seasonal variation of hip fracture at three latitudes
TRANSCRIPT
Seasonal variation of hip fracture at three latitudes
Stuart Douglasa,$, Agnes Bunyanb, Kwok Hing Chiuc, Bruce Twaddled,Nicola Ma�ullib,*
aDepartment of Medicine & Therapeutics, University of Aberdeen Medical School, Polwarth Building, Foresterhill, Aberdeen AB25 2ZD, UKbDepartment of Orthopaedic Surgery, University of Aberdeen Medical School, Polwarth Building, Foresterhill, Aberdeen AB25 2ZD, UK
cDepartment of Orthopaedics and Traumatology, Chinese University of Hong Kong, Prince of Wales Hospital, Clinical Services Building, Shatin,
New Territories, Hong Kong, People's Republic of ChinadDepartment of Orthopaedic Surgery, Auckland Hospital, Auckland, New Zealand
Accepted 19 July 1999
Abstract
We studied the seasonal variation of hip fracture admissions at three di�erent latitudes: Scotland (568 North; 54,399admissions); Shatin, Hong Kong (228 North; 4180 admissions); and Auckland, New Zealand (368 South; 2257 admissions). Wecalculated the extent of seasonal variation (amplitude) and the time of year of the peak value (acrophase) by ®tting a sine curve
to monthly data using cosinor analysis. A signi®cant seasonal variation was found in all three countries, at a high level inScotland ( p < 0.01) and Hong Kong ( p < 0.001), but just signi®cant in New Zealand ( p< 0.05). The extent of the seasonalchange was very similar in Scotland and New Zealand, but, as expected, the peak in New Zealand (early September) was
approximately six months ahead of Scotland (mid February). In Hong Kong, the amplitude was three times greater than inScotland and the peak occurred a month earlier. There is neither snow nor ice in Hong Kong, and this provides powerfulevidence against a major in¯uence of conditions underfoot causing extra falls in winter. In Scotland there was a signi®cantincrease in the proportion of deaths in winter as compared to summer. The Scotland/Hong Kong amplitude di�erence is
striking, but it is unknown whether this has a genetic or environmental explanation. The cause of seasonal death di�erence to agiven injury is also unknown. Possible mechanisms are discussed, but the purpose is to report two new epidemiological features,without wild speculative hypotheses. The ®ndings should be viewed as leads to further epidemiological, clinical and more basic
research. # 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Hip fracture; Season; Seasonal; Latitude
1. Introduction
Seasonal variation in the rate of proximal femoralfractures was ®rst suggested in 1955 in Dundee,Scotland [1] based on a small number of patients.Since then, reports with larger numbers have comefrom several parts of the World: Australia [2]; UnitedStates [3,4]; Canada [5]; and Hong Kong [6]. Thesehave con®rmed an excess in winter months, but not allreports have con®rmed seasonality [7,8].
Three aetiological mechanisms for winter excess
have been suggested. Increased numbers of falls in
winter in cold weather due to conditions underfoot
[9] could be the explanation. However, others have
reported as many or more fractures occurring
indoors [10,11] as outdoors. Seasonal increase in
impaired balance in the elderly in the winter is
another logical possibility [12]. Possibly, 6% of frac-
tures are spontaneous [13], and the fall is subsequent.
Thirdly, bone quality may also be impaired in winter
[14±17].
Seasonality has previously been described at various
latitudes [1±6]. We collected three series at di�erent
latitudes to make an epidemiological comparison of
some features of seasonality, such as extent of seasonal
Injury, Int. J. Care Injured 31 (2000) 11±19
0020-1383/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S0020-1383(99 )00192 -8
www.elsevier.com/locate/injury
$ Deceased
* Corresponding author. Tel.: +44-1224-840-959; fax: +44-1224-
685-373.
E-mail address: n.ma�[email protected] (N. Ma�ulli).
variation and position of the peak. Data fromScotland (558 North), Shatin, Hong Kong (228 North)and Auckland, New Zealand (368 South) wereobtained.
2. Patients and methods
Scotland (568 North) Ð We obtained data from theScottish Health Service Information and StatisticsDivision (ISD) relating to all Scottish hospital dis-charges for hip fracture during the period 1991±1995.The information collected is based upon ScottishMorbidity Records (SMR), which are collected locallyat hospitals throughout Scotland, and subsequentlyamalgamated by the ISD. In the period under study,the population of Scotland remained constant at ap-proximately 5.1 million inhabitants (O�ce ofPopulation Censuses and Surveys 1998) [18]. All indi-viduals discharged after a hospital admission inScotland are included in the SMR data [19]. The num-bers of admissions were 54,399 over the years 1991±1995, a sample su�ciently large to determine whetherthere was seasonal change in the monthly death rate.Deaths within hospitals were established monthly. Theaggregated monthly data were used as a total over 5years, i.e. each January value was added together togive the aggregated January value used in the analyses.This procedure was also applied to the Hong Kongand New Zealand data.
Shatin, Hong Kong (228 North) Ð Data came froma single hospital, the Prince of Wales Hospital,Chinese University of Hong Kong, Shatin, HongKong. Four-thousand, one hundred and eightlypatients were identi®ed from computerised orthopaedicrecords over the years 1985±1995.
Auckland, New Zealand (368 South) Ð Data camefrom a single hospital, Auckland Hospital, Auckland,New Zealand. Two-thousand, two hundred and ®fty-seven patients were available from computerised ortho-paedic records for the years 1993±1996.
3. Data accuracy
The accuracy of epidemiological data collected insuch a fashion needs to be questioned. The infor-mation in Scotland (ICD coding) came from nationallycollated data coded under the ICD 9 code 820 (frac-ture of the neck of femur). In the ISD system, frac-tures were not divided into cervical andpertrochanteric fractures, and we have therefore nodata in this respect. The ISD conducts quality assess-ment projects on a regular basis. In the latest oneavailable, the accuracy of SMR main diagnosis data is89.9%, and the accuracy of main surgical procedure
coding is 90.7% [19]. SMR-1 is an episode basedpatient information system relating to all patients dis-charged from non-psychiatric, non-obstetric depart-ments in Scottish hospitals. A record is raised when apatient is discharged from hospital having been an in-patient or a day-case patient, or changes consultant, oris transferred to another hospital. An SMR-1 return isgenerated on discharge (dead or alive) from hospital.The system uses the International Classi®cation ofDiseases (ICD) coding system. Using SMRI (ScottishMorbidity Records) returns, the numbers of admis-sions classi®ed under ICD9 820 were assembled. TheISD data were double checked individually, and foundto be accurate.
The data from Hong Kong and New Zealand camefrom individual hospital orthopaedic records. One ofus (AB) visited both Hong Kong and New Zealand toco-ordinate data collection.
4. Statistical analysis
4.1. Correction for unequal month length
Each monthly total was corrected to the value itwould have had in a month of 31 days. The value forFebruary was multiplied by 1.097, and that for 30 daymonths by 1.033.
4.2. Cosinor analysis
A computer programme called cosinor analysis wasused [20,21]. This is equivalent to ®tting a cosine orsine curve to the data, and can be undertaken usingstandard statistical software. The analyses use the 12monthly totals or averages to determine how much ofthe seasonal variation can be explained by a sinecurve. The programme also ®ts the best curve to thedata by least squares. The output from the programmeincludes the following.
1. The extent of seasonal variation: the amplitude A isthe largest distance from the mean to the peak andis half of the total seasonal variation.
2. A multiple correlation coe�cient R and a corre-sponding signi®cance level ( p ). R is the proportionof the between-month variation explained by thesine curve. R 2 is the proportion of varianceexplained by a cosine function. The signi®cancelevel is based on an F test. Of the (12ÿ1=11)degrees of freedom available, two are used to esti-mate the curve, leaving nine degrees of freedom forthe deviations from the curve, based on the unex-plained variation. The con®dence intervals are forthe ®tted model.
3. The values of co-e�cient of sin(t ) and of cos(t ) are
S. Douglas et al. / Injury, Int. J. Care Injured 31 (2000) 11±1912
used to de®ne the day of the year on the curve withthe peak value, using the formula
365
360� tan ÿ1
coefficient of sin t
coefficient of cos t
4.3. Normal Approximation to the Poisson Distribution(NAPD)
This method establishes whether a particular monthvalue di�ered signi®cantly from the mean, i.e. whetherit was a peak or a trough. While the monthly datamay not ®t a sine curve, individual monthly valuesmay di�er from the mean monthly ®gure. The methodassumes that the population is large compared to thenumber of cases, that the cases are detected indepen-dently, and that no individual develops the conditionunder study twice. If the mean frequency is largeenough, a normal approximation can be used to calcu-late 95% con®dence intervals from the mean, but thesewill be approximate and should be treated with cau-tion. The monthly mean is compared with the annualmean. Since the model is not very powerful, only thosevalues having a p value of <0.01 are considered.
5. Results
5.1. Original data
Table 1 provides the original data on aggregatedmonthly values and a total for each series. For eachcountry the totals were: Scotland, 54399; Shatin, HongKong, 4180; and Auckland, New Zealand, 2257. The®gures are shown by sex and the proportion of malesin each series was Scotland 21.8%, Hong Kong29.8%, and New Zealand 31.7%.
The data for temperature are also shown. Everymonth in Hong Kong is warmer than in Scotland. TheScottish summer temperatures are below winter tem-peratures in Hong Kong.
In Auckland, summers are a little warmer thanScottish summers, and winters much less cold. Whenthe New Zealand temperature data are advanced bysix months so that January is compared with July,every month in New Zealand is colder than HongKong.
5.2. Cosinor analysis
This is shown in Table 2 and Fig. 1. The results arehighly signi®cant for the whole series in Scotland( p < 0.01) and Hong Kong ( p < 0.001) with a peakin February (13th) in Scotland and January (13th) inHong Kong. The extent of the seasonal variation isT
able
1
Aggregatedmonthly
values
andatotalforeach
series
Months
Scotland
HongKong
New
Zealand
Tem
perature
8CNew
Zealand
Total
allages
Males
allages
Fem
ales
allages
Total
allages
Males
allages
Fem
ales
allages
Total
allages
Males
allages
Fem
ales
allages
Scotland
HongKong
8Cstarts
January
8Cstarts
July
January
5099
1145
3954
467
133
334
169
52
117
3.5
15.5
19.5
10.5
February
4597
1029
3568
370
115
255
144
45
99
3.5
15
19.5
11
March
4772
1002
3770
391
104
287
180
56
124
517.5
18.5
12.5
April
4400
917
3483
289
99
190
155
54
101
7.5
21.5
16.0
14.3
May
4673
1064
3609
354
111
243
192
50
142
10
25.5
14.0
15.5
June
4451
978
3473
288
85
203
188
57
131
13
27.5
11.5
17.5
July
4370
939
3431
297
101
196
210
77
133
14.5
28.5
10.5
19.5
August
4288
950
3338
287
77
210
208
75
133
14.5
28.5
11.0
19.5
September
4185
913
3272
301
109
192
185
56
129
12.5
27.0
12.5
18.5
October
4380
952
3428
350
99
251
200
73
127
9.8
25.6
9.0
16
Novem
ber
4289
913
3376
378
102
276
204
60
144
6.5
20.5
15.5
14
Decem
ber
4895
1080
3815
408
111
297
222
61
161
4.5
17.5
17.5
11.5
Total
54,399
11,882
42,517
4180
1246
2934
2257
716
1541
%100
21.8
78.2
100
29.8
70.2
100
31.7
68.3
S. Douglas et al. / Injury, Int. J. Care Injured 31 (2000) 11±19 13
three times as great in Hong Kong as in Scotland.Similar results are found in females alone and in allpatients (males+females) 70 years and over. The ®nd-ings are very convincing. However, the result in malesalone is just signi®cant in both series and, for thosepatients under 70 years, no signi®cance was found.
For New Zealand (Table 2, Fig. 2), the seasonality®ndings are less convincing. For the total series (andmales alone and those <70 years), there was signi®-cance, but with a p value only reaching the 5% level.
5.3. NAPD
Under the assumptions outlined in the Methods sec-tion, the Poisson distribution was a good model. Allthe ®ndings in Scotland and Hong Kong with p valuesat <0.01± < 0.001 on cosinor analysis had signi®cantpeaks (Table 2). In the Scotland data, peaks were alsofound in males and for all patients under 70 years(Table 2). To illustrate the technique, Fig. 3 showsScottish data. No ®ndings in the New Zealand datawere signi®cant.
5.4. Temperature (Table 2)
The extent of seasonal change (amplitude) is similarin Hong Kong and New Zealand, and in both casesthe extent of seasonal change is much less thanScotland.
5.5. Acrophase position
The peak in Hong Kong is in mid January, and inScotland it occurs in mid February. The high signi®-cance values justify these conclusions, but much lessconvincing is the late August early September peak inNew Zealand, with signi®cance only at the 5% level.This Southern Hemisphere ®nding is six months aheadof the Scotland data.
5.6. Deaths in Scotland (Table 3)
In every month, the percentage of deaths amongstmales was greater than in females. The main monthlydeath rate in males was 9.3% and 7.1% in females.
Table 2
Cosinor analysis
R P Amplitude Mean Acro NAPD ( p < 0.01)
Month Date Peak Troughs
Scotland
Total 0.88 < 0.01 7.3 4618 February 13 February December January
February
June August September
October November
Males 0.71 < 0.05 7.3 1008 February 11 February January February NS
Females 0.92 < 0.001 7.2 3609 February 12 February December January July August September
October
< 70 0.65 NS 7.3 768 January/February ± December January NS
r70 0.90 < 0.001 7.4 3849 February 17 February January February March July August September
October
Hong Kong
Total 0.91 < 0.001 20.1 354 January 13 January December January April June July August
Males 0.73 < 0.05 14 105 January/February 23 January NS August
Females 0.89 < 1.001 22.9 248 January 10 January December January June July August
September
< 70a 0.39 NS 8.1 90 ± ± NS August
r70a 0.89 < 0.001 24.4 257 January 10 January December January
February
April June July August
September
New Zealand
Total 0.7 < 0.05 10.3 191 August/September 7 September NS NS
Males 0.76 < 0.05 16.1 60 August/September 31 August NS NS
Females 0.49 NS 8.5 130 ± ± NS NS
< 70a 0.74 < 0.05 13.3 31 October 8 October NS NS
r70a 0.55 NS 9.4 133 ± ± NS NS
Temperature
Scotland 1.0 < 0.001 65.1 8 July/August 26 June NS January February
Hong Kong 0.99 < 0.001 30.9 22 July/August 22 July NS NS
New Zealand 0.99 < 0.001 29.3 15 January/February 26 January NS NS
a Ages of some patients not known.
S. Douglas et al. / Injury, Int. J. Care Injured 31 (2000) 11±1914
Fig. 1. Results of cosinor analysis in Scotland and Hong Kong. The upper part shows corrected monthly values on the vertical axis, with the
details of R and p values, and amplitude (vertical arrow). The lower part provides the same data, but with the vertical axis as a percentage of the
mean. Note the much smaller extent of the seasonal variation in Scotland. (Continuous line=®tted sine curve; discontinuous curves=95% con®-
dence intervals). The horizontal axis shows months of the year. The discontinuous horizontal line is the mean value.
Table 3
% Death by month in Scotland
Males and females % of monthly admission Males % of monthly admission Females % of monthly admission
January 424 8.32 115 10.04 30.9 7.82
February 407 8.65 115 12.58 288 8.07
March 366 7.67 102 10.18 264 7.00
April 354 8.05 76 8.29 278 7.98
May 330 7.06 90 8.46 240 6.65
June 287 6.45 76 7.77 211 6.08
July 320 7.32 89 9.48 231 6.73
August 290 6.76 77 8.11 213 6.38
September 271 6.48 74 8.11 197 6.02
October 342 7.81 76 7.98 266 7.76
November 350 8.76 91 9.97 259 7.67
December 399 8.15 115 10.63 284 7.44
Cosinor analysis
R 0.86 0.75 0.78
p < 0.01 < 0.05 < 0.05
Amplitude % 12.0 15.8 11.1
Mean 7 9 7
Acrophase January January January
When 12 monthly ®gures for males and for femaleswere analysed using the t-test, the males showed sig-ni®cantly greater death rates (t= 4.61, p < 0.001).
The ®ndings on seasonality of death are shown inTable 3 and Fig. 4. The percentage of death was seaso-nal, with approximately one third more deaths inFebruary than in July. The cosinor analysis on thosemonthly percentages was signi®cant. For each patient,for example, admitted in February the risk of deathwas greater than in July.
6. Discussion
The results of this study agree with most previous
studies [1±6] where seasonality with a winter peak hasbeen found. Two reports have failed to con®rm anysigni®cant seasonal change [7,8]. However, to establishseasonal variation, su�cient numbers are required, andthis may explain the discrepancy between these tworeports and the main body of the literature on thissubject. The Scottish data were assembled nationally,and may therefore not have the quality as that col-lected at single hospitals in Hong Kong and NewZealand. Recognised limitations of the Scottish dataare described in the Patients and Methods section.Some may criticise the comparison of national datawith that collected from one hospital site. However,national data for Hong Kong and New Zealand arenot available, and, given the health system organis-ation in Hong Kong and in New Zealand, all patientswith a hip fracture in their respective catchment areaswere admitted to the hospitals in the study. Given thepresent status of data collection, the reality of a studysuch as this necessitates acceptance of what is avail-able, even if this may be short of scienti®c ideal.
The results from the cosinor analysis of data fromHong Kong and Scotland provide very convincing evi-dence of seasonality. This method is very powerfulprovided the data ®t a sine curve. There are, however,limitations to this method [22]. For example, if therewere more than one peak, ®tting a sine curve would beinappropriate. For that reason, the NormalApproximation to Poisson was used. This is a lesspowerful method, and only values at the 1% level ofsigni®cance were used. These ®ndings support the con-clusions made on cosinor analysis in Scotland andHong Kong, but not in New Zealand. However, theAuckland ®ndings on cosinor analysis were signi®cantat the 5% level; the numbers in this SouthernHemisphere country were smaller, and the conclusionshave to be drawn with less certainty. The convincing
Fig. 2. Cosinor analysis results on the New Zealand data. The verti-
cal axis shows corrected monthly values and the horizontal axis
shows months of the year. The details of p and R values and the
amplitude are given. Comparison should be made with Fig. 1 (upper
left): the peak in New Zealand is about six months advanced.
(Continuous line=®tted sine curve; discontinuous curves=95 percent
con®dence limits). The horizontal discontinuous line is the mean
monthly value.
Fig. 3. Normal approximation to Poisson distribution on the
Scottish data. The vertical axis gives the number of patients, and the
horizontal axis the month of the year. During December, January,
February and March, the 99% con®dence intervals are above the
monthly mean, and in July, August, September, October and
November they are below.
Fig. 4. The vertical axis gives a monthly percentage of death and the
horizontal axis gives months of the year. The cosinor analysis pro-
vides the ®tted curve with its 95% con®dence intervals. The peak
was January. Signi®cance and amplitude values are provided.
Proportionately, there are more deaths in winter than in summer.
S. Douglas et al. / Injury, Int. J. Care Injured 31 (2000) 11±1916
seasonality in Hong Kong may be due to the ease ofdemonstration when the amplitude is high.
The striking feature in the results is betweenScotland and Hong Kong. In Scotland, where it iscolder, the amplitude of seasonal variation amplitudeof hip fractures is signi®cantly lower than in HongKong. The high amplitude seasonal variation in HongKong is very powerful evidence against snow and iceon the ground being the central reason for the seasonalvariation, since these conditions underfoot do notoccur in Hong Kong. This con®rms and extends theobservations made by Chiu et al [6], based on a popu-lation from Hong Kong Island served by the QueenMary Hospital, Hong Kong University.
Seasonality in Auckland, although less signi®cant,was present. The coldest months in Scotland areDecember and January, and, in Auckland, July andAugust. The peak of hip fractures was in midFebruary in Scotland, and early September inAuckland, i.e. the expected di�erence of six months. Inboth countries, the peak was slightly beyond the cold-est time of the year.
The reason for the increased number of hip fracturesin winter is not understood. With advancing age, theelderly stagger and may well fall, and impaired co-ordination is more marked in winter than in summer[12]. If elderly people do feel the cold more, they arealso more likely to put on extra clothes, which in turnmay make them more clumsy. Winters are darker aswell as colder, and this may add to the number of fallsin this season. In comparison with wrist fractures, hipfractures tend to occur relatively more frequentlyindoors.
Photoperiod and sunlight could provide a seasonaltime setter for bone re-modelling, and bone mineraldensity (BMD) is raised in summer as compared towinter [14±17]. Bergstralh et al. [14] found BMD inthe lumbar spine to be 1.4% greater in late summerthan in winter. It could be that, in a seasonal re-mod-elling balanced process, this is towards bone resorptionin autumn/winter and bone formation in spring/sum-mer. There is extensive literature on seasonal andother changes in bone related biochemical parameters[23±34]. These include 25 hydroxyvitamin D3, para-thyroid hormone, osteocalcin, alkaline phosphataseand urinary pyridinium crosslinks. These are not dis-cussed here in detail, but they may re¯ect winter`weakness' in bone.
With the extent of seasonal variation in Hong Kongbeing much greater than in Scotland, the role of coldtemperatures as responsible for the seasonality of hipfracture must be questioned. The e�ect cannot resultfrom absolute fall in temperature. Scotland and HongKong are at very di�erent latitudes, raising the ques-tion whether photoperiod (the ratio of daylight tohours of darkness) is a key in¯uence.
Photoperiod and latitude have constant relation-ships. Weather has a general relationship to latitude,but the climate can be very di�erent at the same lati-tude in two geographical sites. In recent years, thepineal hormone melatonin has been readily measured,and understanding of the physiology of the photo-neuroendocrine system has improved [35]. Melatonin issecreted during the hours of darkness, and may there-fore in¯uence the time keeping of circadian rhythm.Seasonal changes of melatonin circadian rhythm havebeen described [35±37].
An interesting ®nding was that the risk of death forthe individual is greater in winter than in summer. Thebody's reaction to a standard traumatic stress mightbe less adequate in winter than in summer. Anotherpossible explanation could be that this is due to anaccompanying increment in winter from death due toarterial and venous thromboembolic disease.Respiratory infection might also add to the excessdeaths in winter.
In this discussion only two of the possible environ-mental and climatic variables Ð temperature andphotoperiod have been considered. Although hip frac-ture is associated with cold weather, this does notimply a causal relationship with temperature.
In seeking a reason for the greater amplitude inHong Kong, it is possible that the di�erences arise notfrom the climate, but from di�erences in the diseaseamongst Chinese. There is evidence that the frequencyof hip fracture is greater in Caucasian than Asianpopulations (Japanese [38], Chinese [39]). One possibleexplanation is a shorter hip axis in Asians [40].Collagen is the commonest protein in bone and poly-morphisms of collagen I alpha 1 gene are di�erent inAsians and Caucasians [41]. It is possible that changesin collagen rather than bone mineral density underlieseasonality of hip fractures.
The ®nding that males with a hip fracture su�erfrom a greater mortality rate than females is in accord-ance with other studies [42±45].
In conclusion, identi®cation of seasonal changeshave already encouraged clinical studies of vitamin Dand calcium in the prevention of osteoporotic fracturesand further larger studies are ongoing. Following bio-chemical studies demonstrating the correction ofundercarboxylation of osteocalcin by vitamin K, clini-cal trials are underway to test the e�cacy of vitaminsK in patients with osteoporosis. When haemorrhagicdisease of the newborn was common, it had a similarseasonal rhythm as fracture of the hip [10].
Acknowledgements
At the University of Aberdeen, we are gratefulto Professor J.D. Hutchison, Department of
S. Douglas et al. / Injury, Int. J. Care Injured 31 (2000) 11±19 17
Orthopaedics, for guidance and co-operation.Secretarial help was provided by Mrs N. Duncan,Department of Medicine and Therapeutics, and byMiss Linda Lothian, Department of OrthopaedicSurgery. The artwork was performed by theDepartment of Medical Illustration. One of us (ASD)was supported by funds which arose from theMaryland Medical Research Institute.
References
[1] Stewart IM. Fracture of the neck of the femur: incidence and
implications. British Medical Journal 1955;1:698±701.
[2] Lau EMC, Gillespie BG, Valenti L, et al. The seasonality of hip
fracture and its relationship with weather conditions in New
South Wales. Australian Journal of Public Health 1995;19:76±
80.
[3] Jacobsen SJ, Goldberg J, Miles TP, et al. Seasonal variation in
the incidence of hip fracture among white persons aged 65 years
and older in the United States, 1984±1987. American Journal of
Epidemiology 1991;133:996±1004.
[4] Jacobsen SJ, Sargent DJ, O'Fallon WM, et al. Population based
study of the contribution of cold weather to hip fracture season-
ality. American Journal of Epidemiology 1995;141:79±83.
[5] Levy AR, Mayo NE, Grimard G. Rates of transcervical and
peretrochantiric hip fractures in the province of Quebec,
Canada, 1981±1992. American Journal of Epidemiology
1995;142:428±36.
[6] Chiu KP, Ng TP, Chow SP. Seasonal variation of fractures of
the hip in elderly persons. Injury 1996;27:333±6.
[7] Parker J, Martin S. Falls, hip fractures and weather. European
Journal of Epidemiology 1991;10:441±2.
[8] Pedrazzoni M, Alfano FS, Malvi C, et al. Seasonal variation in
the incidence of hip fractures in Emilia-Romagna and Parma.
Bone 1993;14:857±63.
[9] Ralis ZA. Epidemic of fractures during periods of snow and ice.
British Medical Journal 1981;1:603±5.
[10] Douglas AS. Seasonality of hip fracture and haemorrhagic dis-
ease of the newborn. Scottish Medical Journal 1993;38:37±40.
[11] Shapiro S, Adelson L, Oseasohn R. A study of racial di�erence,
seasonal variation and temporal variation in hip fracture Ð re-
lated mortality, Cuyahoga Country, Ohio, 1958±1962, 1968±
1972. Journal of Chronic Diseases 1977;30:185±92.
[12] Bastow MD, Rawlings J, Allison SP. Undernutrition hypother-
mia and injury in elderly women with fractured femur: an injury
response to altered metabolism? Lancet 1983;i:143±6.
[13] Parker MH, Twemlow TR. Spontaneous hip fractures. Acta
Orthopaedica Scandinavica 1997;68:325±6.
[14] Bergstralh EJ, Sinaki M, O�ord KP, et al. E�ect of season on
physical activity score, back extensor muscle strength and lum-
bar bone mineral density. Journal of Bone and Mineral
Research 1990;5:371±7.
[15] Krùlner B. Seasonal variation of lumbar spine bone mineral
content in normal women. Calci®ed Tissue International
1983;35:114±8.
[16] Riggs BL, Wahner HW, Dunn WL, et al. Di�erential changes
in bone mineral density of the appendicular and axial skeleton
with ageing. Journal of Clinical Investigation 1981;67:328±35.
[17] Rosen CJ, Morrison A, Zhou H, et al. Elderly women in north-
ern New England exhibit seasonal changes in bone mineral den-
sity and calciotropic hormones. Bone and Mineral Research
1994;25:83±92.
[18] O�ce of Population Censuses and Surveys. London, ONS
Library, 1 Drummond Gate, Pimlico, London SW1V 2QQ,
]1998 http://www.statistics.gov.uk/stats/ ukin®gs/pop.htm.
[19] Harley K, Jones C. Quality of Scottish Morbidity Record
(SMR) data. Health Bulletin 1996;54:410±7.
[20] Halberg F, Johnson EA, Nelson W, et al.
Authorhythmommetry Ð procedures for physiological self
measurement and their analysis. Physiology Teacher 1972;1:1±
11.
[21] Russell D, Douglas AS, Allan TM. Changing seasonality of
birth Ð a possible environmental e�ect. Journal of
Epidemiology and Community Health 1993;47:362±7.
[22] Minors DS, Waterhouse JM. Analysis of biological time series.
In: Prendt J, Minous DS, Waterhouse JM, editors. Biological
rhythms in clinical practice. London, Sydney, Singapore,
Toronto, Wellington: Wright, 1989. p. 272±93 Chapter 14.
[23] Stamp TCB, Round JM. Seasonal changes in human plasma
levels of 25-hyrodxyvitamin D. Nature 1974;247:563±5.
[24] Bouilion RA, Auwerx JH, Lissens WD, et al. Vitamin D status
in the elderly; seasonal substrate de®ciency causes 1,25-didroxy
cholecalciferol de®ciency. American Journal of Clinical
Nutrition 1987;45:755±6.
[25] Krall EA, Sahyoun N, Tannenbaum S, et al. E�ect of vitamin
D intake on seasonal variations in parathyroid hormone se-
cretion in postmenopausal women. New England Journal of
Medicine 1989;32:1777±83.
[26] Enorring JV, Slatis P, Weber TH, et al. Serum levels of 25
hydroxyvitamin D, 24,25-dihydroxyvitamin D and parathyroid
hormone in patients with femoral neck fracture in Southern
Finland. Clinical Endocrinology 1982;17:189±94.
[27] Lips P, Hackeng WHL, Jongen MJM, et al. Seasonal variation
in serum concentration of parathyroid hormone in elderly
people. Journal of Clinical Endocrinological Metabolism
1983;57:204±6.
[28] Meller Y, Restenbaum RS, Galkinsky D, et al. Seasonal vari-
ation in serum levels of vitamin D metabolites and parathor-
mone in geriatric patients with fractures in southern Israel.
Israeli Journal of Medicine 1986;22:8±11.
[29] Price PA, Baukol BE. 1,25(OH)2D3 increases the synthesis of
the vitamin K dependent bone protein by osteosarcoma cells.
Journal of Biological Chemistry 1980;255:1660±3.
[30] Szulc P, Chapuy M-C, Meunier PJ, et al. Serum undercarboxy-
lated osteocalcin is a marker of risk of hip fracture in elderly
women. Journal of Clinical Investigations 1993;91:1769±74.
[31] Thomsen K, Erickson EF, Jorgenson JCR, et al. Seasonal vari-
ation of serum bone GLA protein. Scandinavian Journal of
Clinical and Laboratory Investigations 1989;49:605±11.
[32] Tuba J, Cantor MM, Siemens H. Phosphatases and inorganic
phosphorus in normal human serum. Journal of Laboratory
and Clinical Medicine 1947;32:194±5.
[33] Devgun MS, Paterson CR, Martin BT. Seasonal changes in the
activity of serum alkaline phosphatase. Enzyme 1981;26:301±5.
[34] Douglas AS, Miller MH, Reid DM, et al. Seasonal di�erences
in biochemical parameters in bone remodelling. Journal of
Clinical Pathology 1996;49:284±9.
[35] Arendt J. Melatonin and the mammalian (pineal) gland.
London, Glasgow, Weinheim, New York, Tokyo, Melbourne,
Madras: Chapman & Hall, 1995.
[36] Bojkowski C, Arendt J. Annual changes in 6 sulphatoxymelato-
nin excretion in man. Acta Endrocrinologica 1988;117:470±6.
[37] Laasko M-L, Parkka-Heiskanen T, Alila A, et al. Twenty four-
hour rhythms in relation to natural photoperiod: a ®eld study
in humans. Journal of Biological Rhythms 1994;9:283±94.
[38] Ross PD, Norinatsu H, Davis JW, et al. A comparison of hip
fracture incidence among native Japanese, Japanese Americans
and American Caucasians. Journal of Epidemiology
1991;133:801±9.
[39] Xu L, Lu A, Zhao X, et al. Very low rates of hip fracture in
S. Douglas et al. / Injury, Int. J. Care Injured 31 (2000) 11±1918
Beijing, People's Republic of China, the Beijing Osteoporosis
project. Journal of Epidemiology 1996;144:901±7.
[40] Cummings SR, Cauley JA, Palermo L, et al. Racial di�erences
in hip axis lengths might explain racial di�erences in rates of
hip fracture. Study of Osteoporotic Fractures Research Group
Osteoporosis International 1994;4:226±9.
[41] Beavan S, Prentice A, Dibba B, et al. Polymorphism of the
collagen type Ialpha1 gene and ethnic di�erences in hip-
fracture rates. New England Journal of Medicine
1998;30(339):351±2.
[42] Fox HJ, Pooler J, Prothero D, et al. Factors a�ecting the
outcome after proximal femoral fractures. Injury 1994;25:297±
300.
[43] Myers AH, Robinson EG, Van Natta ML, et al. Hip fractures
among the elderly: factors associated with in-hospital mortality.
American Journal of Epidemiology 1991;134:1128±37.
[44] Nather A, Seow CS, Iau P, et al. Morbidity and mortality for
elderly patients with fractured neck of femur treated by hemiar-
throplasty. Injury 1995;26:187±90.
[45] Schroder HM, Erlandsen M. Age and sex as determinants of
mortality after hip fracture: 3,895 patients followed for 2.5±18.5
years. Journal of Orthopaedic Trauma 1993;7:525±31.
S. Douglas et al. / Injury, Int. J. Care Injured 31 (2000) 11±19 19