ORIGINAL ARTICLE
Serum Vitamin D and Subsequent Occurrence of
Type 2 Diabetes
Paul Knekt, Maarit Laaksonen, Catharina Mattila, Tommi Härkänen, Jukka Marniemi,
Markku Heliövaara, Harri Rissanen, Jukka Montonen, and Antti Reunanen
Background: Low vitamin D status has been suggested as a risk
factor for type 2 diabetes. Although the epidemiologic evidence is
scarce, 2 recent studies have suggested an association. The present
study investigated the relation of serum vitamin D with type 2
diabetes incidence using pooled data from these 2 cohorts.
Methods: Two nested case-control studies, collected by the Finnish
Mobile Clinic in 1973–1980, were pooled for analysis. The study
populations consisted of men and women aged 40 –74 years and free
of diabetes at baseline. During a follow-up period of 22 years, 412
incident type 2 diabetes cases occurred, and 986 controls were
selected by individual matching. Serum vitamin D (serum 25(OH)D)
was determined from frozen samples, stored at baseline. Pooled
estimates of the relationship between serum vitamin D concentration
and type 2 diabetes incidence were calculated.
Results: Men had higher serum vitamin D concentrations than
women and showed a reduced risk of type 2 diabetes in their highest
vitamin D quartile. The relative odds between the highest and lowest
quartiles was 0.28 (95% confidence interval ⫽ 0.10 – 0.81) in men
and 1.14 (0.60 –2.17) in women after adjustment for smoking, body
mass index, physical activity, and education.
Conclusions: The results support the hypothesis that high vitamin D
status provides protection against type 2 diabetes. Residual con-
founding may contribute to this association.
(Epidemiology 2008;19: 666 – 671)
V itamin D plays an important role not only in the patho-
genesis of skeletal disorders and calcium homeostasis,
but also in the development of several chronic conditions.1,2
It has been suggested that vitamin D deficiency is a risk factor
for type 1 diabetes mellitus.3,4 Vitamin D deficiency impairs
insulin secretion of pancreatic ␤-cells4 – 6 and increases insu-
lin resistance,7 which are major factors in the pathogenesis of
type 2 diabetes. Accordingly, low vitamin D might also be a
risk factor for type 2 diabetes.8
Submitted 1 June 2007; accepted 22 January 2008; posted 20 May 2008.
From the National Public Health Institute, Department of Health and Func-
tional Capacity, Mannerheimintie 166, 00300 Helsinki, Finland.
Correspondence: Paul Knekt, National Public Health Institute, Mannerhei-
mintie 166, 00300 Helsinki, Finland. E-mail: paul.knekt@ktl.fi.
Copyright © 2008 by Lippincott Williams & Wilkins
ISSN: 1044-3983/08/1905-0666
DOI: 10.1097/EDE.0b013e318176b8ad
666
The epidemiologic evidence of an association between
vitamin D deficiency and type 2 diabetes is mainly based on
cross-sectional studies.9 –14 The only prospective study deal-
ing the effect of vitamin D deficiency on the incidence of type
2 diabetes was based on the intake of vitamin D.15 Vitamin D
intake is a weak proxy measure of vitamin D status; tissue
vitamin D gives a more reliable picture.16 The incidence of
type 2 diabetes in relation to serum vitamin D levels has been
investigated in 2 small prospective population studies.17,18
Both studies showed an inverse association between serum
vitamin D status and occurrence of type 2 diabetes. In the
present study, we further investigated the link between serum
vitamin D and type 2 diabetes by pooling the primary data
from these 2 Finnish cohorts.
METHODS
Subjects and Methods
The present study is based on 2 cohorts, the Finnish
Mobile Clinic Health Examination Survey,19 carried out in
1973–1976, and the Mini-Finland Health Survey,20 carried
out in 1978 –1980. The earlier data consisted of 19,518 men
and women, aged 20 years and older, from 12 municipalities
in different parts of Finland. The later consisted of 8000
persons from 40 geographical areas and was a representative
sample of the Finnish population of adults aged 30 years and
over. Among those free of type 2 diabetes at baseline, there
were 3327 persons 40 – 69 years of age (in the earlier study)
and 4176 persons 40 –74 years of age (in the latter study)
(Table 1).
Data on education, smoking, leisure time physical ac-
tivity, previous diseases, and previous and present medica-
tion, (eg, for hypertension), were self-reported on a question-
naire or in a health interview.19,21 Subjects classified their
leisure-time physical activity as (1) none or little; (2) walk-
ing, cycling, or related light activities at least 4 hours per
week; (3) ball games, jogging, or related activities at least 3
hours per week; or (4) regular vigorous exercise. Height and
weight were measured, and body mass index (BMI) was
calculated (kg/m2). Casual blood pressure was measured with
the auscultatory method, and blood samples were collected
and stored at ⫺20°C. Serum total cholesterol, triglycerides,
insulin, and plasma fasting glucose were measured. The
Epidemiology • Volume 19, Number 5, September 2008
Epidemiology • Volume 19, Number 5, September 2008 Serum Vitamin D and Type 2 Diabetes

TABLE 1. Basic Characteristics, by Sex, of the Nested Case-Control Studies Included in the Pooled Analysis of Serum Vitamin
D and Type 2 Diabetes Risk
Mean
Concentrationa Mean Concentrationa (Range) in Quartiles of Serum
(Range) Vitamin D
Study Follow-up Size of Age No. No. Serum Vitamin
Population Period Population (yrs) Cases Controls D (nmol/L) 1 (Lowest) 2 3 4 (Highest)

Pooled 7503 412 986 42.55 (9–148) 22.30 (9–29) 34.64 (30–39) 45.85 (40–53) 69.11 (54–148)
Finnish Mobile 1973–1994
Clinic Health
Examination
Survey
Men 1628 40–74 105 206 46.01 (11–148) 23.52 (11–32) 37.04 (33–41) 48.95 (42–57) 74.54 (58–148)
Women 1699 40–74 125 246 39.30 (11–109) 22.56 (11–28) 32.53 (29–36) 42.61 (37–48) 62.51 (49–109)
Mini-Finland 1978–1994
Health Survey
Men 1948 40–69 83 245 47.33 (11–117) 23.91 (11–32) 38.47 (33–44) 53.78 (45–62) 75.60 (63–117)
Women 2228 40–69 99 289 38.80 (9–105) 20.40 (9–26) 31.39 (27–36) 42.14 (37–48) 62.40 (49–105)
a
Measured from controls.

serum samples were kept frozen in the same storage at
⫺20°C until 2002 (for the earlier study) and 2003 (for the
later study), when the serum vitamin D concentrations (serum
25(OH)D) were determined using the radioimmunoassay
(RIA, DiaSorin, MI). The interassay coefficient of variation
of the 25-OH-vitamin D measurements was 7.8% at the mean
level of 47.3 nmol/L (n ⫽ 167). The serum vitamin D levels
were higher in men than in women and associated with month
of serum measurement. The sex and age-adjusted correlation
coefficient for the month and vitamin D association was 0.44
in controls.
Diabetes cases at baseline were identified by information
given by the participants and by fasting glucose values.19,20 All
previously known cases or persons newly diagnosed with dia-
betes at baseline were excluded from the analyses. Diabetes
cases occurring during the follow-up were identified based on
the registry of reimbursement for costs of diabetes medication.
According to the Finnish sickness insurance legislation, diabetes
patients needing drug treatment are allowed certain drugs free of
charge. To get this drug allowance, a certificate must be obtained
from the physician in charge, describing the diagnostic criteria
applied when the diabetes was diagnosed. The certificate is
accepted after confirmation by special advisers at the Social
Insurance Institution,22 which maintains a central register of all
patients receiving drug reimbursement. Participants in the
present study populations were linked to this register with the
unique social-security code assigned to each Finnish citizen.
Follow-up time was defined as the number of days from
the baseline examination to the dates of type 2 diabetes
occurrence, death, or withdrawal (ie, end of follow-up),
whichever came first. The follow-up varied from 17 years for
the later cohort to 22 years for the earlier cohort (Table 1).
During the follow-up period, the number of incident type 2
diabetes cases identified from a nationwide registry was 230
for the earlier cohort and 182 for the later cohort (Table 1).
All medical certificates of these cases were checked, and
every case met the World Health Organization diagnostic
criteria for type 2 diabetes mellitus.23
A nested case-control design was adopted for both sets
of data. Two controls per case were selected from the earlier
cohort and 3 per case from the later cohort by individual
matching for sex, age and municipality. Controls were drawn
from the same municipality as the type 2 diabetes cases, and
age was matched using nearest available matching. The
differences in age between cases and controls varied from 0
to 2 years for 95% of the patients and from 3 to 5 in 5%.
Length of follow-up was controlled for. The group at risk
from which controls were selected comprised all persons who
were free of type 2 diabetes until the date of diabetes
diagnosis of the case. Matching for municipality also con-
trolled for the month of baseline examination (ie, for possible
seasonal variation in exposure to vitamin D coming from the
sunlight) and for duration of storage of serum samples.24
Statistical Methods
The conditional logistic model25 was used to assess the
association between vitamin D intake and type 2 diabetes risk
in the single subcohorts. To avoid assumptions of the shape
of the relationship between vitamin D exposure and type 2
diabetes incidence in the statistical analyses, relative odds
(odds ratios 关ORs兴) were estimated for quartiles of vitamin D.
Two-sided 95% confidence intervals (CIs) were estimated.
The P value for trend was calculated by including vitamin D
as a continuous variable in the model.
We defined 2 main models, one of which included
serum vitamin D and age only, and another that also included
the a priori potential confounding factors of body mass index,
physical exercise, smoking, and education. Modification of

© 2008 Lippincott Williams & Wilkins 667

Knekt et al
the effect of different risk factors on the association between
vitamin D and type 2 diabetes incidence was explored by
including an interaction term between the vitamin D variable
as a continuous variable and the potential effect-modifying
factor (ie, sex, age, season, hypertension, body mass index,
and the factors serum cholesterol and blood pressure) as a
categorical variable.
The pooling methodology is described in more detail
elsewhere26 and only briefly described here. The subcohort-
specific logs of relative odds were combined, weighting them
by the inverse of their variance in a random-effects model.27
The P value for test of trend was based on a Wald test of the
pooled estimates. Pooled P value for test of interaction was
obtained using the squared Wald statistic in which the
squared pooled estimate of the interaction coefficient was
divided by its variance and referring the Wald statistic to a ␹2
distribution with 1 degree of freedom. Heterogeneity among
the study-specific relative odds was tested using the asymp-
totic DerSimonian and Laird Q statistic.27 The potential
modification of the effect of exposure (heterogeneity) due to
sex was tested by the Wald test.28 The calculations were
performed using SAS (version 9.1; SAS Institute, Cary, NC).
RESULTS
An inverse association between age-adjusted serum
vitamin D and type 2 diabetes incidence was found in the
pooled population of individuals (Table 2). The age-adjusted
relative odds (OR) of the disease comparing the highest with
the lowest quartiles of the serum concentration of vitamin D
was 0.60 (CI ⫽ 0.37– 0.96; P for trend ⫽ 0.06; P for
heterogeneity ⫽ 0.46) (Fig. 1). The pooled relative odds are
apparently an appropriate summary of the data since testing
for heterogeneity among substudies did not indicate signifi-
cant differences (P for heterogeneity ⫽ 0.46). After further
FIGURE 1. Study- and sex-specific and pooled age-adjusted
relative odds of type 2 diabetes comparing highest and lowest
quartiles of serum vitamin D. The black squares and horizontal
lines represent study- and sex-specific ORs and 95% CIs,
respectively. The area of the black squares reflects the study-
and sex-specific weight (inverse of the standard error). The
diamond represents the pooled ORs and 95% CI. The vertical
dashed line represents the pooled relative risk.
668
Epidemiology • Volume 19, Number 5, September 2008
adjustment for the potential confounding factors of body mass
index, physical activity, smoking, and education, the OR was
unchanged (OR ⫽ 0.60; CI ⫽ 0.25–1.48; P for trend ⫽ 0.38; P
for heterogeneity ⫽ 0.08). The increase in heterogeneity was
due to different effects in men and women for adjustment for
body mass index. This was eliminated by stratifying the analysis
by sex. The adjusted relative odds were 0.28 (CI ⫽ 0.10 – 0.81;
P for trend ⬍0.001; P for heterogeneity ⫽ 0.44) for men and
1.14 (CI ⫽ 0.60 –2.17; P for trend ⫽ 0.89; P for heterogeneity
⫽ 0.64) for women (Table 3). Further adjustment for serum
cholesterol and blood pressure or exclusion of the cases occur-
ring during the first 5 years of follow-up did not notably alter the
results. Inclusion of an interaction term between vitamin D and
age, body mass index, serum cholesterol, blood pressure, and
season did not notably alter the results (data not shown).
DISCUSSION
High serum concentration of vitamin D was related to
a reduced incidence of type 2 diabetes in our study based on
pooling 2 nested case-control studies. Men in the highest
quartile of serum vitamin D had an 82% lower risk compared
with those in the lowest quartile after adjustment for body
mass index, physical activity, smoking, and education. Fur-
ther adjustment for the intermediate factors serum cholesterol
and blood pressure did not change the results.
These data suggest that vitamin D may provide protec-
tion against type 2 diabetes mellitus. As far as we know, only
one previous prospective study on this topic has been pub-
lished.15 That study, which also reported an association, was
based on vitamin D intake. Intake does not include variation
in vitamin D due to sunlight.
The effect of vitamin D on pancreatic ␤-cells and
subsequent insulin release is mediated through vitamin D
receptor, and thus the major focus of the association of
vitamin D deficiency and diabetes mellitus has been the
function of this receptor in humans and experimental ani-
mals.6,8 Polymorphism of this receptor leading to impaired
function has been observed especially in type 1 diabetes, but
observations have also included individuals with type 2
diabetes.29 Vitamin D is thought to promote insulin secretion
by increasing the cytosolic calcium concentration in ␤-cells.
The effect of insufficient vitamin D supply on insulin resis-
tance has not been investigated as closely as the effects on
insulin secretion. However, according to a recent metabolic
study using a hyperglycemic clamp technique on healthy
persons,7 serum vitamin D concentration is negatively corre-
lated with first-phase insulin response, which indicates insu-
lin resistance. Furthermore, treatment of women with type 2
diabetes by vitamin D supplements decreases insulin resis-
tance.30 In cross-sectional study of the third National Health
and Nutrition Examination Surveys (NHANES), serum vita-
min D concentration was inversely correlated with newly
detected diabetes and Homeostasis Model Assessment index
© 2008 Lippincott Williams & Wilkins
Epidemiology • Volume 19, Number 5, September 2008 Serum Vitamin D and Type 2 Diabetes

TABLE 2. Association of Quartiles of Serum Vitamin D With Type 2 Diabetes for the 2 Studies
Vitamin D Quartilea
P for
b
No. No. 1 (Lowest) 2 3 4 (Highest) P for P for Heterogeneity
Cases Controls OR OR (95% CI) OR (95% CI) OR (95% CI) Trend Heterogeneity by Sex

Model Ac
Pooled 412 986 1.00 0.99 (0.60–1.62) 0.99 (0.68–1.44) 0.60 (0.37–0.96) 0.06 0.46 0.67
FMC
Men 105 206 1.00 1.06 (0.53–2.11) 1.28 (0.62–2.63) 0.70 (0.30–1.62) 0.24
Women 125 246 1.00 0.53 (0.28–1.03) 0.70 (0.36–1.35) 0.50 (0.25–0.9998) 0.04
MFH
Men 83 245 1.00 0.94 (0.49–1.80) 0.76 (0.37–1.57) 0.32 (0.13–0.80) 0.02
Women 99 289 1.00 1.86 (0.92–3.74) 1.52 (0.73–3.20) 1.06 (0.49–2.30) 0.72
Model Bd
Pooled 405 969 1.00 1.00 (0.60–1.69) 1.06 (0.57–1.97) 0.60 (0.25–1.48) 0.38 0.08 0.14
FMC
Men 102 198 1.00 0.73 (0.28–1.95) 1.44 (0.53–3.87) 0.49 (0.15–1.64) 0.06
Women 121 239 1.00 0.77 (0.32–1.82) 1.37 (0.57–3.28) 0.91 (0.37–2.23) 0.66
MFH
Men 83 243 1.00 0.77 (0.33–1.80) 0.40 (0.16–1.03) 0.17 (0.05–0.52) ⬍0.001
Women 99 289 1.00 2.08 (0.93–4.66) 1.55 (0.65–3.71) 1.45 (0.58–3.62) 0.83
Model Ce
Pooled 307 770 1.00 0.91 (0.50–1.67) 1.13 (0.60–2.13) 0.48 (0.14–1.73) 0.09 0.03 0.12
FMC
Men 76 146 1.00 0.90 (0.31–2.63) 1.68 (0.54–5.22) 0.41 (0.10–1.65) 0.05
Women 88 206 1.00 0.82 (0.32–2.11) 0.88 (0.33–2.30) 0.60 (0.21–1.71) 0.21
MFH
Men 64 187 1.00 0.44 (0.15–1.25) 0.51 (0.18–1.50) 0.07 (0.02–0.35) 0.001
Women 79 231 1.00 1.97 (0.74–5.20) 2.22 (0.78–6.26) 2.17 (0.71–6.63) 0.31
Model Df
Pooled 403 961 1.00 1.07 (0.55–2.05) 1.16 (0.56–2.40) 0.67 (0.23–1.96) 0.14 0.04 0.13
FMC
Men 102 198 1.00 0.95 (0.33–2.67) 1.96 (0.64–5.97) 0.63 (0.18–2.26) 0.16
Women 121 239 1.00 0.77 (0.30–1.96) 1.35 (0.53–3.46) 1.14 (0.42–3.16) 0.90
MFH
Men 81 240 1.00 0.61 (0.23–1.62) 0.37 (0.13–1.05) 0.12 (0.03–0.46) 0.60
Women 99 284 1.00 2.53 (1.09–5.86) 1.77 (0.70–4.46) 1.74 (0.66–4.54) 0.001
a
Measured from controls. The range of vitamin D quartiles is described in Table 1.
b
Reference category.
c
Model A: adjusted for age.
d
Model B: adjusted for age, body mass index (kg/m2: ⬍23, 23–24.9, 25–27.4, 27.5–29.9, 30⫹), physical activity (inactive, occasionally, or regularly active), smoking (never,
past, current smoker, and dose), and education (prestage, basic level, intermediate, or high level).
e
Model C: model B with the exclusion of diabetes cases during first 5 years of follow-up.
f
Model D: model B plus blood pressure and cholesterol.
FMC indicates Finnish Mobile Clinic Health Examination Survey; MFH, Mini-Finland Health Survey.

for insulin resistance, but not with Homeostasis Model As-
sessment index for ␤-cell function14 and, in a separate study,
with metabolic syndrome.31 Thus, according to these obser-
vations, vitamin D deficiency seems to be associated with the
2 central components of type 2 diabetes pathogenesis: im-
paired insulin secretion and peripheral insulin resistance.
The strengths of the present study are the prospective
study design; the use of serum vitamin D concentration
covering the total amount of the vitamin available (due both
to diet and to sunlight); the pooled data giving greater power
to detect an association; and better ability to investigate
potential effect modification.
There are some methodologic factors that may have
suppressed the true associations or have caused artificial
associations. First, we could not obtain a comprehensive
picture of vitamin D exposure. The baseline examination was
carried out during different parts of the year and vitamin D
levels vary strongly by sunlight exposure. No measurements
were carried out during July, thus depressing the seasonal
variation in vitamin D values. Study of the interaction be-

© 2008 Lippincott Williams & Wilkins 669

Knekt et al Epidemiology • Volume 19, Number 5, September 2008

TABLE 3. Pooled Relative Odds of Type 2 Diabetes Comparing the Highest

and the Lowest Quartiles of Serum Vitamin Da According to Sex
No. Cases No. Controls OR (95% CI) P for Trend P for Heterogeneity

Model Ab
Men 188 451 0.49 (0.23–1.03) 0.01 0.39
Women 224 535 0.71 (0.34–1.49) 0.20 0.20
Model Bb
Men 185 441 0.28 (0.10–0.81) ⬍0.001 0.44
Women 220 528 1.14 (0.60–2.17) 0.89 0.64
Model Cb
Men 140 333 0.18 (0.03–0.97) ⬍0.001 0.50
Women 167 437 1.12 (0.32–3.96) 0.92 0.11
Model Db
Men 183 438 0.28 (0.05–1.43) 0.01 0.23
Women 220 523 1.42 (0.71–2.84) 0.75 0.66
a
Measured from controls.
b
As described in Table 2 footnotes.

tween serum vitamin D concentration and season (sunny vs.
dark period) confirmed a lack of bias due to season, however.
Another potential source of bias is the long storage time of
the serum samples before serum vitamin D determinations.
The few studies published so far have given contradictory
results, with a reduction in vitamin D concentration during 11
months’ storage of plasma samples at ⫺18°C32 and only a
slight change after 22 months’ storage.33
Furthermore, possible changes in dietary habits during
the 22-year follow-up period may have altered the vitamin D
status over time and thus have weakened the association
between vitamin D status and type 2 diabetes incidence. For
example, fish intake (and, accordingly, vitamin D intake) has
increased during the follow-up period.34 Second, since our
outcome variable included patients receiving reimbursed
medication for treatment of type 2 diabetes, all individuals
treating their disease with dietary changes are included in the
control group. This will cause conservative estimates of the
strength of association between vitamin D status and type 2
diabetes risk. Third, a more detailed study of the complex
patterns of different nutrients in the diet would be necessary
to obtain a deeper understanding of the benefits of vitamin D,
and therefore the lack of information on dietary factors may
have oversimplified our interpretations. Fourth, our study
population (especially the women) had low levels of vitamin
D, which may be due to the low intake of fish and short period
of sun in Finland; therefore, the findings of this study cannot
be generalized to populations with high levels of vitamin D.
Fifth, although the main nondietary risk factors of type 2
diabetes were adjusted for in this study,35 there could none-
theless be uncontrolled residual confounding, especially due
to dietary factors but also unmeasured nondietary factors.
Finally, the reduced risk could be associated with a healthy
lifestyle or some other beneficial factor related to the sources
of vitamin D.
In conclusion, the results of the present pooled study
support the hypothesis that low vitamin D status predicts
development of type 2 diabetes. Since the results may be due
to confounding by dietary and lifestyle factors, further studies
are needed before firm conclusions can be made about the
role of vitamin D in diabetes prevention.
REFERENCES
1. Peterlik M, Cross HS. Vitamin D and calcium deficits predispose for
multiple chronic diseases. Eur J Clin Invest. 2005;35:290 –304.
2. Grant WB. Epidemiology of disease risks in relation to vitamin D
insufficiency. Prog Biophys Mol Biol. 2006;92:65–79.
3. Hyppönen E. Micronutrients and the risk of type 1 diabetes: vitamin D,
vitamin E, and nicotinamide. Nutr Rev. 2004;62:340 –347.
4. Mathieu C, Badenhoop K. Vitamin D and type 1 diabetes mellitus: state
of the art. Trends Endocrinol Metab. 2005;16:261–266.
5. Lee S, Clark SA, Gill RK, et al. 1,25-Dihydroxyvitamin D3 and
pancreatic beta-cell function: vitamin D receptors, gene expression, and
insulin secretion. Endocrinology. 1994;134:1602–1610.
6. Reis AF, Hauache OM, Velho G. Vitamin D endocrine system and the
genetic susceptibility to diabetes, obesity and vascular disease. A review
of evidence. Diabetes Metab. 2005;31:318 –325.
7. Chiu KC, Chu A, Go VL, et al. Hypovitaminosis D is associated with
insulin resistance and beta cell dysfunction. Am J Clin Nutr. 2004;79:
820 – 825.
8. Mathieu C, Gysemans C, Giulietti A, et al. Vitamin D and diabetes.
Diabetologia. 2005;48:1247–1257.
9. Baynes KC, Boucher BJ, Feskens EJ, et al. Vitamin D, glucose tolerance
and insulinaemia in elderly men. Diabetologia. 1997;40:344 –347.
10. Boucher BJ, Mannan N, Noonan K, et al. Glucose intolerance and
impairment of insulin secretion in relation to vitamin D deficiency in
east London Asians. Diabetologia. 1995;38:1239 –1245.
11. Di Cesar DJ, Ploutz-Snyder R, Weinstock RS, et al. Vitamin D defi-
ciency is more common in type 2 than in type 1 diabetes. Diabetes Care.
2006;29:174.
12. Pittas AG, Lau J, Hu FB, et al. The role of vitamin D and calcium in type
2 diabetes. A systematic review and meta-analysis. J Clin Endocrinol
Metab. 2007;92:2017–2029.
13. Scragg R, Holdaway I, Singh V, et al. Serum 25-hydroxyvitamin D3
levels decreased in impaired glucose tolerance and diabetes mellitus.
Diabetes Res Clin Pract. 1995;27:181–188.
14. Scragg R, Sowers M, Bell C. Serum 25-hydroxyvitamin D, diabetes, and
ethnicity in the Third National Health and Nutrition Examination Sur-
vey. Diabetes Care. 2004;27:2813–2818.

670 © 2008 Lippincott Williams & Wilkins

Epidemiology • Volume 19, Number 5, September 2008
15. Pittas AG, Dawson-Hughes B, Li T, et al. Vitamin D and calcium intake
in relation to type 2 diabetes in women. Diabetes Care. 2006;29:650 –
656.
16. Holick MF. Vitamin D. In: Shils ME, Olson JA, Shike M, eds. Modern
Nutrition in Health and Disease. Philadelphia: Williams & Wilkins;
1994.
17. Laaksonen M, Knekt P, Marniemi J, et al. Serum vitamin D and
occurrence of subsequent type 2 diabetes. Public Health Nutr. 2006;9:
232.
18. Mattila C, Knekt P, Männistö S, et al. Serum 25-hydroxyvitamin D
concentration and subsequent risk of type 2 diabetes. Diabetes Care.
2007;30:2569 –2570.
19. Reunanen A, Aromaa A, Pyörälä K, et al. The Social Insurance Insti-
tution’s coronary heart disease study. Baseline data and 5-year mortality
experience. Acta Med Scand Suppl. 1983;673:1–120.
20. Aromaa A, Heliövaara M, Impivaara O. Aims, methods and study
population. Part 1. In: Aromaa A, Heliovaara M, Impivaara O, et al, eds.
The Execution of the Mini-Finland Health Survey 关In Finnish, English
summary兴. Helsinki and Turku, Finland: Publications of the Social
Insurance Institution, Finland, ML: 88;1989.
21. Aromaa A, Heliövaara M, Impivaara O, et al. Health, Functional
Limitations, and Need for Care in Finland. Basic results from the
Mini-Finland Health Survey 关In Finnish, English summary兴. Helsinki
and Turku, Finland; Publications of the Social Insurance Institution AL:
32;1989.
22. Reunanen A, Kangas T, Martikainen J, et al. Nationwide survey of
comorbidity, use, and costs of all medications in Finnish diabetic
individuals. Diabetes Care. 2000;23:1265–1271.
23. World Health Organization. Diabetes mellitus: report of a WHO study
group. Geneva: WHO; 1985.
24. Knekt P, Marniemi J, Teppo L, et al. Is low selenium status a risk factor
for lung cancer? Am J Epidemiol. 1998;148:975–982.
© 2008 Lippincott Williams & Wilkins
Serum Vitamin D and Type 2 Diabetes
25. Breslow NE, Day NE. Statistical Methods in Cancer Research. Vol. 1:
The analysis of case-control studies. Lyon: International Agency for
Research on Cancer Scientific Publications; 1980.
26. Knekt P, Ritz J, Pereira MA, et al. Antioxidant vitamins and coronary
heart disease risk: a pooled analysis of 9 cohorts. Am J Clin Nutr.
2004;80:1508 –1520.
27. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin
Trials. 1986;7:177–188.
28. Stram DO. Meta-analysis of published data using a linear mixed-effects
model. Biometrics. 1996;52:536 –544.
29. Oh JY, Barrett-Connor E. Association between vitamin D receptor
polymorphism and type 2 diabetes or metabolic syndrome in commu-
nity-dwelling older adults: the Rancho Bernardo Study. Metabolism.
2002;51:356 –359.
30. Borissova AM, Tankova T, Kirilov G, et al. The effect of vitamin D3 on
insulin secretion and peripheral insulin sensitivity in type 2 diabetic
patients. Int J Clin Pract. 2003;57:258 –261.
31. Ford ES, Ajani UA, McGuire LC, et al. Concentrations of serum vitamin
D and the metabolic syndrome among U.S. adults. Diabetes Care.
2005;28:1228 –1230.
32. Norris RL, Thomas MJ, Craswell PW. Assessment of a two-step high-
performance liquid chromatographic assay using dual-wavelength ultra-
violet monitoring for 25-hydroxyergocalciferol and 25-hydroxycholecal-
ciferol in human serum or plasma. J Chromatogr. 1986;381:53– 61.
33. Stamp TC, Round JM. Seasonal changes in human plasma levels of
25-hydroxyvitamin D. Nature. 1974;247:563–565.
34. Järvinen R, Seppänen R, Knekt P. Short-term and long-term reproduc-
ibility of dietary history interview data. Int J Epidemiol. 1993;22:520 –
527.
35. van Dam RM. The epidemiology of lifestyle and risk for type 2 diabetes.
Eur J Epidemiol. 2003;18:1115–1125.
671