- Email: [email protected]
Seasonal variation in high density lipoprotein cholesterol
Atherosclerosis, 100 (1993) 257-265 0 1993 Elsevier Scientific Publishers Ireland, Ltd. All rights reserved. 0021-9150/93/$06.00
257
Printed and Published in Ireland
ATHERO 05032
Seasonal variation in high density lipoprotein cholesterol Matti M2ntt%ria, Kaija Javelab, Pekka Koskinen”, Jarmo Pikkarainen”, Vesa Manninena, Jussi K. Huttunenc and M.H. Frick” “First Department
ofMedicine and bDepartment of Clinical Chemistry, Helsinki University and ‘The National Public Health Institute. Helsinki
(Finland)
(Received 6 October, 1992) (Revised, received 22 January, 1993) (Accepted 4 February, 1993)
Summary We investigated the seasonal variation in high density lipoprotein cholesterol (HDL) in 142 dyslipidemic (non-HDL-cholesterol 2 5.2 mmoY1) middle-aged men in the placebo group of the Helsinki Heart Study over the 5-year trial period. A seasonal pattern was found in HDL fluctuation, with a 4.5% drop during mid-winter (5-year mean 1.192 f 0.265 mmol/l) compared with a stable level (5-year mean 1.248 f 0.281 mmol/l) during the rest of the year (P < 0.001). A less pronounced seasonal variation in HDL was observed in 85 subjects receiving gemfibrozil. Although affecting pretrial HDL level in cross-sectional analyses, age, alcohol consumption, dietary adherence, physical activity and serum triglycerides had no influence on the seasonality of HDL variation. Smoking had a slight attenuating effect on the variation pattern. Pretrial HDL was influenced by relative weight, but there was also an inverse relationship between HDL and body weight variations, i.e. the annual drop in HDL coincided with the annual peak in body weight. However, seasonal HDL variation was not directly reflected in the annual variation in CHD incidence.
Key words: Cholesterol; HDL-cholesterol;
Seasonal variation; Gemtibrozil
Introduction Fluctuation in biological measurements is a universal phenomenon, caused by both physiological factors and laboratory variation. In humans, after elimination of the analytical variation, seasonal cycles have been detected in several biochemical components, including serum lipid and lipoprotein concentrations [l-6]. However, some studies have
failed to demonstrate seasonal cycles in serum concentration of high density lipoprotein cholesterol (HDL) [7,8]. The intra-individual coefficients of variation in HDL have ranged from 2% to 14%, depending on the population studied [5,7,10]. It has been estimated that more than 60% of this within-person variation is caused by biological factors while the rest is the result of analytical variation [7]. At present the data seem to favour a
Correspondence to: Matti Manttari, MD, First Department of Medicine, Helsinki University Central Hospital, Harrtmaninkatu 3, 00290 Helsinki, Finland. Tel.: 358-O-624988;Fax: 358-O-624998.
S!q OS ‘S~tXIa)U!
JlaaM-PI-Z1
*(.wpa~o~)
$E +I!13
aql papI.IWE
30 %U!UU$iaq aqJ UIOJ3 q,JnOJ aql pUE
JaqUIaAON
rpq
pd~q.md
IaqtuaDaa
pp
Iyun
(w2!ntJ) laqrualdag pp Igun w2hy 30 8uyuu@q aq ‘(@qy) aunt pi pm W4 30 aq~ r.uo~3pq %[email protected] uroy puo3as aq1 ‘(heruqa~) qw?~ p!r.u Igun
A.wnlqad
30 %u!uu$aq
aJaM s~!s!n )sig ayl snqL aq$o~ ~UI~JO~E asodlnd
*ue%aq
aql
uxo13 apaur
6aql ywour lsig
paureu ale SI!S!A aql ‘laded
sy,
30
aql JOT .Joa6 30 sun) au.n?s aql I’I? s6a~p
aJaM put? syaaM 9 30 aweq auy a U!~J!M aqd asaqL ~S!A yufp qDea w sas@ua aql 103
go01 ume~p
SBM a[dtues poolq v *po!lad rCpn)s aqj JaAo 61 -[enuut! saug lno3 pauyuualap alaM sIana p!dg ]o.wo~ hgvnb rclowoqvl pm suoyvugu~a~ap p@g
sa!pnls loquo9-ase9 palsau 30 slow03 wol3 papal -as aJaM Lpnls ywyed sg 103 watqns aqJ uoyyndod pauaaxx ayJ30 ~?ql 01 as013 hrah SBM s~u~dpg.nzd SHH 30 uognq!qs!p IoJalsaloq+qaH ayl aI!qM ‘ahln3 Io.ta)saIoqa .xaddn aq1 u.10~3 uaux pal=Ias p!d!l
SHH
aJaM qteap
*(paluasaJd
Jaq)o pue ~e3uauIuo.I~A
spoq$aur
aql
IOU wap)
30 suoydnasap watqns
ys!~ auqaseq
sIaAa1 row3
aq,
~tqyys
nroq
pay
papalas
OSIE SBM 10~ ‘uoy Ja3yp IOU
‘s~psy~aw.xeq3
.n?ad-S alaldmo3
SBM uoyDalag
augasaq
SHH
aqL
*slu!od-pua
jeOol tunlas aql30 uoua$!lD
ql!M
90 I) Dyap!dqsAp
le
paww
JO (hp@
2.1)
L~opue~
aqj
IV
.ssauIl! aiaM
stt~
IoJalsaloq3
aql
QuaaJ3s hreqa
aA!aDaI
laql!a
%uysruIoJduIo3
aq$ I!S!A Jaylo
ha
suoyouyurralap
OMJ $I? I/IOU.UU Z’S P
snup
pauypno
SBM
1al01) lolalsal
aIq@Ia
Aprils aql103
Igsnpu!
aArJ pue
sap pa%
alaM svxtFd!BiIFrad SHH
aJv lnq
‘[1z-61]
@no!AaJd
uaaq aAeq sllnsaJ pue u%!sap dpms 30
dnoJ8
troy
pw3
)tz Sl!S!A 3!Ui13
pue y+n %ugasuno:,
Irzolqgurakl
aJaM s)uodprFred LayI 3! pue
aAyr0asuoD
(lolalsaloq2-_?aH
uuuas .qaql3! *sa!uadwoD
alay
JoaA-s aql Ouunp (SHH)
reps aq) zp~
oqwqd
slpsaJ
u!
ay,L
l.waH
apn$s
aql
saquasap
sluedpy
Jaded
s!qL
uoitvpzdod Xpnls
pa%-alppF 82 ‘aI1
ad&
6qqaaq
luas
ay) UI~I!M [Vz‘fZ]
SHH
pua
aua$da%e
DB!PJBD p?dpuud
Jap-BaJ aqj 112panuyuoD aql
.oqa3tqd augaseq
30 aal
%yuaa_ws
-0q3 TaH-uou
aJaM sWqnc;
-ua%a )uaT.uuraAo% 0~1 dq palCoIduxa ‘ss-()t uaw 996 81 uIol3 pauaws *dva!lq paqsyqnd *poyad guyaH -.wd
30 hjqol!eAt?
*uauI
911 ad&
pua
putF sarpnls aJaM asaqL
aq] ‘uos!~vduxo~
p?uosEas
le~o$ aq$ u10.13dpua~y~u8~s UI pabt?lds!p p!dq
pasleq
8 pug q)!~
oqaDtqd
p!dg 30 uognqys!p
atuvs ?aH
SHH
‘I alqeL wi?p uo
aqt
ql!M
UossqD~JpalJ
ICay1pue uogelndod
~?eyp~e3 pua uogcw23uy p2ipJwoAw
pUl2 ~E)l?3-UON ‘SIEA.IaJU~ q)UOUI-f puoDas
01 pa)aDo[lt! JO aH3
*[()z‘6r] AIsno!Aa.Idpaysgqnd uaaq anay salqa! -JBA dn-MOllO3 put? augaseq -ua p.IoDaJ 01 pasn palyaa
~91 aql st! sapgaurrouqg
UO UCXU S$J %UOUW pa!pnlS
‘I!ZO’qIJUIa8
loquo3-asa uogopti?a -elndod p!p 4aqL
ad&
(AI
-aJdal
*acwappu! aH3
I ‘ZP
30 uoyepe~
[snuua
*[fj I ‘L I] sa!pnts
-!ppv
Iwo.was
-eyerl
paDuangu!
afi
ql!ht inq s!qi
pau!uua~ap
lauoyas-ssol3
S’IZMuog
aM lC~pi?uog
u! IaAal TaH
$leqj s1013tFJ ql!M
‘TaH
uoyy+aA
aq,
30 lanaI ~t?u.uou dIaAgaIa1
‘joJaisajoq3
TaH-uou
t? UI uoypaa
paieAa[a
?(IH
aqi
30 ixa)uoD
UI *[91-011
laded
ql!b
paia%gsaAu!
qsp
uaaq @?nsn
loJa$saloqa
ode)
IV
TaH
aqL
u!aiordodyodo
*a*! ‘iuauodwo3
%uypuodsaJroD
u!al
SJ! pue prdq aql aJOux q8nOq)ltF
OSIa aAt?q SUIJO3aA?M pawqdwo~
paisa%%ns uaaq
qioq
‘saA=M s!
au!s
leqi
Lq
aqL
B
aH3
say sp!dg
Itzuost?as aqi 30 %uoAaIaJ (1-v
JO@U
-oJdodqodE
pue (1aH)
spa3y2 uo!ie+w~
uoy
aAaq aM
%u!uaa.ws ~01~3
u! pa!pnls
uuuas u! uoyey~
'a's) suea)q
aql u! papauaJ
aql JaqlaqM
aAeq
pa$elaJlo:,
(P'L-9'0a%uel)8;:; (6Z'Z-SL'Oasu=) 08Z'I (Z'OI-I'9DEW PS'L
yndod
(6'LI)O'ZPI
(8’Z) Z’9Z (W I’8P
‘[p]
alqv=h xa%suaarad pue (a%UeJJO
OlEKWId NO
t10 S3I.LSIXKL%‘XVH33NI-I3SV8
SKEI18flS ZPI 3H.L
I 318V.L
‘[&9‘s]
paq!lDsap
uo!itryA
isaq
TaH
Alpyauraq~ew
30
apow
pwowas
852
259
visits occurred precisely at the same time of the year throughout the study period. The recruitment of HHS participants occurred in two phases, the first starting in January 1981 and the second in January 1982. For simplicity, the figures in this paper are presented as if the follow-up of the entire study population started in 1982. To eliminate the trend of increasing levels with time, the data in some figures are presented as ‘pooled’. In these figures the values are the mean levels for that visit during the whole study period, e.g. the HDL level for the May visit is the mean of all HDL determinations for May visits. A fasting sample was only required semiannually, when serum triglycerides were also determined (May and November visits). The blood was drawn using the Venoject@ collecting system, centrifuged at the end of the working day and the sera sent daily for analyses to the central biochemical laboratory of the National Public Health Institute in Helsinki. The sera collected on Fridays were kept refrigerated over the weekend and sent on the following Monday. The delay between the collection of blood samples and their analysis varied from 3 to 5 days. All HDL and other lipid determinations at the central laboratory were made by a permanent team of 3-5 laboratory technicians, supervised by a chemist. During the visit periods the whole team worked solely on the HHS lipid samples, free of other duties. The calibration procedure was conducted daily using commercially available primary standards. For the maintenance of good precision 5 control sera were analysed in each block of 24 samples. During the HHS pilot study in autumn 1980 [17] all analyses had been made in duplicate. However, as the results did not differ by more than l%, only single determinations were made during the study period. For internal quality control, two lyophilized commercially available control sera and three deep-frozen human sera pools were used. One control pool covered needs for 3-6 months. When a new pool was introduced, both the old and the new pool were analyzed in parallel for 1 week. Every day 15-20% of the samples of the previous day were reanalyzed and routine statistics used to compare the results. For the external quality assessment, WHO reference samples were analyzed 4 times a year.
Within-batch variation, determined by duplicate analyses of the control sera, ranged daily from 0.2% to 1.O% and within-day variation from 0.5% to 1.3%. The day-to-day variation of the control ranged from 0.5% to 2.1%. The within-day accuracy ranged from 0 to 2.1% and the day-to-day accuracy from 0.2% to 2.8%. The accuracy of the external quality control varied from -6% to -9%, due to the different precipitation methods used (WHO used phosphotungstate, while magnesiumdextran sulphate was used in the HHS). Statistical analyses The significances of the differences in continuous variables were estimated with the t-test or the analysis of variance (ANOVA), while for discrete variables the x2 test was applied. Pearson’s correlation coefficients were calculated to describe the association between the lifestyle variables and HDL levels at the pre-trial visit. The significances of the detected differences in mean HDL levels were compared by averaging each individual’s measurements at the time of nadir/zenith (February visit) and comparing these with the averaged values of all other visits using paired ttest. ANOVA was used when estimating the effect of lifestyle and other measured variables on HDL variation. Results
Variation in HDL cholesterol HDL levels increased with time and there was a 0.035 mmol/l(2.9%) difference in the mean annual (4 determinations) levels between the first (1.224 mmol/l) and the last (1.259 mmol/l) study year. There was a clear fall in HDL level at the February visit every year (Fig. l), with a mean decrease of 0.056 mmol/l (4.5%) over the whole study period compared with the mean of the other 3 visits (P < 0.001, paired t-test). The mean HDL-cholesterol levels of visits with fasting samples were 1.255 & 0.290 mmol/l in May and 1.247 f 0.285 mm0111in November (P = 0.11 paired t-test). The mean levels from the non-fasting samples were 1.238 f 0.278 mmoyl in August and 1.192 f 0.265 mmol/l in February, a mean difference of 0.046 mmol/l or 3.7% (P < 0.001, paired t-test). Figure 2 describes the association between
260
............. ............. .. ........ ...
_.. .......
zlJLI:z
..............
I
I
I
I
NOV
*w
Feb
May
TIME Fig. 1. Mean HDL-cholesterol levels (averaged over the 5-year trial period) in mmolil by visit in 142 study participants on placebo.
individual averaged values in February and other time points combined in the study population. Effect of modlyying factors
The effect of biological and behavioral factors on HDL variation was studied by comparing the differences in HDL between February and averaged non-February visits in various categories of these factors. In continuous variables the categorization was based on median values. At the
pretrial visit, relative weight had a significant negative association with HDL level (r = -0.17, P < 0.05). However, relative weight had no influence on the pattern of variation (Table 2). There was a 0.052 mmoyl non-significant (P = 0.3) difference in pre-trial HDL between smokers and nonsmokers, and smoking had a slight attenuating effect on variation. Sedentary subjects exhibited identical HDL variation to the physically active, while the pretrial level in the latter group was 0.067 mmol/l higher (P = 0.2). The effects of age, alcohol consumption, blood pressures and pretrial total cholesterol on HDL variation were not significant (data not shown). The metabolism of TG-rich VLDL and HDL is closely interrelated, and there tends to be a strong negative correlation between the serum levels. In this population, too, TG had a profound negative effect (r = -0.37, P < 0.001) on pretrial HDL, but a similar seasonal variation pattern was seen in subjects with baseline TG either above or below the median pretrial value of 1.53 mmol/l. Although most pretrial factors had no modifying effect on HDL variation, we investigated the simultaneous variation in biologically meaningful factors. We could not estimate the contribution of TG, since these were analysed only semiannually
,’ .:
/’ ,’ 8’
,
_I’
,‘*
0.6.
.
0,’
.’
,,’
0.8
1.0
1.2
1.4
1.6
TABLE 2
,’
P < 0.001 Y = FEB*0.99266 + 0.06422
I/ /
0.6
??
,’
4’
,’ ,’ ’ 8’
.
,’
,*‘r
t
s’ .
;’
,
4’
.
1.6
2.0
2.2
HDLIN FEBRUARY(mmol/l) Fig. 2. Linear regression analysis of the association between HDL-cholesterol levels (mmol/l) at February visit and other visits combined.
MEAN DIFFERENCES (SD.) IN HDL-CHOLESTEROL (mmoliI) BETWEEN FEBRUARY AND AVERAGED NONFEBRUARY VISITS IN CATEGORIES OF PRETRIAL RECORDED VARIABLE Pretrial variable
Difference in HDLcholesterol
Body mass, kp/m2 c25.5 -0.053 zt 225.5 -0.057 f Cigarette smoking No -0.068 f Yes -0.035 f Physical activity Sedentary -0.051 zt Active -0.063 f Serum triglycerides, mmovl < 1.53 -0.056 f z1.53 -0.054 zt
P Value
0.089 0.082
0.40
0.088 0.078
0.03
0.082 0.091
0.45
0.093 0.078
0.44
261
I!
;
1
:
1
ii hji ii ii :: j: i: i i-L
May Aug NW Feb May Aug NW Feb May Aug Nov Feb May Aug Nov Feb May
TIME IN 3 MONTHS INTERVALS Fig. 3. Mean serum total cholesterol
levels during
the 5-year trial period
and neither determination occurred during the nadir of the HDL cycle. Like HDL cholesterol, total cholesterol level increased slightly with time (Fig. 3.), with a 0.18 rnmol/l difference in mean levels between the first and last study year (6.94 mmol/l vs. 7.07 mmol/l). When data from the entire follow-up period were pooled, total cholesterol had the lowest value in August (6.93 f 0.75 mmol/l) and the highest in November (7.11 f 0.77
in 142 study participants
on placebo.
mmol/l), after which it decreased through February (6.99 f 0.73 mmol/l) and May (7.03 & 0.74 mmol/l) to the nadir in August. Mean body weight variation exhibited a significant negative correlation with mean HDL variation (Fig. 4). The differences in averaged HDL levels in May, August and November were not significantly different using variance analysis, while at the February visit there was a significantly lower level
May Aug Nov Feb May Aug Nov Feb May Aug Nov Feb May Aug NOVFeb May TIME IN 3 MONTHS INTERVALS Fig. 4. Mean values of HDL-cholesterol
and body weight during
the 5-year trial period
in 142 study participants
on placebo.
262
May Aug NOVFeb May Aug Nov Feb May AW Nov Feb May Aug Nov Feb May TIME IN 3 MONTHS
INTERVALS
Fig. 5. Mean values of HDL-cholesterol and body weight during the Syear trial period in 85 subjects on gernlibrozil.
compared with all other visits (P < 0.001). The variation in body weight was tested after removing the trend by calculating the differences between the mean level during the whole study period and the level at each visit. As illustrated in Fig. 4, the body weight exhibited a variation pattern that is probably best described by a sine wave with the nadir of -0.50 i 0.77 kg in August, and zenith in February with a deviation of +0.31 f 0.66 kg from the mean level. There was also a minor seasonal HDL variation pattern in the 85 subjects on gemfibrozil. The mean level was higher compared with subjects on placebo, and ranged from 1.335 f 0.337 mmoY1 in August to 1.378 f 0.353 mmol/l in May. The value for the February visit was 1.340 & 0.337 rnmov1. The variation patterns of HDL and body weight are given in Fig. 5. The fluctuations in HDL over the year in subjects on gemfibrozil ranged from -1% to +3% of the mean annual values. There was a drop in HDL during the last year with no obvious explanation, since there was no change in compliance as estimated by the number of returned capsules (data not presented).
in CHD incidence, the cardiac end-points in the placebo groups were arranged according to the month of the event, and quarterly rates are presented in Fig. 6. There was a steady increase in CHD events until late summer, while the numbers were lower during the last quarter of the 5 study years. Discussion Our finding from 142 subjects over 5 years is in accordance with previous studies describing seasonal variation in HDL. In this population the mean level reached a nadir during mid-winter, followed by a stable higher level during the rest of
Effect on CHD incidence
A total of 140 cardiac end-points occurred in the HHS during the 5-year study period. Nine in the placebo group were ‘silent’ infarctions detected in routine annual ECG and these were excluded from the present analyses. To study the annual variation
W GEMFIBR 0 PLACEBO
TIME IN 3 MONTHS INTERVALS
Fig. 6. Quarterly distribution of cardiac end-points in 2035 subjects on placebo and 2046 subjects on gemfibrozil during the 5year trial period in the Helsinki Heart Study.
263
the year. The finding was very consistent even at the individual level. The lipid acceptance criterion in the HHS consecutive determinations of nontwo HDL 1 5.2 mmol/l, 4 to 8 (-12) weeks apart selected subjects from the upper third of serum total cholesterol distribution. On the other hand, the HDL distribution in HHS participants was relatively normal and very close to that in the overall screened population [22]. This suggests that the discovered variation might not be restricted to a hypercholesterolemic population only. The differences in variation patterns also indicate that the mechanisms involved in the seasonal variation of serum total cholesterol and HDL have somewhat different physiological backgrounds. The seasonal variation influences both the HDL cholesterol and apolipoprotein components. However, the patterns of variation in these components are in fact quite opposite, the lowest HDL cholesterol occurring in mid-winter, when the apo AI level is highest [4]. This might indicate that the control of the composition, i.e. cholesterol content, of the HDL particles has more influence on the seasonal pattern of HDL cholesterol variation than fluctuations in the secretion of the protein components. One possible explanation for the decrease in HDL during mid-winter is freezing of the samples during transport to the laboratory. Although we cannot totally eliminate this possibility, all available precautions were taken to prevent freezing. The nurses were asked not to leave envelopes containing the samples in outdoor post boxes, but to deposit them inside post offices. The envelopes were also provided with specific warning labels and the Post Office made special ‘warm-car’ transport available. Laboratory variation is a very unlikely explanation for our findings. The extremely small variation coefficients in HDLdeterminations in the central laboratory were the result of a rigorous quality control procedure. This was achieved by a permanent staff concentrating totally on HHS lipid determinations, a mandatory feature for long term trials of lipid manipulation. One hypothesis to explain the detected correlation in the seasonal variations of body weight and HDL cholesterol is that a reduced physical activity
during the cold winter months could lead to weight gain and simultaneous reduction in the activity of lipoprotein lipase [25] with a reduced HDL level as a result. On the other hand, seasonal changes in smoking habits or alcohol consumption could also be involved, but no data of the potential changes in these behavioural variables is available. Either due to a cross-sectional study design or the fact that determinations from an individual have not always been made at the same time of year throughout the entire study period, analyses using vector algebra [4-61 have been required in previous reports. Our own study design with only four annual lipid determinations and four cycles would not have permitted the use of sophisticated statistics without gross generalisations. On the other hand, the positive amplitude deflection in the sine curve may have remained unnoticed without the use of a periodic regression analysis [4]. In fact when our data were examined using the Fourier component analysis of the SystatSygraph@ package, a sine wave with minor harmonic components was detected, giving the best description of the variation pattern (data not shown). Most studies have detected a variation mode in HDL cholesterol that is best described by sine curves [4-71, although more complicated waveforms have also been suggested [ 111. It is of interest that the largest variation has usually been discovered in subjects living in areas without major seasonal changes, e.g. in Israeli and southern USA. Another interesting finding is that the lowest HDL levels are usually found in summer, in complete contrast to our population, which seems to follow a ‘Nordic’ pattern [4] with a nadir in mid-winter. In the LRC study the seasonal zeniths of HDL and body weight were only 1 month apart while the zenith of TG occurred when the weight was lowest. The authors interpreted this to mean that seasonal variation in HDL occurs despite, rather than because of, seasonal weight changes
Fl. Seasonal variation in cardiovascular mortality and morbidity has been observed in several countries. In Finland and New Zealand [26-281 the peak in the general population occurs during midwinter. In our study population the incidence in the placebo group increased steadily towards the
264
summer months and then declined, making a direct association with HDL variation highly improbable. Even this weak trend was completely abolished by gemfibrozil. Due to the small numbers, however, no firm conclusions are justifiable. In conclusion, a seasonal variation in serum HDL cholesterol levels was detected among 142 dyslipidemic middle-aged men during 5 years of follow-up. There was a nadir during mid-winter and a relatively stable level throughout the rest of the year. Factors known to influence HDL in cross sectional studies had no major effect on this variation pattern. HDL variation exhibited a negative correlation with body weight variation, i.e. HDL was lowest when body weight peaked. The HDL variation pattern was not reflected in the quarterly incidences of coronary heart disease.
8
9
10
11
12
13
Acknowledgments 14
The experienced advice of DOS Juni Palmgren, Ph.D. on statistical analyses is highly appreciated. This study was supported (M.M.) by the Paavo Nurmi Foundation. References Paloheimo, J., Seasonal variations of serum lipids in healthy men, Ann. Med. Exper. Biol. Fenn., 39 (1961) Suppl. 8, 1. Van Gent, C.M., van der Voort, H. and Hessel, L.W., High-density lipoprotein cholesterol monthly variation and association with cardiovascular risk factors in 1000 forty-year-old Dutch citizens, Chn. Chim. Acta, 88 (1978) 155. Sasaki, J., Kumagae, G., Sata, T., Ikeda, M., Tsutsumi, S. and Arakawa, K., Seasonal variation of serum high density lipoprotein cholesterol levels in men, Atherosclerosis, 48 (1983) 167. Fager, G., Wiklund, O., Olofsson, S.-O. and Bondjers, G., Seasonal variations in serum lipid and apolipoprotein levels evaluated by periodic regression analyses, J. Chron. Dis., 35 (1982) 643. Gordon, D.J., Trost, D.C., Hyde, J. et al., Seasonal cholesterol cycles: the Lipid Research Clinics Coronary Primary Prevention Trial placebo group, Circulation, 76 (1987) 1224. Gordon, D.J., Hyde, J., Trost, D.C. et al., Cyclic seasonal variation in plasma lipid and lipoprotein levels; The Lipid Research Clinics Coronary Primary Prevention Trial placebo group, J. Clin. Epidemiol., 41 (1988) 679. Demacker, P.N.M., Schade, R.W.B., Jansen, R.T.P. and
15
16
17
18
19
20
21
22
Van’t Laar, A., Intra-individual variation of serum cholesterol, triglycerides and high density lipoprotein cholesterol in normal humans, Atherosclerosis, 45 (1982) 259. Mjos, O.D., Rao, S.N., Bjoru, L. et al., A longitudinal study of the biological variability of plasma lipoproteins in healthy young adults, Atherosclerosis, 34 (1979) 75. Robinson, D., Bevan, E.A., Hinohara, S. and Takahashi, T., Seasonal variation in serum cholesterol levels - evidence from the UK and Japan, Atherosclerosis, 95 (1992) 15 Rotterdam, E.P., Katan, M.B. and Knuiman, J.T., Importance of time interval between repeated measurements of total or high-density lipoprotein cholesterol when estimating an individual’s baseline concentrations, Chn. Chem., 33 (1987) 1913. Harlap, S., Kark, J.D., Baras, M., Eisenberg, S. and Stein, Y., Seasonal changes in plasma lipid and lipoprotein levels in Jerusalem, Isr. J. Med. Sci., 18 (1982) 1158. Hegstedt, D.M. and Nicolosi, R.J., Individual variation in serum cholesterol levels, Proc. Nat]. Acad. Sci. USA, 84 (1987) 6259. Gardner, M.J. and Heady, J.A., Some effects of withinperson variability in epidemiological studies, J. Chron. Dis., 26 (1973) 781. Pocock, S.J., Ashby, D., Shaper, A.G., Walker, M. and Broughton, B.M.G., Diurnal variation in serum biochemical and haematological measurements, J. Clin. Pathol., 42 (1989) 172. Thompson, S.G. and Pocock, S.J., The variability of serum cholesterol measurements: implications for screening and monitoring, J. Clin. Epidemiol., 43 (1990) 783. Mogadam, M., Ahmed, SW., Mensch, A.H. and Godwin, I.D., Within-person fluctuations of serum cholesterol and lipoproteins, Arch. Intern. Med., 150 (1990) 1645. Haffner, S.M., Applebaum-Bowden, D., Wahl, P.W. et al., Epidemiological correlates of high density lipoprotein subfractions, apolipoproteins A-I, A-II and D, and lecithin cholesterol acyhransferase. Effects of smoking, alcohol and adiposity, Arteriosclerosis, 5 (1985) 169. Diehl, A.K., Fuller, J.H., Mattock, M.B., Salter, A.M,, El-Gohari, R. and Keen, H., The relationship of high density lipoprotein subfractions to alcohol consumption, other lifestyle factors, and coronary heart disease, Atherosclerosis, 69 (1988) 145. Manttlri, M., Elo, O., Frick, M.H. et al., The Helsinki study: basic design and randomization procedure, Eur. Heart J., 8 (1987) Suppl. I, 1. Frick, M.H., Elo, O., Haapa, K. et al., Helsinki Heart Study: primary-prevention trial with gemfibrozil in middle-aged men with dyslipidemia. Safety of treatment, changes in risk factors, and incidence of coronary heart disease, N. Engl. J. Med., 317 (1987) 1237. Manninen, V., Elo, O., Frick, M.H. et al., Lipid alterations and decline in the incidence of coronary heart disease in the Helsinki Heart Study, J. Am. Med. Assoc., 260 (1988) 641. Manninen, V., Huttunen, J.K., Tenkanen, L., Heinonen, O.P., M&m&i, M. and Frick, M.H., High-density lipoprotein cholesterol as a risk factor for coronary heart
265
23
24
25
disease in the Helsinki Heart Study, In: Miller, N.E. (Ed.), High Density Lipoproteins and Atherosclerosis II, Excerpta Medica, ICS 826, Amsterdam, 1989, pp. 35-42. Aalto-Set&, K., Kontula, K., Mlnttlri, M. et al., DNA polymorphism of apolipoprotein B and AI/C111 genes and response to gemfibrozil treatment: a study in middle-aged hyperlipidemic men, Clin. Pharmacol. Ther., 50 (1991) 208. Manttlri, M., Koskinen, P., Ehnholm, C., Huttunen, J.K. and Manninen, V., Apolipoprotein E polymorphism influences the serum cholesterol response to dietary intervention, Metabolism, 40 (1991) 217. Kuusi, T.. Ehnholm, C., Viikari, J. et al., Postheparin
26 27
28
plasma lipoprotein and hepatic lipase are determinants of hypo and hyperalphalipoproteinemia, J. Lipid Res., 30 (1989) 1117. Sama, S., Romo, M. and Siltanen, P., Myocardial infarction and weather, Ann. Clin. Res., 9 (1977) 222. Marshall, R.J., Scragg, R. and Bourke, P., An analysis of the seasonal variation of coronary heart disease and respiratory disease mortality in New Zealand. Int. J. Epidemiol., 17 (1988) 325 Douglas, A.S., Russel, D. and Allan, T.M., Seasonal, regional and secular variations in cardiovascular and cerebrovascular mortality in New Zealand, Aust. N. Z. J. Med., 20 (1990) 669.
257
Printed and Published in Ireland
ATHERO 05032
Seasonal variation in high density lipoprotein cholesterol Matti M2ntt%ria, Kaija Javelab, Pekka Koskinen”, Jarmo Pikkarainen”, Vesa Manninena, Jussi K. Huttunenc and M.H. Frick” “First Department
ofMedicine and bDepartment of Clinical Chemistry, Helsinki University and ‘The National Public Health Institute. Helsinki
(Finland)
(Received 6 October, 1992) (Revised, received 22 January, 1993) (Accepted 4 February, 1993)
Summary We investigated the seasonal variation in high density lipoprotein cholesterol (HDL) in 142 dyslipidemic (non-HDL-cholesterol 2 5.2 mmoY1) middle-aged men in the placebo group of the Helsinki Heart Study over the 5-year trial period. A seasonal pattern was found in HDL fluctuation, with a 4.5% drop during mid-winter (5-year mean 1.192 f 0.265 mmol/l) compared with a stable level (5-year mean 1.248 f 0.281 mmol/l) during the rest of the year (P < 0.001). A less pronounced seasonal variation in HDL was observed in 85 subjects receiving gemfibrozil. Although affecting pretrial HDL level in cross-sectional analyses, age, alcohol consumption, dietary adherence, physical activity and serum triglycerides had no influence on the seasonality of HDL variation. Smoking had a slight attenuating effect on the variation pattern. Pretrial HDL was influenced by relative weight, but there was also an inverse relationship between HDL and body weight variations, i.e. the annual drop in HDL coincided with the annual peak in body weight. However, seasonal HDL variation was not directly reflected in the annual variation in CHD incidence.
Key words: Cholesterol; HDL-cholesterol;
Seasonal variation; Gemtibrozil
Introduction Fluctuation in biological measurements is a universal phenomenon, caused by both physiological factors and laboratory variation. In humans, after elimination of the analytical variation, seasonal cycles have been detected in several biochemical components, including serum lipid and lipoprotein concentrations [l-6]. However, some studies have
failed to demonstrate seasonal cycles in serum concentration of high density lipoprotein cholesterol (HDL) [7,8]. The intra-individual coefficients of variation in HDL have ranged from 2% to 14%, depending on the population studied [5,7,10]. It has been estimated that more than 60% of this within-person variation is caused by biological factors while the rest is the result of analytical variation [7]. At present the data seem to favour a
Correspondence to: Matti Manttari, MD, First Department of Medicine, Helsinki University Central Hospital, Harrtmaninkatu 3, 00290 Helsinki, Finland. Tel.: 358-O-624988;Fax: 358-O-624998.
S!q OS ‘S~tXIa)U!
JlaaM-PI-Z1
*(.wpa~o~)
$E +I!13
aql papI.IWE
30 %U!UU$iaq aqJ UIOJ3 q,JnOJ aql pUE
JaqUIaAON
rpq
pd~q.md
IaqtuaDaa
pp
Iyun
(w2!ntJ) laqrualdag pp Igun w2hy 30 8uyuu@q aq ‘(@qy) aunt pi pm W4 30 aq~ r.uo~3pq %[email protected] uroy puo3as aq1 ‘(heruqa~) qw?~ p!r.u Igun
A.wnlqad
30 %u!uu$aq
aJaM s~!s!n )sig ayl snqL aq$o~ ~UI~JO~E asodlnd
*ue%aq
aql
uxo13 apaur
6aql ywour lsig
paureu ale SI!S!A aql ‘laded
sy,
30
aql JOT .Joa6 30 sun) au.n?s aql I’I? s6a~p
aJaM put? syaaM 9 30 aweq auy a U!~J!M aqd asaqL ~S!A yufp qDea w sas@ua aql 103
go01 ume~p
SBM a[dtues poolq v *po!lad rCpn)s aqj JaAo 61 -[enuut! saug lno3 pauyuualap alaM sIana p!dg ]o.wo~ hgvnb rclowoqvl pm suoyvugu~a~ap p@g
sa!pnls loquo9-ase9 palsau 30 slow03 wol3 papal -as aJaM Lpnls ywyed sg 103 watqns aqJ uoyyndod pauaaxx ayJ30 ~?ql 01 as013 hrah SBM s~u~dpg.nzd SHH 30 uognq!qs!p IoJalsaloq+qaH ayl aI!qM ‘ahln3 Io.ta)saIoqa .xaddn aq1 u.10~3 uaux pal=Ias p!d!l
SHH
aJaM qteap
*(paluasaJd
Jaq)o pue ~e3uauIuo.I~A
spoq$aur
aql
IOU wap)
30 suoydnasap watqns
ys!~ auqaseq
sIaAa1 row3
aq,
~tqyys
nroq
pay
papalas
OSIE SBM 10~ ‘uoy Ja3yp IOU
‘s~psy~aw.xeq3
.n?ad-S alaldmo3
SBM uoyDalag
augasaq
SHH
aqL
*slu!od-pua
jeOol tunlas aql30 uoua$!lD
ql!M
90 I) Dyap!dqsAp
le
paww
JO (hp@
2.1)
L~opue~
aqj
IV
.ssauIl! aiaM
stt~
IoJalsaloq3
aql
QuaaJ3s hreqa
aA!aDaI
laql!a
%uysruIoJduIo3
aq$ I!S!A Jaylo
ha
suoyouyurralap
OMJ $I? I/IOU.UU Z’S P
snup
pauypno
SBM
1al01) lolalsal
aIq@Ia
Aprils aql103
Igsnpu!
aArJ pue
sap pa%
alaM svxtFd!BiIFrad SHH
aJv lnq
‘[1z-61]
@no!AaJd
uaaq aAeq sllnsaJ pue u%!sap dpms 30
dnoJ8
troy
pw3
)tz Sl!S!A 3!Ui13
pue y+n %ugasuno:,
Irzolqgurakl
aJaM s)uodprFred LayI 3! pue
aAyr0asuoD
(lolalsaloq2-_?aH
uuuas .qaql3! *sa!uadwoD
alay
JoaA-s aql Ouunp (SHH)
reps aq) zp~
oqwqd
slpsaJ
u!
ay,L
l.waH
apn$s
aql
saquasap
sluedpy
Jaded
s!qL
uoitvpzdod Xpnls
pa%-alppF 82 ‘aI1
ad&
6qqaaq
luas
ay) UI~I!M [Vz‘fZ]
SHH
pua
aua$da%e
DB!PJBD p?dpuud
Jap-BaJ aqj 112panuyuoD aql
.oqa3tqd augaseq
30 aal
%yuaa_ws
-0q3 TaH-uou
aJaM sWqnc;
-ua%a )uaT.uuraAo% 0~1 dq palCoIduxa ‘ss-()t uaw 996 81 uIol3 pauaws *dva!lq paqsyqnd *poyad guyaH -.wd
30 hjqol!eAt?
*uauI
911 ad&
pua
putF sarpnls aJaM asaqL
aq] ‘uos!~vduxo~
p?uosEas
le~o$ aq$ u10.13dpua~y~u8~s UI pabt?lds!p p!dq
pasleq
8 pug q)!~
oqaDtqd
p!dg 30 uognqys!p
atuvs ?aH
SHH
‘I alqeL wi?p uo
aqt
ql!M
UossqD~JpalJ
ICay1pue uogelndod
~?eyp~e3 pua uogcw23uy p2ipJwoAw
pUl2 ~E)l?3-UON ‘SIEA.IaJU~ q)UOUI-f puoDas
01 pa)aDo[lt! JO aH3
*[()z‘6r] AIsno!Aa.Idpaysgqnd uaaq anay salqa! -JBA dn-MOllO3 put? augaseq -ua p.IoDaJ 01 pasn palyaa
~91 aql st! sapgaurrouqg
UO UCXU S$J %UOUW pa!pnlS
‘I!ZO’qIJUIa8
loquo3-asa uogopti?a -elndod p!p 4aqL
ad&
(AI
-aJdal
*acwappu! aH3
I ‘ZP
30 uoyepe~
[snuua
*[fj I ‘L I] sa!pnts
-!ppv
Iwo.was
-eyerl
paDuangu!
afi
ql!ht inq s!qi
pau!uua~ap
lauoyas-ssol3
S’IZMuog
aM lC~pi?uog
u! IaAal TaH
$leqj s1013tFJ ql!M
‘TaH
uoyy+aA
aq,
30 lanaI ~t?u.uou dIaAgaIa1
‘joJaisajoq3
TaH-uou
t? UI uoypaa
paieAa[a
?(IH
aqi
30 ixa)uoD
UI *[91-011
laded
ql!b
paia%gsaAu!
qsp
uaaq @?nsn
loJa$saloqa
ode)
IV
TaH
aqL
u!aiordodyodo
*a*! ‘iuauodwo3
%uypuodsaJroD
u!al
SJ! pue prdq aql aJOux q8nOq)ltF
OSIa aAt?q SUIJO3aA?M pawqdwo~
paisa%%ns uaaq
qioq
‘saA=M s!
au!s
leqi
Lq
aqL
B
aH3
say sp!dg
Itzuost?as aqi 30 %uoAaIaJ (1-v
JO@U
-oJdodqodE
pue (1aH)
spa3y2 uo!ie+w~
uoy
aAaq aM
%u!uaa.ws ~01~3
u! pa!pnls
uuuas u! uoyey~
'a's) suea)q
aql u! papauaJ
aql JaqlaqM
aAeq
pa$elaJlo:,
(P'L-9'0a%uel)8;:; (6Z'Z-SL'Oasu=) 08Z'I (Z'OI-I'9DEW PS'L
yndod
(6'LI)O'ZPI
(8’Z) Z’9Z (W I’8P
‘[p]
alqv=h xa%suaarad pue (a%UeJJO
OlEKWId NO
t10 S3I.LSIXKL%‘XVH33NI-I3SV8
SKEI18flS ZPI 3H.L
I 318V.L
‘[&9‘s]
paq!lDsap
uo!itryA
isaq
TaH
Alpyauraq~ew
30
apow
pwowas
852
259
visits occurred precisely at the same time of the year throughout the study period. The recruitment of HHS participants occurred in two phases, the first starting in January 1981 and the second in January 1982. For simplicity, the figures in this paper are presented as if the follow-up of the entire study population started in 1982. To eliminate the trend of increasing levels with time, the data in some figures are presented as ‘pooled’. In these figures the values are the mean levels for that visit during the whole study period, e.g. the HDL level for the May visit is the mean of all HDL determinations for May visits. A fasting sample was only required semiannually, when serum triglycerides were also determined (May and November visits). The blood was drawn using the Venoject@ collecting system, centrifuged at the end of the working day and the sera sent daily for analyses to the central biochemical laboratory of the National Public Health Institute in Helsinki. The sera collected on Fridays were kept refrigerated over the weekend and sent on the following Monday. The delay between the collection of blood samples and their analysis varied from 3 to 5 days. All HDL and other lipid determinations at the central laboratory were made by a permanent team of 3-5 laboratory technicians, supervised by a chemist. During the visit periods the whole team worked solely on the HHS lipid samples, free of other duties. The calibration procedure was conducted daily using commercially available primary standards. For the maintenance of good precision 5 control sera were analysed in each block of 24 samples. During the HHS pilot study in autumn 1980 [17] all analyses had been made in duplicate. However, as the results did not differ by more than l%, only single determinations were made during the study period. For internal quality control, two lyophilized commercially available control sera and three deep-frozen human sera pools were used. One control pool covered needs for 3-6 months. When a new pool was introduced, both the old and the new pool were analyzed in parallel for 1 week. Every day 15-20% of the samples of the previous day were reanalyzed and routine statistics used to compare the results. For the external quality assessment, WHO reference samples were analyzed 4 times a year.
Within-batch variation, determined by duplicate analyses of the control sera, ranged daily from 0.2% to 1.O% and within-day variation from 0.5% to 1.3%. The day-to-day variation of the control ranged from 0.5% to 2.1%. The within-day accuracy ranged from 0 to 2.1% and the day-to-day accuracy from 0.2% to 2.8%. The accuracy of the external quality control varied from -6% to -9%, due to the different precipitation methods used (WHO used phosphotungstate, while magnesiumdextran sulphate was used in the HHS). Statistical analyses The significances of the differences in continuous variables were estimated with the t-test or the analysis of variance (ANOVA), while for discrete variables the x2 test was applied. Pearson’s correlation coefficients were calculated to describe the association between the lifestyle variables and HDL levels at the pre-trial visit. The significances of the detected differences in mean HDL levels were compared by averaging each individual’s measurements at the time of nadir/zenith (February visit) and comparing these with the averaged values of all other visits using paired ttest. ANOVA was used when estimating the effect of lifestyle and other measured variables on HDL variation. Results
Variation in HDL cholesterol HDL levels increased with time and there was a 0.035 mmol/l(2.9%) difference in the mean annual (4 determinations) levels between the first (1.224 mmol/l) and the last (1.259 mmol/l) study year. There was a clear fall in HDL level at the February visit every year (Fig. l), with a mean decrease of 0.056 mmol/l (4.5%) over the whole study period compared with the mean of the other 3 visits (P < 0.001, paired t-test). The mean HDL-cholesterol levels of visits with fasting samples were 1.255 & 0.290 mmol/l in May and 1.247 f 0.285 mm0111in November (P = 0.11 paired t-test). The mean levels from the non-fasting samples were 1.238 f 0.278 mmoyl in August and 1.192 f 0.265 mmol/l in February, a mean difference of 0.046 mmol/l or 3.7% (P < 0.001, paired t-test). Figure 2 describes the association between
260
............. ............. .. ........ ...
_.. .......
zlJLI:z
..............
I
I
I
I
NOV
*w
Feb
May
TIME Fig. 1. Mean HDL-cholesterol levels (averaged over the 5-year trial period) in mmolil by visit in 142 study participants on placebo.
individual averaged values in February and other time points combined in the study population. Effect of modlyying factors
The effect of biological and behavioral factors on HDL variation was studied by comparing the differences in HDL between February and averaged non-February visits in various categories of these factors. In continuous variables the categorization was based on median values. At the
pretrial visit, relative weight had a significant negative association with HDL level (r = -0.17, P < 0.05). However, relative weight had no influence on the pattern of variation (Table 2). There was a 0.052 mmoyl non-significant (P = 0.3) difference in pre-trial HDL between smokers and nonsmokers, and smoking had a slight attenuating effect on variation. Sedentary subjects exhibited identical HDL variation to the physically active, while the pretrial level in the latter group was 0.067 mmol/l higher (P = 0.2). The effects of age, alcohol consumption, blood pressures and pretrial total cholesterol on HDL variation were not significant (data not shown). The metabolism of TG-rich VLDL and HDL is closely interrelated, and there tends to be a strong negative correlation between the serum levels. In this population, too, TG had a profound negative effect (r = -0.37, P < 0.001) on pretrial HDL, but a similar seasonal variation pattern was seen in subjects with baseline TG either above or below the median pretrial value of 1.53 mmol/l. Although most pretrial factors had no modifying effect on HDL variation, we investigated the simultaneous variation in biologically meaningful factors. We could not estimate the contribution of TG, since these were analysed only semiannually
,’ .:
/’ ,’ 8’
,
_I’
,‘*
0.6.
.
0,’
.’
,,’
0.8
1.0
1.2
1.4
1.6
TABLE 2
,’
P < 0.001 Y = FEB*0.99266 + 0.06422
I/ /
0.6
??
,’
4’
,’ ,’ ’ 8’
.
,’
,*‘r
t
s’ .
;’
,
4’
.
1.6
2.0
2.2
HDLIN FEBRUARY(mmol/l) Fig. 2. Linear regression analysis of the association between HDL-cholesterol levels (mmol/l) at February visit and other visits combined.
MEAN DIFFERENCES (SD.) IN HDL-CHOLESTEROL (mmoliI) BETWEEN FEBRUARY AND AVERAGED NONFEBRUARY VISITS IN CATEGORIES OF PRETRIAL RECORDED VARIABLE Pretrial variable
Difference in HDLcholesterol
Body mass, kp/m2 c25.5 -0.053 zt 225.5 -0.057 f Cigarette smoking No -0.068 f Yes -0.035 f Physical activity Sedentary -0.051 zt Active -0.063 f Serum triglycerides, mmovl < 1.53 -0.056 f z1.53 -0.054 zt
P Value
0.089 0.082
0.40
0.088 0.078
0.03
0.082 0.091
0.45
0.093 0.078
0.44
261
I!
;
1
:
1
ii hji ii ii :: j: i: i i-L
May Aug NW Feb May Aug NW Feb May Aug Nov Feb May Aug Nov Feb May
TIME IN 3 MONTHS INTERVALS Fig. 3. Mean serum total cholesterol
levels during
the 5-year trial period
and neither determination occurred during the nadir of the HDL cycle. Like HDL cholesterol, total cholesterol level increased slightly with time (Fig. 3.), with a 0.18 rnmol/l difference in mean levels between the first and last study year (6.94 mmol/l vs. 7.07 mmol/l). When data from the entire follow-up period were pooled, total cholesterol had the lowest value in August (6.93 f 0.75 mmol/l) and the highest in November (7.11 f 0.77
in 142 study participants
on placebo.
mmol/l), after which it decreased through February (6.99 f 0.73 mmol/l) and May (7.03 & 0.74 mmol/l) to the nadir in August. Mean body weight variation exhibited a significant negative correlation with mean HDL variation (Fig. 4). The differences in averaged HDL levels in May, August and November were not significantly different using variance analysis, while at the February visit there was a significantly lower level
May Aug Nov Feb May Aug Nov Feb May Aug Nov Feb May Aug NOVFeb May TIME IN 3 MONTHS INTERVALS Fig. 4. Mean values of HDL-cholesterol
and body weight during
the 5-year trial period
in 142 study participants
on placebo.
262
May Aug NOVFeb May Aug Nov Feb May AW Nov Feb May Aug Nov Feb May TIME IN 3 MONTHS
INTERVALS
Fig. 5. Mean values of HDL-cholesterol and body weight during the Syear trial period in 85 subjects on gernlibrozil.
compared with all other visits (P < 0.001). The variation in body weight was tested after removing the trend by calculating the differences between the mean level during the whole study period and the level at each visit. As illustrated in Fig. 4, the body weight exhibited a variation pattern that is probably best described by a sine wave with the nadir of -0.50 i 0.77 kg in August, and zenith in February with a deviation of +0.31 f 0.66 kg from the mean level. There was also a minor seasonal HDL variation pattern in the 85 subjects on gemfibrozil. The mean level was higher compared with subjects on placebo, and ranged from 1.335 f 0.337 mmoY1 in August to 1.378 f 0.353 mmol/l in May. The value for the February visit was 1.340 & 0.337 rnmov1. The variation patterns of HDL and body weight are given in Fig. 5. The fluctuations in HDL over the year in subjects on gemfibrozil ranged from -1% to +3% of the mean annual values. There was a drop in HDL during the last year with no obvious explanation, since there was no change in compliance as estimated by the number of returned capsules (data not presented).
in CHD incidence, the cardiac end-points in the placebo groups were arranged according to the month of the event, and quarterly rates are presented in Fig. 6. There was a steady increase in CHD events until late summer, while the numbers were lower during the last quarter of the 5 study years. Discussion Our finding from 142 subjects over 5 years is in accordance with previous studies describing seasonal variation in HDL. In this population the mean level reached a nadir during mid-winter, followed by a stable higher level during the rest of
Effect on CHD incidence
A total of 140 cardiac end-points occurred in the HHS during the 5-year study period. Nine in the placebo group were ‘silent’ infarctions detected in routine annual ECG and these were excluded from the present analyses. To study the annual variation
W GEMFIBR 0 PLACEBO
TIME IN 3 MONTHS INTERVALS
Fig. 6. Quarterly distribution of cardiac end-points in 2035 subjects on placebo and 2046 subjects on gemfibrozil during the 5year trial period in the Helsinki Heart Study.
263
the year. The finding was very consistent even at the individual level. The lipid acceptance criterion in the HHS consecutive determinations of nontwo HDL 1 5.2 mmol/l, 4 to 8 (-12) weeks apart selected subjects from the upper third of serum total cholesterol distribution. On the other hand, the HDL distribution in HHS participants was relatively normal and very close to that in the overall screened population [22]. This suggests that the discovered variation might not be restricted to a hypercholesterolemic population only. The differences in variation patterns also indicate that the mechanisms involved in the seasonal variation of serum total cholesterol and HDL have somewhat different physiological backgrounds. The seasonal variation influences both the HDL cholesterol and apolipoprotein components. However, the patterns of variation in these components are in fact quite opposite, the lowest HDL cholesterol occurring in mid-winter, when the apo AI level is highest [4]. This might indicate that the control of the composition, i.e. cholesterol content, of the HDL particles has more influence on the seasonal pattern of HDL cholesterol variation than fluctuations in the secretion of the protein components. One possible explanation for the decrease in HDL during mid-winter is freezing of the samples during transport to the laboratory. Although we cannot totally eliminate this possibility, all available precautions were taken to prevent freezing. The nurses were asked not to leave envelopes containing the samples in outdoor post boxes, but to deposit them inside post offices. The envelopes were also provided with specific warning labels and the Post Office made special ‘warm-car’ transport available. Laboratory variation is a very unlikely explanation for our findings. The extremely small variation coefficients in HDLdeterminations in the central laboratory were the result of a rigorous quality control procedure. This was achieved by a permanent staff concentrating totally on HHS lipid determinations, a mandatory feature for long term trials of lipid manipulation. One hypothesis to explain the detected correlation in the seasonal variations of body weight and HDL cholesterol is that a reduced physical activity
during the cold winter months could lead to weight gain and simultaneous reduction in the activity of lipoprotein lipase [25] with a reduced HDL level as a result. On the other hand, seasonal changes in smoking habits or alcohol consumption could also be involved, but no data of the potential changes in these behavioural variables is available. Either due to a cross-sectional study design or the fact that determinations from an individual have not always been made at the same time of year throughout the entire study period, analyses using vector algebra [4-61 have been required in previous reports. Our own study design with only four annual lipid determinations and four cycles would not have permitted the use of sophisticated statistics without gross generalisations. On the other hand, the positive amplitude deflection in the sine curve may have remained unnoticed without the use of a periodic regression analysis [4]. In fact when our data were examined using the Fourier component analysis of the SystatSygraph@ package, a sine wave with minor harmonic components was detected, giving the best description of the variation pattern (data not shown). Most studies have detected a variation mode in HDL cholesterol that is best described by sine curves [4-71, although more complicated waveforms have also been suggested [ 111. It is of interest that the largest variation has usually been discovered in subjects living in areas without major seasonal changes, e.g. in Israeli and southern USA. Another interesting finding is that the lowest HDL levels are usually found in summer, in complete contrast to our population, which seems to follow a ‘Nordic’ pattern [4] with a nadir in mid-winter. In the LRC study the seasonal zeniths of HDL and body weight were only 1 month apart while the zenith of TG occurred when the weight was lowest. The authors interpreted this to mean that seasonal variation in HDL occurs despite, rather than because of, seasonal weight changes
Fl. Seasonal variation in cardiovascular mortality and morbidity has been observed in several countries. In Finland and New Zealand [26-281 the peak in the general population occurs during midwinter. In our study population the incidence in the placebo group increased steadily towards the
264
summer months and then declined, making a direct association with HDL variation highly improbable. Even this weak trend was completely abolished by gemfibrozil. Due to the small numbers, however, no firm conclusions are justifiable. In conclusion, a seasonal variation in serum HDL cholesterol levels was detected among 142 dyslipidemic middle-aged men during 5 years of follow-up. There was a nadir during mid-winter and a relatively stable level throughout the rest of the year. Factors known to influence HDL in cross sectional studies had no major effect on this variation pattern. HDL variation exhibited a negative correlation with body weight variation, i.e. HDL was lowest when body weight peaked. The HDL variation pattern was not reflected in the quarterly incidences of coronary heart disease.
8
9
10
11
12
13
Acknowledgments 14
The experienced advice of DOS Juni Palmgren, Ph.D. on statistical analyses is highly appreciated. This study was supported (M.M.) by the Paavo Nurmi Foundation. References Paloheimo, J., Seasonal variations of serum lipids in healthy men, Ann. Med. Exper. Biol. Fenn., 39 (1961) Suppl. 8, 1. Van Gent, C.M., van der Voort, H. and Hessel, L.W., High-density lipoprotein cholesterol monthly variation and association with cardiovascular risk factors in 1000 forty-year-old Dutch citizens, Chn. Chim. Acta, 88 (1978) 155. Sasaki, J., Kumagae, G., Sata, T., Ikeda, M., Tsutsumi, S. and Arakawa, K., Seasonal variation of serum high density lipoprotein cholesterol levels in men, Atherosclerosis, 48 (1983) 167. Fager, G., Wiklund, O., Olofsson, S.-O. and Bondjers, G., Seasonal variations in serum lipid and apolipoprotein levels evaluated by periodic regression analyses, J. Chron. Dis., 35 (1982) 643. Gordon, D.J., Trost, D.C., Hyde, J. et al., Seasonal cholesterol cycles: the Lipid Research Clinics Coronary Primary Prevention Trial placebo group, Circulation, 76 (1987) 1224. Gordon, D.J., Hyde, J., Trost, D.C. et al., Cyclic seasonal variation in plasma lipid and lipoprotein levels; The Lipid Research Clinics Coronary Primary Prevention Trial placebo group, J. Clin. Epidemiol., 41 (1988) 679. Demacker, P.N.M., Schade, R.W.B., Jansen, R.T.P. and
15
16
17
18
19
20
21
22
Van’t Laar, A., Intra-individual variation of serum cholesterol, triglycerides and high density lipoprotein cholesterol in normal humans, Atherosclerosis, 45 (1982) 259. Mjos, O.D., Rao, S.N., Bjoru, L. et al., A longitudinal study of the biological variability of plasma lipoproteins in healthy young adults, Atherosclerosis, 34 (1979) 75. Robinson, D., Bevan, E.A., Hinohara, S. and Takahashi, T., Seasonal variation in serum cholesterol levels - evidence from the UK and Japan, Atherosclerosis, 95 (1992) 15 Rotterdam, E.P., Katan, M.B. and Knuiman, J.T., Importance of time interval between repeated measurements of total or high-density lipoprotein cholesterol when estimating an individual’s baseline concentrations, Chn. Chem., 33 (1987) 1913. Harlap, S., Kark, J.D., Baras, M., Eisenberg, S. and Stein, Y., Seasonal changes in plasma lipid and lipoprotein levels in Jerusalem, Isr. J. Med. Sci., 18 (1982) 1158. Hegstedt, D.M. and Nicolosi, R.J., Individual variation in serum cholesterol levels, Proc. Nat]. Acad. Sci. USA, 84 (1987) 6259. Gardner, M.J. and Heady, J.A., Some effects of withinperson variability in epidemiological studies, J. Chron. Dis., 26 (1973) 781. Pocock, S.J., Ashby, D., Shaper, A.G., Walker, M. and Broughton, B.M.G., Diurnal variation in serum biochemical and haematological measurements, J. Clin. Pathol., 42 (1989) 172. Thompson, S.G. and Pocock, S.J., The variability of serum cholesterol measurements: implications for screening and monitoring, J. Clin. Epidemiol., 43 (1990) 783. Mogadam, M., Ahmed, SW., Mensch, A.H. and Godwin, I.D., Within-person fluctuations of serum cholesterol and lipoproteins, Arch. Intern. Med., 150 (1990) 1645. Haffner, S.M., Applebaum-Bowden, D., Wahl, P.W. et al., Epidemiological correlates of high density lipoprotein subfractions, apolipoproteins A-I, A-II and D, and lecithin cholesterol acyhransferase. Effects of smoking, alcohol and adiposity, Arteriosclerosis, 5 (1985) 169. Diehl, A.K., Fuller, J.H., Mattock, M.B., Salter, A.M,, El-Gohari, R. and Keen, H., The relationship of high density lipoprotein subfractions to alcohol consumption, other lifestyle factors, and coronary heart disease, Atherosclerosis, 69 (1988) 145. Manttlri, M., Elo, O., Frick, M.H. et al., The Helsinki study: basic design and randomization procedure, Eur. Heart J., 8 (1987) Suppl. I, 1. Frick, M.H., Elo, O., Haapa, K. et al., Helsinki Heart Study: primary-prevention trial with gemfibrozil in middle-aged men with dyslipidemia. Safety of treatment, changes in risk factors, and incidence of coronary heart disease, N. Engl. J. Med., 317 (1987) 1237. Manninen, V., Elo, O., Frick, M.H. et al., Lipid alterations and decline in the incidence of coronary heart disease in the Helsinki Heart Study, J. Am. Med. Assoc., 260 (1988) 641. Manninen, V., Huttunen, J.K., Tenkanen, L., Heinonen, O.P., M&m&i, M. and Frick, M.H., High-density lipoprotein cholesterol as a risk factor for coronary heart
265
23
24
25
disease in the Helsinki Heart Study, In: Miller, N.E. (Ed.), High Density Lipoproteins and Atherosclerosis II, Excerpta Medica, ICS 826, Amsterdam, 1989, pp. 35-42. Aalto-Set&, K., Kontula, K., Mlnttlri, M. et al., DNA polymorphism of apolipoprotein B and AI/C111 genes and response to gemfibrozil treatment: a study in middle-aged hyperlipidemic men, Clin. Pharmacol. Ther., 50 (1991) 208. Manttlri, M., Koskinen, P., Ehnholm, C., Huttunen, J.K. and Manninen, V., Apolipoprotein E polymorphism influences the serum cholesterol response to dietary intervention, Metabolism, 40 (1991) 217. Kuusi, T.. Ehnholm, C., Viikari, J. et al., Postheparin
26 27
28
plasma lipoprotein and hepatic lipase are determinants of hypo and hyperalphalipoproteinemia, J. Lipid Res., 30 (1989) 1117. Sama, S., Romo, M. and Siltanen, P., Myocardial infarction and weather, Ann. Clin. Res., 9 (1977) 222. Marshall, R.J., Scragg, R. and Bourke, P., An analysis of the seasonal variation of coronary heart disease and respiratory disease mortality in New Zealand. Int. J. Epidemiol., 17 (1988) 325 Douglas, A.S., Russel, D. and Allan, T.M., Seasonal, regional and secular variations in cardiovascular and cerebrovascular mortality in New Zealand, Aust. N. Z. J. Med., 20 (1990) 669.