点击显示 收起
1 From the Diabetes Research Centre, Aker and Ullevål University Hospitals, Department of Paediatrics, Ullevål University Hospital, Oslo (LCS and GJ), and the Division of Epidemiology, Norwegian Institute of Public Health, Oslo (LCS).
2 Supported by a grant (148359/330) from the Research Council of Norway (to LCS and GJ), the Norwegian Diabetes Association, Aker Diabetes Research Foundation, Novo Nordisk Fonden, Novo Nordisk Scandinavia AS, and TINE Norwegian Diaries. Peter Möller AS, a manufacturer of cod liver oil in Norway, provided a research grant to support part of this study. 3 Address reprint requests to LC Stene, Division of Epidemiology, Norwegian Institute of Public Health, PO Box 4404, Nydalen, N-0403 Oslo, Norway. E-mail: lars.christian.stene{at}fhi.no
ABSTRACT
Background: In Norway, cod liver oil is an important source of dietary vitamin D and the long-chain n-3 fatty acids eicosapentaenoic acid and docosahexaenoic acid, all of which have biological properties of potential relevance for the prevention of type 1 diabetes.
Objective: The main objective was to investigate whether the use of dietary cod liver oil or other vitamin D supplements, either by the mother during pregnancy or by the child during the first year of life, is associated with a lower risk of type 1 diabetes among children.
Design: We designed a nationwide case-control study in Norway with 545 cases of childhood-onset type 1 diabetes and 1668 population control subjects. Families were contacted by mail, and they completed a questionnaire on the frequency of use of cod liver oil and other vitamin D supplements and other relevant factors.
Results: Use of cod liver oil in the first year of life was associated with a significantly lower risk of type 1 diabetes (adjusted odds ratio: 0.74; 95% CI: 0.56, 0.99). Use of other vitamin D supplements during the first year of life and maternal use of cod liver oil or other vitamin D supplements during pregnancy were not associated with type 1 diabetes.
Conclusion: Cod liver oil may reduce the risk of type 1 diabetes, perhaps through the antiinflammatory effects of long-chain n-3 fatty acids.
Key Words: Diabetes mellitus • insulin-dependent diabetes mellitus • diet • fatty acids • vitamin D • case-control study • population-based study • childhood
INTRODUCTION
Type 1 diabetes is among the most prevalent chronic diseases with onset in childhood. It results from an immune-mediated destruction of pancreatic ß cells and is linked to genes in the HLA complex on chromosome 6p21 (1). However, genetic susceptibility is not sufficient for development of disease. Although the environmental triggers of the disease are essentially unknown, early diet is among the strongest candidates, together with viral infections (2). Many studies have investigated the role of breastfeeding and the timing of introduction of cow milk (2). Other aspects of diet have also been investigated in relation to the development of type 1 diabetes, such as nitrosamines (3) and ecologic relations with population-level consumption of cow milk (4), fat, proteins, and meat (5). A few studies have focused on immunomodulatory effects of vitamin D, which may be relevant in the prevention of type 1 diabetes (6). After 1,25-dihydroxyvitamin D was found to prevent autoimmune diabetes in nonobese diabetic mice, 2 epidemiologic studies showed associations between the use of vitamin D supplements in the first year of life and a lower risk of type 1 diabetes (7, 8).
Cod liver oil is an important source of both vitamin D and the long-chain n-3 fatty acids docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA) in the Norwegian population (9, 10). In Norway, dietary vitamin D supplementation is recommended from infancy, preferably in the form of cod liver oil (11). Long-chain n-3 fatty acids are incorporated into cell membranes and have antiinflammatory properties that may be relevant for the prevention of type 1 diabetes, such as decreased expression of HLA class II molecules on activated human monocytes (12) and reduced expression of interleukin 1ß (13). The long-chain n-3 fatty acids play an important role in eicosanoid metabolism, and there is evidence for aberrant prostaglandin metabolism in children with type 1 diabetes (14, 15). Together, these data suggest that the antiinflammatory n-3 fatty acids such as DHA and EPA may reduce the risk of disease development. The n-3 fatty acid and vitamin D status of a newborn depend on that of the mother during pregnancy (16, 17). We conducted a pilot case-control study to test the hypothesis that cod liver oil, taken either by the mother during pregnancy or by the child during the first year of life, is associated with a lower risk of type 1 diabetes among children (18). We have now conducted a much larger case-controlstudy, with cases of type 1 diabetes of more recent onset. As an additional improvement over the previous study, we also collected information on the frequency of use and the age at start of supplementation in the first year of life.
The main objective of the present study was to test the hypothesis that the use of cod liver oil or other vitamin D supplements, either by the mother during pregnancy or by the child during the first year of life, is associated with a lower risk of type 1 diabetes among children. Secondary objectives were to estimate the associations of duration of exclusive breastfeeding, age at introduction of solid foods, maternal education, and maternal smoking during pregnancy with the risk of childhood-onset type 1 diabetes.
SUBJECTS AND METHODS
All new cases of type 1 diabetes diagnosed in children aged < 15 y have been prospectively registered in Norway since 1989 with a high degree of ascertainment with the use of the EURODIAB criteria (19). On the basis of the results of a pilot study conducted in one county of Norway (18), we designed a nationwide case-control study of type 1 diabetes, excluding anyone who was part of the pilot study. All children in the diabetes registry with a date of diagnosis between 1997 and 2000 and who were born between 1 January 1985 and 31 December 1999 (n = 801) were eligible for the study. Additionally, 3000 randomly selected control children from the national population registry who were born in the same time period were eligible. In Norway, the national population registry includes the name, date of birth, unique personal identification number, and address of residence of all inhabitants, and all inhabitants are legally required to report to the national population registry within 8 d of changing residence. During the fall of 2001, the families of all eligible children were mailed a written explanation of the purpose of the study, a consent form to be signed by the participants and their parents or guardians, and a 4-page questionnaire. The questionnaire focused on environment and diet during early childhood and during the pregnancy with the index child. Of all eligible children, 53 cases and 33 controls had moved without notifying the population registry and could not be contacted; 203 eligible cases and 1299 eligible controls did not return the questionnaire. This left 545 cases (73% response) and 1668 controls (56% response) in the study. The response was similar in all 19 counties of Norway and for boys and girls. There was a slightly lower response among the very youngest and the very oldest, but no single 1-y age group had a < 50% response rate. Questionnaires returned by mail were optically scanned and checked for possible errors before data analysis. Written informed consent was obtained from all participants, and the study was approved by the Regional Ethics Committee and the National Data Inspectorate. The mean (± SD) age at diagnosis among the participating cases was 8.80 ± 3.47 y.
Variables
In the questionnaire, the child with type 1 diabetes or the control child was defined as the participant. The mothers replied to the following questions:
"Did the participant’s mother use cod liver oil during the pregnancy with the participant?"
"If yes, how often?" (The precoded frequency categories for this question were "less than once a week," "about one to four days per week," and "nearly every day.")
"Did the participant’s mother use multivitamin or mineral supplements during the pregnancy with the participant?"
"If yes, how often?" (Precoded frequency categories were as above.)
"If yes, what kind of multivitamin or mineral supplements did you use?" [The precoded categories were 3 of the most commonly used brands of multivitamin supplements for adults in Norway (all of which contained 5 µg vitamin D) and "other, specify."]
"Did the participant get cod liver oil during the first year of life?"
"If yes, from what age?"
"If yes, how often?" (Precoded frequency categories were as above.)
"Did the participant get a dietary supplement with vitamin D (other than cod liver oil) during the first year of life?"
"If yes, from what age?"
"If yes, how often?" (Precoded frequency categories were as above.)
"If yes, what kind?" (The precoded categories were 3 of the most commonly used brands in Norway and "other, specify.")
Because the recommended daily dose of cod liver oil and other supplements is specified on the bottle or box and because of the relatively long recall period for most mothers, no attempts were made to obtain information on amounts taken (the recommended daily dose of cod liver oil is 5 mL, providing 0.6 g DHA, 0.4 g EPA, and 10 µg vitamin D). The recommended dose of other vitamin D supplements for infants also contains 10 µg vitamin D. For all supplements, a frequency of use of less than once per week was regarded as nonuse.
For information on infant feeding, the mothers were asked:
"Was the participant breastfed?"
"To what age did the participant get only breast milk and no other food (excluding water, cod liver oil, or other vitamin D supplements)?"
"At what age did the participant cease to receive breast milk?"
"At what age did the child first receive solid foods?"
As a marker for socioeconomic status, we used highest maternal education at the time of delivery of the index child. Six precoded categories were specified (Table 1). For smoking during pregnancy, the mothers were asked "Did the participant’s mother smoke daily or nearly daily during the pregnancy with the participant?" If the mothers smoked during pregnancy, they were asked to specify whether they smoked during the first part of the pregnancy, during the last part of pregnancy, or during most of the pregnancy. Because very few answered that they smoked only during the last part of the pregnancy, these respondents were pooled with those who said they smoked during the first part for a joint category of "during parts of pregnancy" for the data analysis. The questionnaire also included a section inquiring about all siblings and parents, such as their year of birth; whether they had diabetes; if yes, whether they use insulin; and the year of onset of diabetes. Maternal age at delivery (20) was calculated as the difference between the date of birth of the participant (child) and that of his or her mother. Information on the sex and date of birth of all participants was obtained from the national population registry. For the other variables, a varying number of participants had missing data as specified in the footnotes to the tables.
View this table:
TABLE 1. Descriptive characteristics of cases with type 1 diabetes and randomly selected population control subjects in Norway
Statistical analysis
For our prestudy power calculations, we estimated that we would recruit 600 cases and 1800 controls. On the basis of our pilot study, which indicated a prevalence of maternal use of cod liver oil during pregnancy of 17% and a prevalence of use of cod liver oil during the first year of life of 34%, we calculated that we would have =" BORDER="0"> 84% power to detect a significant association in the present study if the true odds ratio for maternal use of cod liver oil was 0.67. For children’s use of cod liver oil, the corresponding power was 97%. If the true odds ratio were 0.50, we had > 95% power to detect significant associations in the present study, even for dietary supplements used by < 10% of the population. All power calculations were based on two-sided tests and 5% significance levels. All data analyses were done by using SPSS, version 11 (SPSS Inc, Chicago).
Odds ratios with 95% CIs were estimated by using logistic regression with type 1 diabetes as the dependent variable. We decided before the study that all of the variables listed in the section above would be included in adjusted analyses, regardless of statistical significance. We investigated whether the associations were of similar magnitude for different sexes and age groups by stratified analysis and by testing the respective two-way interaction terms in the logistic regression models (21). Continuous variables were categorized and checked for logit-linearity as were categorical variables with > 2 categories, such as maternal education. Cutoffs for categorization of continuous variables were based on the distribution among the control subjects and were decided on before the associations with type 1 diabetes were analyzed. Maternal education fitted a logit-linear dose-response pattern in the regression models when coded 1, 2,...6 for the 6 levels (see Table 1), whereas the other variables were entered as dummy variables, allowing for nonlinear relations. Age at data collection was entered with 5 categories (< 5, 5–7.9, 8–10.9, 11–13.9, and =" BORDER="0"> 14 y). P values < 0.05 or 95% CIs not overlapping the null value 1.00 for the odds ratio were regarded as statistically significant.
RESULTS
In the unadjusted analyses, there was a significant relation between longer duration of exclusive breastfeeding and lower risk of type 1 diabetes and between higher maternal education and lower risk of type 1 diabetes ( Table 1). These relations persisted after adjustment for potential confounders, although the clear dose-response relation with maternal education was less pronounced and not significant. Smoking during pregnancy tended to be more common among the mothers of cases than among the mothers of controls, but there was no association after adjustment for potential confounders (Table 1). There was no clear association between type 1 diabetes and age at introduction of solid foods, maternal age at delivery, or number of siblings (Table 1). As expected, type 1 diabetes among at least one sibling or parent was related to a much higher risk of type 1 diabetes.
There was no clear association between maternal use of cod liver oil or other vitamin D–containing supplements during pregnancy and type 1 diabetes among children, in either the unadjusted or the adjusted analyses (Table 2). In contrast, use of cod liver oil during the first year of life was associated with a significantly lower risk of type 1 diabetes, even after adjustment for a large number of potential confounders, such as duration of breastfeeding, age at introduction of solid foods, maternal education, maternal age at delivery, and history of diabetes in the family (Table 3). The dose-response patterns with frequency of use and age at start were not particularly clear.
View this table:
TABLE 2. Maternal use of cod liver oil and other dietary vitamin D supplements during pregnancy among cases of type 1 diabetes and randomly selected control subjects in Norway
View this table:
TABLE 3. Use of cod liver oil and other dietary vitamin D supplements during the first year of life among cases of type 1 diabetes and randomly selected control subjects in Norway
Use of other vitamin D supplements in the first year of life tended to be somewhat more common among cases than among controls (Table 3), but the association was not significant, and the suggestive association disappeared in the adjusted analyses. Nearly all participants who had used other vitamin D supplements had used 1 of 3 common multivitamin preparations, and the distribution of specific type of supplements was similar among the cases and controls (data not shown).
The association of risk of diabetes with use of cod liver oil during the first year of life was somewhat stronger among girls than among boys, and the association with breastfeeding was slightly stronger among boys than among girls; however, these interactions with sex were not significant (P = 0.07 and P = 0.06, respectively). Otherwise, all associations were similar among boys and girls and in different age groups (data not shown). Also, after adjustment for only subsets of the adjustment variables, the results were essentially unchanged. Because only 9 cases of type 1 diabetes were included from Vest-Agder county, the county where we did our pilot study, a meaningful subgroup analysis of this county could not be done. However, further adjustment for county of residence in the regression model did not influence the results (data not shown).
DISCUSSION
The main finding of this study was that use of cod liver oil during the first year of life was associated with a lower risk of type 1 diabetes. Except for our earlier pilot study (18), the present study is the first to evaluate the effect of intake of cod liver oil or long-chain n-3 fatty acids on risk of type 1 diabetes in humans. In the pilot study, a moderately strong but nonsignificant association was shown between use of cod liver oil in the first year of life and risk of type 1 diabetes, which is consistent with the results of present study. The relatively strong association between maternal use of cod liver oil during pregnancy and lower risk of type 1 diabetes in the pilot study (18) was not reproduced in the present study. The reason for the conflicting results is not known. Lack of power should not be the problem in the present study, although the number of cases from the pilot county included in the present study was too small for meaningful subgroup analysis. However, because the results were similar after adjustment for county of residence, it is unlikely that possible differences in background diet between counties explain our present findings.
One Norwegian study found a lower incidence of type 1 diabetes in fishery municipalities than in the rest of the country (22), compatible with a protective effect of marine fatty acids, although ecologic studies must be interpreted with caution. Long-chain fatty acids have been shown to reduce the risk of chemically induced diabetes in experimental animals (23) and to induce changes in Th1-Th2 balance in the gut-associated immune system but not in the pancreas of BB rats (24). Other studies using animal models found essential fatty acid deficiency to reduce the risk of diabetes (25). Under any circumstance, it is difficult to generalize the results from animal models to human disease (26).
Strengths and limitations of the present study
A major strength of the present study is that it was a large population-based study and the use of cod liver oil, which has a distinct taste and smell, is relatively easy to recall. However, although we adjusted for several potential confounders, we cannot rule out the possibility of unmeasured confounding factors. With the use of a case-control design with questionnaire data, there is always the possibility of recall bias, but the direction of such possible bias is not obvious. We would expect that any such bias would be similar for maternal use of supplements and for use of other vitamin D supplements by children. The lack of significant association for these other factors is thus an argument against recall bias as an explanation for our main finding. Furthermore, the possible relation between cod liver oil or vitamin D supplements and type 1 diabetes is not known by the public, and there is usually no particular focus on these dietary factors in the treatment of childhood-onset type 1 diabetes.
We did not have access to blood samples for measurement of biomarkers of n-3 fatty acids or of vitamin D status among the participants in the present study. A few small studies found lower n-3 fatty acids status (27) and lower serum concentrations of 25-dihydroxyvitamin D3 (28, 29) in patients with type 1 diabetes than in healthy control children, although the results of other studies do not support the latter finding (30). However, biomarkers of exposure are problematic in case-control studies because they may be affected by the disease or by the treatment. Cod liver oil is an important source of both vitamin D (9) and the long-chain n-3 fatty acids EPA and DHA in the Norwegian population (10). A daily dose of cod liver oil contributes 0.6 g DHA and 0.4 g EPA. Although there are a few other sources of these fatty acids in the diet, such as fatty fish, those who regularly use cod liver oil have, on average, a much higher dietary intake of DHA and EPA than do those who do not use cod liver oil (10). The same argument applies to vitamin D. The normal diets of adults in Norway contain an average of about one-half to one-third of the recommended daily intake of vitamin D, depending on age group (9). The diet of unsupplemented Norwegian children aged 0–1 y is likely to be even lower in vitamin D, particularly if the child is exclusively breastfed. A complicating factor is that vitamin D is also produced in the skin as a result of sun exposure (31). In the future, biomarkers should be used to investigate the relations between n-3 fatty acids or vitamin D and the risk of developing islet autoimmunity or type 1 diabetes in some of the ongoing prospective studies among children at genetically increased risk of type 1 diabetes (32).
We focused on exposure in utero and during the first year of life. We thought this would help to avoid confusion of different periods in the child’s life and would be easier for the mothers to recall. Some children were only 1 y old at the time of data collection, and these time windows roughly correspond to the concept that potential environmental factors influencing the pathogenesis of type 1 diabetes may operate early in life (33). However, cod liver oil or other vitamin D supplements may well play a role after the first year of life as well. Our decision to regard those who started using cod liver oil or other supplements at age 13 mo or later as nonusers probably led to an attenuation of the observed association. Previous findings of an association between use of vitamin D supplements (other than cod liver oil) in the first year of life and a lower risk of type 1 diabetes (7, 8) were not supported by the present study. The reasons for this difference are not known, but it may in part be explained by differences in study design and in customs for use and recall of vitamin D supplements in different populations. As discussed in our previous paper (18), some data indicate that intake or bioavailability of vitamin D may be higher from cod liver oil than from other supplements, at least for maternal supplementation during pregnancy.
Possible explanations of the main finding
Because there was no association between risk of type 1 diabetes and use of other vitamin D supplements, our data support a possible effect of the long-chain n-3 fatty acids EPA and DHA, at least in the range of doses used in the Norwegian population. These fatty acids may influence gene expression, have antiinflammatory effects, and have been shown to be relevant in the prevention and treatment of several chronic diseases (34). The long-chain n-3 fatty acids participate in eicosanoid metabolism. Chase et al (14) found indications of changes in prostaglandin synthesis in children with type 1 diabetes. Litherland et al (15) reported an increased expression of prostaglandin synthase 2 (also known as COX-2) on activated monocytes of persons with type 1 diabetes. The aberrant prostaglandin metabolism was also present among nondiabetic subjects at genetically increased risk of type 1 diabetes who were positive for islet cell autoantibodies (15). This suggests that the defect in prostaglandin metabolism causes an increased risk of type 1 diabetes. DHA and EPA may act as inhibitors of prostaglandin synthase 2 (35). Our main finding may be explained by this pathway or other antiinflammatory effects (12, 13) of EPA and DHA. Such antiinflammatory effects may be of particular importance in the pancreas and perhaps also in the gut (36).
There is an ongoing debate as to whether the association between longer duration of breastfeeding and lower risk of type 1 diabetes found in many studies can be explained by recall bias (37). The fact that breast milk, as opposed to most infant formulas, contains some n-3 fatty acids (38) is one possible and novel explanation for the relation.
Conclusion
We found a significant association between the use of cod liver oil during the first year of life and a lower risk of type 1 diabetes, suggesting that cod liver oil may reduce the risk of type 1 diabetes, perhaps through the effects of long-chain n-3 fatty acids. If our results are corroborated in prospective studies with genetically high-risk persons or persons with autoantibodies predictive of type 1 diabetes, cod liver oil or individual fatty acids such as DHA may be candidates for preventive intervention trials.
ACKNOWLEDGMENTS
GJ was responsible for the Norwegian Childhood Diabetes Registry, and all members of the Norwegian Childhood Diabetes Study Group (listed below) participated in the registration of new-onset cases of type 1 diabetes. Both GJ and LCS participated in writing the protocol for the present study. LCS constructed the questionnaire, collected the data, performed the data cleaning and data analysis, and wrote the manuscript with input from GJ. The funding sources had no involvement in study design, data collection, analysis, or interpretation and no involvement in writing the manuscript or deciding to submit it for publication.
The members of the Norwegian Childhood Diabetes Study Group were as follows: Henning Aabech, Barneavdelingen, Sykehuset Østfold, 1603 Fredrikstad; Helge Vogt, Barneavdelingen, Sentralsykehuset i Akershus, 1474 Nordbyhagen; Kolbeinn Gudmundsson, Barneklinikken, Rikshospitalet, 0032 Oslo; Hans-Jacob Bangstad, Knut Dahl-Jørgensen, and Geir Joner, Barnesenteret, Ullevål sykehus, 0407 Oslo; Øystein Aagenæs, Norsk Diabetikersenter, Sponhoggveien 19, 0284 Oslo; Olav Flesvig, Barneavdelingen, Sentralsjukehuset i Hedmark, 2418 Elverum; Ola Talleraas, Barneavdelingen, Oppland sentralsykehus, 2629 Lillehammer; Halvor Bævre, Pediatrisk poliklinikk, Oppland sentralsykehus, 2819 Gjøvik; Kjell Stensvold, Barneavdelingen, Buskerud sentralsykehus, 3004 Drammen; Bjørn Halvorsen, Barneavdelingen, Vestfold sentralsykehus, 3103 Tønsberg; Kristin Hodneckvam, Barneavdelingen, Telemark sentralsjukehus, 3710 Skien; Ole Kr Danielsen, Barneavdelingen, Aust-Agder sentralsjukehus, 4809 Arendal; Jorunn Ulriksen and Geir Stangeland, Barneavdelingen, Vest-Agder sentralsykehus, 4604 Kristiansand; Jon Bland, Barneavdelingen, Sentralsjukehuset i Rogaland, 4068 Stavanger; Dag Roness, Barneavdelingen, Fylkessjukehuset i Haugesund, 5513 Haugesund; Oddmund Søvik and Pål R Njølstad, Barneklinikken, Haukeland sykehus, 5021 Bergen; Per Helge Kvistad, Barneavdelingen, Sentralsjukehuset i Song og Fjordane, 6800 Førde; Steinar Spangen, Barneavdelingen, Sentralsjukehuset i Møre og Romsdal, 6026 Ålesund; Per Erik Hæreid, Barneklinikken, Regionsykehuset i Trondheim, 7006 Trondheim; Sigurd Børsting, Barneavdelingen, Innherred Sykehus, 7600 Levanger; Ketil Mevold and Dag Veimo Barneavdelingen, Nordland Sentralsykehus, 8092 Bodø; Harald Dramsdahl, Barnepoliklinikken, Harstad sykehus, 9406 Harstad; Bård Forsdahl Barneavdelingen, Regionsykehuset i Tromsø, 9038 Tromsø; and Kersti Elisabeth Thodenius, Barneavdelingen, Hammerfest sykehus, 9600 Hammerfest.
REFERENCES