from the overall incidence.Since the 1920s, there has been increasinginterest in the possibility that a viral infec-tion may play a role in the onset of IDDM(17,18,41,42). However, virological studiesin patients have not yet produced convinc-ing evidence to prove this hypothesis,although the possibility of numerousviruses in the etiology has been suggested.

In Sardinia, the higher incidence par-ticularly around March and April could beexplained by a precipitating effect of non-specific infections in people who alreadyhave an appreciable degree of (3-celldestruction and possibly experienced infec-tions during the preceding winter (42,43).In recent years, it has been reported that ayearly peak in the incidence of rubella,chickenpox, and mumps occurs in thespring among children ^14 years of age(44). However, the highest number ofIDDM cases has been diagnosed amongchildren aged ^ 9 years in October, a monthwhen the weather is still generally mild andviral infections are infrequent. It can bespeculated that seasonal variation in theincidence of IDDM in Sardinia would bepartly a consequence of the summer holi-days, because schools are closed from thebeginning of July to late September. Familyholidays and trips may cause a delay innoticing early symptoms in children and adelay in diagnosis until October, when viralinfections in Sardinia are infrequent (44).

In Finland, there was a seasonal patternwith one nadir among younger males andtwo nadirs among older males. This doesnot fit the assumption that the seasonalvariation in incidence is caused by a viralinfection. The occurrence of diseases causedby viruses, such as mumps, measles, andcoxsackie B, that have been proposed totrigger IDDM shows an epidemic patternand the duration of the high-incidenceperiod is often short. However, the veryhigh and relatively stable incidence ofIDDM in Finland supports the assumption


Dow


ecember 2021


that if a viral infection were involved in theetiology of diabetes, such a virus should beendemic in the country. The high and stableincidence of IDDM would also require thepresence of chronic or recurrent infectionsand thus a high endemic level of infectionsin the country. Almost all of Finnish chil-dren aged ^14 years have been vaccinatedagainst several common viruses, such asmumps, that have previously beendescribed as causing occasional epidemics.However, despite such a vaccination pro-gram, the incidence of IDDM has remainedhigh and is still increasing in Finland, espe-cially among the youngest age-groups (22).Our results suggest that if a viral infectioncauses the (3-cell damage, the behavior ofsuch a pathogen is very atypical of viruses,which usually cause short-term epidemicsin susceptible populations.

It is very likely that the seasonality inthe diagnosis of IDDM is not related tocausal factors that trigger IDDM, but maybe associated with the expression of symp-toms in people who already are at anadvanced stage in developing the disease,which may take months or years. Manyphysiological parameters (for instance,blood glucose, blood lipids, blood pres-sure, and body weight) and health habits(diet and physical activity) have a seasonalpattern. This may explain the seasonaleffect that we observed in the diagnosis ofIDDM. The beginning of school holidays inJune in Finland may explain the low rate ofnewly diagnosed IDDM in children in thismonth (25). The low incidence of IDDMduring the summer months is a real phe-nomenon made plausible not only by theusually low frequency of potentially pre-cipitating infections (43) but also by a par-allel phenomenon in an animal model inwhich the incidence of diabetes is reducedin the NOD mouse exposed to raised envi-ronmental temperature (45).

Differences in the seasonal pattern ofthe IDDM incidence between Sardinia andFinland cannot be explained by differencesin climate and temperature, the longerwarm period in Sardinia, or differences inother climatic phenomena. The results donot provide evidence in favor of a specificviral etiology of IDDM. There may be trig-gering events at certain times but they arelikely to be unspecific. Nevertheless, thequestion remains unsolved as far as why theincidence of IDDM in these two popula-tions is equally high despite differences inclimate, environment, and genetic back-ground.

References1. Hogglof B, Holmgren G, Wall S: Incidence

of insulin-dependent diabetes mellitusamong children in a North-Swedish popu-lation 1938-1977. Hum Herd 32:408-417,1982

2. Durruty P, Ruiz F, Carcia de los Rios M: Ageat diagnosis and seasonal variation in theonset of insulin-dependent diabetes inChile (Southern Hemisphere). Diabetologia17:357-360, 1979

3. Dahlquist G, Blom L, Holmgren G, HagglofB, Larsson Y, Streky G, Wall S: The epi-demiology of diabetes in Swedish children0-14 years: a six-year prospective study.Diabetologia 28:802-808, 1985

4. Tajima N, LaPorte RE, Hibi I, Kitagawa T,Fujita H, Drash AL: A comparison of theepidemiology of youth-onset insulin-dependent diabetes mellitus between Japanand the United States (Allegheny County,Pennsylvania). Diabetes Care 8 (Suppl.1): 17-23, 1985

5. Allen C, Palta M, Alessio DJD: Incidenceand differences in urban-rural seasonalvariation of type 1 (insulin-dependent) dia-betes in Wisconsin. Diabetologia 29:629-633,1986

6. Siemiatycki J, Colle E, Aubert D, CampbellS, Belmonte MM: The distribution of type 1(insulin-dependent) diabetes mellitus byage, sex, secular trend, seasonality, timeclusters, and space-time clusters: evidencefrom Montreal, 1971-1983. AmJ Epidemiol124:545-560, 1986

7. Joner G, Sovik O: Increasing incidence ofdiabetes mellitus in Norwegian children0-14 years of age 1973-1982. Diabetologia32:79-83, 1989

8. Schranz AG, Prikatsky V: Type 1 diabetes inthe Maltese Islands. Diabet Med 6:228-231,1989

9. Hamman RE, Gay EC, Cruickshanks KJ,Cook M, Lezotte DC, Klingensmith GJ,Chase HP: Colorado IDDM Registry: inci-dence and validation of IDDM in childrenaged 0-17 yr. Diabetes Care 13:499-506,1990

10. Metcalfe MA, Baum JD: Incidence ofinsulin-dependent diabetes in childrenaged under 15 years in the British Islesduring 1988. BMJ 302:443-447, 1991

11. Muntoni S, Stabilini M, Mancosu G, Sta-bilini L, Loddo S, Muntoni Sa: Epidemi-ologia del diabete tipo 1 in Sardegna.Metabolismo Oggi 11:83-89,1994

12. Green A, Andersen PK, Svendsen AJ,Mortensen K: Increasing incidence of earlyonset type 1 (insulin-dependent) diabetesmellitus: a study of Danish male birthcohorts. Diabetologia 35:178-182, 1992

13. Elamin A, Omar MI, Zein K, Tuvemo T:Epidemiology of childhood type I diabetesin Sudan, 1987-1990. Diabetes Care15:1556-1559, 1992

14. Goday A, Castell C, Tresserras R, Canela J,

Tberner JL, Lloveras G: Sex differences inseasonality at onset of type 1 (insulin-dependent) diabetes mellitus in Catalonia(Abstract). Diabetologia 35 (Suppl. 1):A131,1992

15. Kostraba JN, Gay EC, Cai Y: Incidence ofinsulin-dependent diabetes mellitus in Col-orado. Epidemiology 3:232-238, 1992

16. Lipman TH: The epidemiology of type Idiabetes in children 0-14 yr of age inPhiladelphia. Diabetes Care 16:922-925,1993

17. Szopa TM, Titchener PA, Portwood ND,Taylor KW: Diabetes mellitus due toviruses: some recent developments. Dia-betologia 36:687-695, 1993

18. Dahlquist G, Frisk G, Ivarsson S, SvanbergL, Forsgren M, Diderholm H: Indicationsthat maternal coxsackie B infection duringpregnancy is a risk factor of childhood onsetIDDM. Diabetologia 38:1371-1373,1995

19. Karvonen M, Tuomilehto J, Libman I,LaPorte R, for the WHO DIAMOND ProjectGroup: A review of the recent epidemiolog-ical data on the worldwide incidence oftype 1 (insulin-dependent) diabetes melli-tus. Diabetologia 36:883-892,1993

20. Muntoni S, Songini M, Sardinian Collabo-rative Group for Epidemiology of IDDM:High incidence rate of IDDM in Sardinia.Diabetes Care 15:1317-1322,1992

21. Tuomilehto J, Virtala E, Karvonen M,Lounamaa R, Pitkaniemi J, Reunanen A,Tuomilehto-Wolf E, Toivanen L, the DiMeStudy Group: Increase in incidence ofinsulin-dependent diabetes mellitus amongchildren in Finland. Intj Epidemiol 24:984-992.1995

22. Green A, Gale EAM, Patterson CC, for theEURODIAB ACE Study: Incidence of child-hood-onset insulin-dependent diabetesmellitus: the EURODIAB ACE Study. Lancet339:905-909, 1992

23. Somersalo O, Hiekkala H, Rantakallio P Tuu-teri L: Studies in childhood diabetes. Reviewof 359 cases. I. Onset and physical develop-ment. Ann Paediatr Fenn 6:253-274,1960

24. Reunanen A, Akerblom HK, Kaar ML:Prevalence and ten-year (1970-1979) inci-dence of insulin-dependent diabetes inchildren and adolescents in Finland. AdaPaediatr Scand 71:893-899, 1982

25. Karvonen M, Tuomilehto J, Virtala E,Pitkaniemi J, Reunanen A, Tuomilehto-Wolf E, Ckerblom H, for the DiMe StudyGroup: Seasonality in the clinical onset ofinsulin-dependent diabetes mellitus inFinnish children. AmJ Epidemiol 143:167-176.1996

26. Muntoni SA, Stabilini L, Stabilini M, Man-cosu G, Muntoni S: Steadily high IDDMincidence over 4 years in Sardinia. DiabetesCare 18:1600-1601,1995

27. Tuomilehto J, Lounamaa R, Tuomilehto-Wolf E, Reunanen A, Virtala E, Kaprio EA,Akerblom HK: Epidemiology of childhood


Dow


ecember 2021


diabetes mellitus in Finland: background ofa nationwide study of type 1 (insulin-depen-dent) diabetes mellitus. Diabetologia 35:70-76,1992

28. RikkinenK: A Geography of Finland, lsted.Lahti, Finland, University of Helsinki, LahtiResearch and Training Centre, 1992

29. Istituto Nazionale di Statistica, Popolazionee Abitazioni, Fascicolo Regionale Sardegna:13 Censimento Generale della Popolazione edelle Abitazioni. Cagliari, Italy, ISTAT, 1991

30. Sitzia G (Ed.): Epidemiologia Sardegna. Vol.4. Cagliari, Italy, Osservatorio Epidemio-logico Regionale, 1994

31. Bethemont J, Pelletier J: Italy a GeographicalIntroduction. 1st ed. Paris, Bordas, 1983

32. Diabetes Epidemiology Research Interna-tional Group: Geographic patterns of child-hood insulin-dependent diabetes mellitus.Diabetes 37:1113-1119, 1988

33. Jones R, Ford P, Hamman R: Seasonalitycomparisons among groups using incidencedata. Biometrics 44:1131-1144,1988

34. Akaike H: Information theory and anextension of the maximum likelihood prin-ciple. In Second International Symposium on

Information Theory. Petrov BN, Csaki F, Eds.Budapest, Akatemia Kaido, 1973, p.267-281

35. Cavalli-Sforza LL, Piazza A: Humangenomic diversity in Europe: a summary ofrecent research and prospects for thefuture. EurJHum Genet 1:3-18, 1993

36. Christau B, Kromann H, Ortved-AndersenO, Christy M, Buchard K: Incidence, sea-sonal and geographical patterns of juvenile-onset insulin-dependent diabetes mellitus inDenmark. Diabetologia 13:281-284,1977

37. Ludvigsson J, Afoke AO: Seasonality of type1 (insulin-dependent) diabetes mellitus:values of C-peptide, insulin antibodies andhaemoglobin Ale show evidence of morerapid loss of insulin secretion in epidemicpatients. Diabetologia 32:84-91,1989

38. Bingley PJ, Gale EAM: Incidence of insulindependent diabetes in England: a study inthe Oxford region, 1985-6. BMJ 298:558-560,1989

39. Serrano Rios M, Moy CS, Martin Serrano R,Minuesa Asensio A, de Tomas Labat ME,Zarandieta Romero G, Herrera J: Incidenceof type 1 (insulin-dependent) diabetes mel-

litus in subjects 0-14 years of age in theComunidad of Madrid, Spain. Diabetologia33:422-424, 1990

40. Podar T, Tuomilehto-Wolf E, TuomilehtoJ,LaPorte RE, Adojaan B: Insulin-dependentdiabetes mellitus in native Estonians andimmigrants to Estonia. Am J Epidemiol135:1231-1236,1992

41. Gundersen E: Is diabetes infectious origin?J Infect Dis 4:197-202, 1927

42. Gamble DR: A possible viais etiology forjuvenile diabetes. In The Genetics of DiabetesMellitus. Creutzfeld W, KobberlingJ, Neel V,Eds. New York, Springer-Verlag, 1977, p.95-105

43. Bennett PH: Changing concepts of the epi-demiology of insulin-dependent diabetes.Diabetes Care 8 (Suppl. l):29-33,1985

44. Sitzia G (Ed.): Epidemiologia Sardegna. Vol.3. Cagliari, Italy, Osservatorio Epidemi-logico Regionale, 1994

45. Williams AJK, Krug J, Lampeter EF, Mans-field K, Beales PE, Signore A, Gale EA,Pozille P: Raised temperature reduces theincidence of diabetes in the NOD mouse.Diabetologia 33:635-637,1990


Dow


ecember 2021

Comparison of the Seasonal Pattern in the Clinical - Diabetes Care

Documents

Transcript of Comparison of the Seasonal Pattern in the Clinical - Diabetes Care