ORIGINAL COMMUNICATION

Transcription

1 (2004) 58, & 2004 Nature Publishing Group All rights reserved /04 $ ORIGINAL COMMUNICATION Habitual fish consumption and glycated haemoglobin: The EPIC-Norfolk Study A-H Harding 1, NE Day 1, K-T Khaw 1, SA Bingham 1,2, RN Luben 1, A Welsh 1 and NJ Wareham 1 * 1 Department of Public Health and Primary Care, Institute of Public Health, University of Cambridge, Cambridge, UK; and 2 Medical Research Council Dunn Human Nutrition Unit, Cambridge, UK Objective: To investigate the association between habitual fish consumption and a continuous measure of glycaemia. Design: Cross-sectional study. Setting: EPIC-Norfolk, a population-based cohort study of diet and chronic disease. Subjects and methods: In all, 4500 men and 5509 women, aged y, without self-reported diabetes. Diet was assessed by a semiquantitative food frequency questionnaire, and glycaemia was measured by glycated haemoglobin. Results: In women only, in analyses adjusted for age, the HbA 1c level was positively associated with eating fried fish and inversely associated with eating oily fish (b ¼ 0.036, 95% confidence interval (CI): , 0.069; and b ¼ 0.046, 95% CI: 0.086, respectively). These associations were attenuated by adjustment for family history of diabetes, smoking status and physical activity level, but the association with fried fish remained statistically significant (b ¼ 0.033, 95% CI: , 0.066). Adjusting for total energy, alcohol, fruit and vegetable intakes resulted in further attenuation and both associations were no longer statistically significant. In men, there was no evidence that HbA 1c level was associated with fish consumption. Conclusions: The study found no evidence of an association between fish consumption and HbA 1c after taking other lifestyle factors into account. Sponsorship: NJW is a Wellcome Trust Senior Clinical Research Fellow. (2004) 58, doi: /sj.ejcn Keywords: cohort; fish; glycated haemoglobin; lifestyle Introduction The observation that communities eating a traditional diet rich in fish experience low rates of chronic disease (Kromann & Green, 1980; Young et al, 1992) gave rise to a large body of research investigating the health benefits of fish consumption. The protective effects of fish consumption have been attributed to the long-chain n-3 polyunsaturated fatty acids contained in fish, particularly eicosopentanoic and docosahexanoic acids. A beneficial association between *Correspondence: NJ Wareham, Department of Public Health and Primary Care, Institute of Public Health, University of Cambridge, Robinson Way, Cambridge CB2 2SR, UK. njw1004@medschl.cam.ac.uk Guarantor: NJ Wareham. Contributors: A-HH and NJW contributed to the study concept, analysis and interpretation of the data and the drafting of the manuscript. K-TK, SAB, NED and NJW are principal investigators of EPIC-Norfolk. AW and RNL contributed to the study design and data collection and management. All authors contributed to the revision and drafting of the final version. Received 4 December 2002; revised 27 February 2003; accepted 2 March 2003 fish eating and glycaemia may be a consequence of the incorporation of the n-3 fatty acids into cell membranes since the dietary fatty acid profile is reflected in the fatty acid content of skeletal muscle phospholipid membranes (Storlien et al, 1996). Insulin sensitivity is positively related to the proportion of long-chain polyunsaturated fatty acids present in the cell membranes (Borkman et al, 1993). Animal experiments have also demonstrated a positive effect of n-3 fatty acids on insulin sensitivity (Vessby, 2000). However, fish oil intervention studies of people without diabetes have found no evidence of a beneficial effect on insulin sensitivity (Grundt et al, 1995; Toft et al, 1995; Lovegrove et al, 1997; Chan et al, 2002), and in one study of hyperlipidaemic patients, 4 g/day of fish oil was associated with an increase in fasting insulin (Mori et al, 2000). Similarly, intervention studies suggest that there is no effect of fish oil on blood glucose levels in nondiabetic people (Grundt et al, 1995; Toft et al, 1995; Lovegrove et al, 1997), although in hyperlipidaemic patients, fish oil was associated with an increase in fasting plasma glucose (Mori et al, 2000).

2 278 Any health benefits of fish consumption may also be attributable to other constituents of fish. Recent research in animal models suggests that protein subtypes may influence the action of dietary fats (Lavigne et al, 2001; Tremblay et al, 2003). Rats fed high-fat diets were insulin resistant when their source of protein was casein or soy, but were not insulin resistant when the protein source was cod. The dietary cod protein appeared to be a natural insulinsensitising agent. There is limited evidence relating fish consumption to glycaemia in nondiabetic populations. Most of the evidence is observational and the reported associations are inconsistent. Consumption of salmon and seal oil among Alaska Natives (Adler et al, 1994) was associated with a lower risk of impaired glucose tolerance. Habitual intake of a small amount of fish in an elderly population of Dutch men and women (Feskens et al, 1991) was associated with a reduced risk of developing impaired glucose tolerance and type 2 diabetes. However, in population-based studies of adult Inuit and Quebecers, positive associations between marine diet (Bjerregaard et al, 2000) or plasma phospholipid n-3 fatty acids (Dewailly et al, 2001a, b) and blood glucose have been reported. A weight loss intervention study reported that daily fish consumption had no independent effect on blood glucose levels (Mori et al, 1999). We undertook an analysis of the cross-sectional association between fish consumption and a continuous measure of blood glucose, glycated haemoglobin (HbA 1c ), in a Western population of men and women, aged y, without known diabetes. Methods Study population The people in this study were participants in the Norfolk arm of the European Prospective Investigation of Cancer (EPIC), an international multicentre cohort study designed to investigate the relation between diet and cancer (Riboli & Kaaks, 1997). The Norfolk study broadened its scope to include chronic diseases other than cancer. Approval for the study was obtained from the Norfolk Local Research Ethics Committee. EPIC-Norfolk is a population-based cohort study, which recruited volunteers from March 1993 to the end of 1997 from general practices in the city of Norwich and surrounding small towns (Day et al, 1999). All individuals in the age range y in each participating general practice were invited to take part. Those who consented were requested to attend the study clinic for a health check. Of the individuals contacted, 39% consented to take part and (33%) attended the health check. The recruitment target was participants. The measurement of HbA 1c began in November 1995, midway through recruitment of the cohort. The subcohort selected for this analysis comprises all individuals who had HbA 1c measured. Data collection Volunteers completed a detailed health and lifestyle questionnaire. It included questions on personal and family history of diabetes, smoking, diet and physical activity. Three questions addressed the respondent s personal history of diabetes: whether they had ever been told by a doctor that they had diabetes; whether they had modified their diet in the past year due to diabetes; and whether they followed a diabetic diet. A positive response to any of these questions was taken as an indication of prevalent diabetes. Family history of diabetes was covered in a question asking whether any of the respondent s immediate family had diabetes. The age at which diabetes was diagnosed in mother, father and/or siblings was recorded. Smoking was derived from the questions asking whether they had ever smoked as much as one cigarette a day and whether they were current smokers. The weekly consumption of fresh fruit and vegetables was reported. A four-point physical activity index was used, which combined occupational activity and recreational physical activity (hours per week in cycling and other recreational activity) (Wareham et al, in press). Volunteers completed a semiquantitative food frequency questionnaire (FFQ), designed to assess habitual diet in the past year, before attending the health check. The FFQ was based on the questionnaire developed for the US Nurses Health Study (Willett et al, 1985). The frequency categories remained unchanged, but taking information from the British National Food Survey (National Food Survey Committee, 1982), the lists of foods were modified to reflect the important sources of nutrients in the average British diet. Energy and nutrient intakes were estimated by converting food intakes reported in the FFQ into grams using software developed for use by the EPIC-Norfolk study. The FFQ included questions regarding the consumption of fried fish, white fish and oily fish. The FFQ had been validated by comparison with 16-day weighed food records (Bingham et al, 1997). Research nurses at the EPIC-Norfolk clinic carried out standardised health checks. Anthropometric measurements were taken with participants dressed in light clothing and without shoes. Height was measured to the nearest 0.1 cm using a stadiometer, and weight was measured to the nearest 100 g using Salter scales. Body mass index (BMI) was calculated as weight (kg)/height 2 (m). Waist circumference was measured to the nearest 0.1 cm at the smallest circumference between the ribs and iliac crest with the volunteers standing with abdomen relaxed, or at the level of the umbilicus if there was no natural waistline. Hips were measured to the nearest 0.1 cm at the maximum circumference between the iliac crest and the crotch. The waist:hip ratio (WHR) was calculated from these measures. Of those attending the health check, 95% consented to have blood taken. A sample of EDTA-anticoagulated blood was taken for HbA 1c measurement. The blood was stored in a refrigerator at 4 71C until it was transported at ambient temperature to be

3 assayed, within 1 week of sampling. The HbA 1c assays were undertaken using high-performance liquid chromatography on a Bio-rad Diamat (Richmond, CA, USA) (Standing & Taylor, 1992). The coefficient of variation was 3.6% at the lower end of the range (mean 4.9%) and 3.0% at the upper end of the range (mean 9.8%). Statistical analysis Participants with self-reported diabetes were excluded from the analysis since it is probable that they would have changed their diet after diagnosis, or would have altered how they reported it. Food frequency questionnaires were excluded if 10 or more lines had not been completed. Three types of fishffried fish, white fish and oily fishf were considered. Fish consumption was placed into two categories representing non- and occasional fish eaters (less than one portion of ), and regular fish eaters (one or more portions per week). All analyses were stratified by sex. The associations between the potential confounding variables, age, family history of diabetes, smoking status, physical activity level, total energy intake, alcohol, green leafy vegetables and fruit intakes, WHR and BMI, and the exposures and outcome were examined. Linear regression methods were used to investigate the relation between HbA 1c and the consumption of fish in a series of models. The simplest model included only fish intake. The final models were adjusted for age, family history of diabetes, smoking status, physical activity, total energy intake, alcohol, fruit, green leafy vegetables, BMI and WHR. Results Of the participants who had HbA 1c measured, 351 (3. 2%) reported having diabetes. The mean HbA 1c of people reporting diabetes was 7.6% (s.d. 2.02), and of those remaining in the study it was 5.2% (s.d. 0.67). After excluding those with self-reported diabetes and those with incomplete data, participants remained in the study. In total, 4% of the study population did not eat any type of fish. Apart from women eating oily fish, mean HbA 1c was similar for those eating fish regularly and for those eating less than one portion per week (Tables 1 3). However, regular fish eaters tended to be older. Adjusted for age, regular fried fish eaters had higher HbA 1c than those eating less than one portion per week, whereas regular white fish eaters tended to have lower HbA 1c (Figure 1), demonstrating that substantial confounding existed between fish consumption and age. In women only, regular oily fish eaters were slightly younger and had significantly lower HbA 1c (Table 3 and Figure 1). Of the potential confounding variables, age, smoking status, total energy intake, alcohol, fruit and green leafy vegetable intakes were most consistently associated with fish consumption. Eating fried fish regularly was associated with higher BMI, WHR and total energy intake, and with lower fruit and vegetable intakes. White fish and oily fish consumptions were associated with higher total energy, alcohol, fruit and vegetable intakes, and with lower proportions of current smokers and physically inactive people. Tables 4 and 5 summarise the study population s characteristics by quintiles of HbA 1c. For the majority of characteristics, there were significant trends across the quintiles of HbA 1c, and the trends were similar for men 279 Table 1 Characteristics of the study population according to level of fried fish intake: the EPIC-Norfolk Study (n=10 009) Men Women o1 portion fried fish per week X1 portion fried fish per week o1 portion fried fish per week X1 portion fried fish per week HbA 1c (%) 5.27 (0.687) 5.27 (0.690) 5.20 (0.662) 5.22 (0.656) Age (y) 59.2 (9.55) 57.5 (9.59)*** 57.7 (9.60) 57.0 (9.50)* BMI (kg/m 2 ) 26.2 (3.14) 26.7 (3.31)*** 25.8 (4.31) 26.2 (4.28)*** WHR (0.0578) (0.0581)* (0.0629) (0.0641)** Family history of diabetes a 181 (11) 363 (12) 327 (13) 406 (14) Physical activity level (lowest) a 494 (31) 822 (28)* 689 (27) 831 (28) Current smokers a 149 (9) 392 (13)*** 283 (11) 335 (12) Total energy intake (kj) 8800 (2500) 9480 (2610)*** 7700 (2120) 8480 (2320)*** Alcohol intake (g) b 8.59 (2.01, 5.60) 7.50 (2.01, 5.60) 3.30 (0.761, 5.50) 2.52 (0.761, 5.60)** Fruit intake a,c 1043 (66) 1619 (56)*** 2027 (78) 2013 (69)*** Green leafy vegetables a,c 652 (41) 893 (31)*** 1305 (50) 1251 (43)*** *P-value p0.05, **P-value p0.01, ***P-value p P-values relate to comparisons within sex, between those eating less than one portion of and those eating more than one portion per week, within sex. Data are means with standard deviations in parentheses and P-values refer to t-tests. a Data are counts with percents in parentheses and P-values refer to w 2 tests. b Data are medians with interquartile range in parentheses and P-values refer to Kruskal Wallis tests. c At least one portion per day.

4 280 Table 2 Characteristics of the study population according to level of white fish intake: the EPIC-Norfolk Study (n=10 009) Men Women o1 portion white X1 portion white o1 portion white X1 portion white HbA 1c (%) 5.27 (0.693) 5.27 (0.687) 5.18 (0.613) 5.22 (0.671) Age (y) 56.6 (9.57) 58.7 (9.56)*** 55.7 (9.48) 57.8 (9.52)*** BMI (kg/m 2 ) 26.5 (3.29) 26.5 (3.25) 25.8 (4.12) 26.1 (4.35) WHR (0.0595) (0.0575) (0.0597) (0.0646)** Family history of diabetes 173 (14) 371 (11)* 165 (13) 568 (13) Physical activity level (lowest) a 382 (30) 934 (29)** 379 (30) 1141 (27) Current smokers a 189 (15) 352 (11)*** 192 (15) 426 (10)*** Total energy intake (kj) 9000 (2570) 9340 (2590)*** 7790 (2180) 8210 (2280)*** Alcohol intake (g) b 5.63 (1.22, 5.60) 9.40 (2.52, 5.60)*** 1.55 (0, 5.60) 3.57 (0.761, 5.60)*** Fruit intake a,c 647 (51) 2015 (62)*** 816 (65) 3224 (76)*** Green leafy vegetables a,c 339 (27) 1206 (37)*** 434 (24) 2122 (50)*** See Table 1. Table 3 Characteristics of the study population according to level of oily fish intake: the EPIC-Norfolk Study (n=10 009) Men Women o1 portion oily X1 portion oily o1 portion oily X1 portion oily HbA 1c (%) 5.26 (0.675) 5.27 (0.694) 5.26 (0.672) 5.20 (0.654)** Age (y) 57.6 (9.98) 58.3 (9.44)* 57.8 (9.72) 57.2 (9.50)* BMI (kg/m 2 ) 26.6 (3.33) 26.4 (3.23) 26.1 (4.47) 26.0 (4.25) WHR (0.0598) (0.0573)** (0.0627) (0.0638) Family history of diabetes 163 (12) 381 (12) 166 (14) 567 (13) Physical activity level (lowest) a 402 (30) 914 (29) 399 (33) 1121 (26)*** Current smokers a 215 (16) 326 (10)*** 168 (14) 450 (10)** Total energy intake (kj) 8790 (2510) 9440 (2600)*** 7620 (2190) 8250 (2260)*** Alcohol intake (g) b 6.14 (1.55, 5.60) 8.95 (2.35, 5.60)*** 1.55 (0, 5.60) 3.46 (0.761, 5.60)*** Fruit intake a,c 658 (48) 2004 (64)*** 794 (65) 3246 (76)*** Green leafy vegetables a,c 385 (28) 1160 (37)*** 468 (38) 2088 (49)*** See Table 1. and women. Age, BMI, WHR and, in women only, total energy intake were greatest in the top quintile of HbA 1c, while alcohol intake was lowest in this quintile. A larger proportion of the least physically active people were in the top quintile of HbA 1c. Among men, the proportion of current smokers was significantly higher in the top quintile of HbA 1c, while the proportion of men consuming at least one portion of fruit was greatest in the lowest quintile of HbA 1c. In women, there was an inverse association between oily fish intake and the quintile of HbA 1c. The linear regression models (Table 6) indicated that in men there were no statistically significant associations between HbA 1c and the consumption of fish. In women, fried fish consumption adjusted only for age was positively associated with HbA 1c (b ¼ 0.036, 95% CI: , 0.069). The association was attenuated by also adjusting for family history of diabetes, smoking status and physical activity level, but remained statistically significant (b ¼ 0.033, 95% CI: , 0.066). After further adjustment for total energy, alcohol, fruit and green leafy vegetable intakes, the association between HbA 1c and fried fish intake was no longer significant. Oily fish consumption adjusted for age was strongly associated with HbA 1c in women (b ¼ 0.046, 95% CI: 0.086, ). However, the association was no longer statistically significant after adjusting for family history of diabetes, smoking status and physical activity level. Discussion This study investigated the cross-sectional association between the habitual consumption of three categories of fish and HbA 1c, in a population of men and women without selfreported diabetes. There was evidence that women who consumed at least one portion of fried had

5 Mean HbAc (%) Mean HbAc (%) Fried White Oily Oily fish intake (portions) Fried White Oily Oily fish intake (portions) Men < 1 per week >= 1 per week Women < 1 per week >= 1 per week Figure 1 Mean glycated haemoglobin level by sex and by category of fish intake: EPIC-Norfolk study (n ¼ ). *Significant difference between occasional and regular fish eaters (Po0.05). higher HbA 1c than women who ate little or no fried fish, while women who consumed at least one portion of oily fish per week had lower HbA 1c than women who ate little or no oily fish. However, these associations were subject to confounding by other lifestyle factors and were no longer statistically significant in multiple regression models adjusted for age, family history of diabetes, smoking status, physical activity, total energy, alcohol, fruit and green leafy vegetable intakes. The trends in Figure 1 and the results from the regression analyses suggested that the association between HbA 1c and fried fish consumption in men was similar to that in women, but in men the association was not statistically significant. The EPIC-Norfolk study is a population-based cohort study with characteristics similar to those from nationally representative samples (Day et al, 1999). The individuals in this analysis included all those in EPIC-Norfolk who did not report having diabetes, had HbA 1c measured and had complete data. Consequently, there is little evidence of selection bias in this population. Chance was unlikely to be an explanation for the associations observed in this large study. The major issues affecting the inferences that can be drawn from this study are likely to be confounding, and the potential for residual confounding in those variables assessed with relative imprecision. Age was an important confounding variable, but the direction of confounding differed between types of fish intake and for white fish consumption was not consistent between men and women. There was also evidence of confounding with smoking status, physical activity, obesity and the dietary variables, total energy, alcohol, fruit and green leafy vegetable intakes. The association between exposure and outcome may also be affected by measurement error. Fish consumption was assessed by FFQ. Respondents were asked to indicate how often, on average, they ate a medium serving of, for example, fried fish over the past year. The interpretation of medium serving provides scope for interindividual variation, which will result in some reporting error. If this error is random, it will tend to attenuate the relations between variables. However, there is some evidence from validation studies that obese people are more likely to under-report their dietary intake (Bingham & Nelson, 1991), and such systematic differences in reporting would introduce bias into the results. In reporting fish consumption, participants may have misclassified the type of fish eaten, which would also result in effect sizes being underestimated. Seasonal differences in the composition and total amount of fatty acids contained in fish may obscure the distinction between oily and white fish (Soriguer et al, 1997). Since white fish contains little fat compared with oily fish, these fluctuations in fat content may lead to misclassification in terms of the fatty acid content of the fish. This study highlighted the importance of distinguishing between the categories of fish. The positive association between HbA 1c and fried fish consumption may partly be ascribed to the oil used to fry the fish, and possibly to the British habit of eating fried fish with chips. Eating fried fish regularly may also be a marker for a lifestyle that is associated with higher HbA 1c. In this population, regular fried fish eaters tended to have higher BMI, WHR and total energy intake, and lower intakes of alcohol, fruit and vegetables. In men, a higher proportion of regular fried fish eaters were current smokers. Previous studies of this cohort have reported on the benefits of moderate alcohol consumption, fruit and vegetable consumption and the detrimental effect of smoking on HbA 1c (Sargeant et al, 2001a,b; Harding et al, 2002). Among women, eating oily fish was inversely associated with HbA 1c, although the effect was not independent of other lifestyle factors. Regular oily fish eaters had a lower proportion of the least physically active people and current smokers, and tended to have higher total energy, alcohol, fruit and vegetable intakes. Eating oily fish regularly may represent one component of a healthy lifestyle. However, 281

6 282 Table 4 Characteristics of 4500 men by quintiles of glycated haemoglobin (HbA 1c ): the EPIC-Norfolk Study Glycated haemoglobin Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 (4.48%) a (5.00%) a (5.30%) a (5.58%) a (6.21%) a P-value N F HbA 1c (%) 4.48 (0.336) 5.00 (0.0803) 5.30 (0.0799) 5.58 (0.0799) 6.22 (0.762) F Age (y) 55.4 (9.48) 56.7 (9.51) 58.2 (9.66) 59.3 (9.27) 61.7 (8.85) o0.001 BMI (kg/m 2 ) 26.2 (3.03) 26.2 (3.09) 26.5 (3.16) 26.4 (3.32) 27.2 (3.69) o0.001 WHR (0.0580) (0.0564) (0.0565) (0.0568) (0.0602) o0.001 Family history of diabetes b 112 (11) 103 (12) 118 (12) 102 (13) 109 (13) Physical activity level (least active) b 257 (25) 236 (28) 262 (26) 258 (33) 303 (37) o0.001 Current smokers b 82 (8) 80 (9) 111 (11) 121 (16) 147 (18) o0.001 Total energy intake (kj) 9070 (2440) 9330 (2620) 9220 (2535) 9430 (2690) 9260 (2690) Alcohol intake (g) c 9.55 (2.57, 19.0) 9.38 (2.35, 18.1) 8.17 (1.59, 16.2) 6.65 (1.59, 14.0) 6.39 (1.55, 13.5) o0.001 Fruit intake b,d 648 (62) 544 (63) 589 (59) 438 (56) 443 (54) o0.001 Green leafy vegetables b,d 340 (33) 325 (38) 330 (33) 256 (33) 294 (36) Fried fish b,e 684 (66) 532 (62) 646 (64) 511 (66) 540 (66) White fish b,e 744 (71) 629 (73) 710 (71) 572 (74) 573 (70) Oily fish b,e 719 (69) 600 (70) 686 (68) 565 (73) 572 (70) Data are means with standard deviations in parentheses, and P-values refer to a nonparametric test for trend. a Quintile mean HbA 1c. b Data are counts with percents in parentheses, and P-values refer to the w 2 test. c Data are medians with interquartile range in parentheses. d At least one portion per day. e At least one portion per week. Table 5 Characteristics of 5509 women by quintiles of glycated haemoglobin (HbA 1c ): the EPIC-Norfolk Study Glycated haemoglobin Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 (4.48) a (4.96) a (5.28) a (5.53) a (6.15) a P-value N F HbA 1c (%) 4.39 (0.319) 4.96 (0.110) 5.25 (0.0500) 5.53 (0.110) 6.16 (0.698) F Age (y) 52.7 (8.65) 55.4 (9.33) 57.0 (8.92) 59.8 (8.99) 62.8 (8.47) o0.001 BMI (kg/m 2 ) 25.1 (3.82) 25.6 (3.97) 26.0 (4.06) 26.3 (4.48) 27.5 (4.88) o0.001 WHR (0.0664) (0.0607) (0.0590) (0.0617) (0.0618) o0.001 Family history of diabetes b 145 (12) 188 (14) 104 (12) 156 (12) 140 (15) Physical activity level (least active) b 246 (21) 338 (25) 227 (27) 358 (29) 351 (38) o0.001 Current smokers b 111 (10) 143 (11) 97 (12) 140 (11) 127 (14) Total energy intake (kj) 7950 (2240) 7960 (2140) 8190 (2340) 8220 (2230) 8320 (2380) o0.001 Alcohol intake (g) c 4.67 (0.761, 10.3) 4.13 (0.793, 9.61) 2.78 (0.761, 8.01) 2.35 (0.510, 7.50) 1.52 (0, 5.73) o0.001 Fruit intake b,d 851 (73) 992 (73) 626 (74) 929 (74) 642 (70) Green leafy vegetables b,d 526 (45) 631 (47) 384 (46) 597 (48) 418 (46) Fried fish b,e 604 (52) 696 (52) 454 (54) 647 (52) 514 (56) White fish b,e 877 (76) 1041 (77) 639 (76) 980 (79) 714 (78) Oily fish b,e 913 (79) 1093 (81) 656 (78) 969 (78) 665 (73) o0.001 See Table 4. eating oily fish may also represent health awareness and could then be affected by reporting bias. A beneficial effect of oily fish consumption would be consistent with the hypothesis that the long-chain polyunsaturated fatty acids found in oily fish are the active agents in the protective effect of fish. However, these fatty acids may not be the only constituents of fish with health benefits. Establishing the population determinants of glycaemia is important, since there is evidence that the risk of cardiovascular disease increases throughout the range of HbA 1c (Singer et al, 1992; Khaw et al, 2001), and hyperglycaemia is a risk factor for type 2 diabetes (National Diabetes Data Group, 1996; Unwin et al, 1998). An expert workshop that reviewed the health effects of n-3 fatty acids concluded that eating one portion of compared with eating no fish may reduce the risk of fatal coronary heart disease by approximately 40% (de Deckere et al, 1998). In our study, fish consumption was associated with other dietary and

7 Table 6 Regression of glycated haemoglobin on fish intake: the EPIC-Norfolk Study (n=10 009) 283 Regression model Men (n=4500) Women (n=5509) Fried fish consumption Fried fish ( 0.036, 0.045) ( 0.013, 0.056) +Age ( 0.011, 0.072) (0.0033, 0.069) +Family history of diabetes ( 0.011, 0.071) (0.026, 0.068) +Smoking status, physical activity ( 0.021, 0.061) ( , 0.066) +Total energy, alcohol, fruit and vegetable intakes ( 0.028, 0.055) ( , 0.058) +BMI, WHR ( 0.039, 0.045) ( 0.021, 0.045) White fish consumption White fish ( 0.046, 0.043) ( , 0.078) +Age ( 0.076, 0.012) ( 0.049, 0.030) +Family history of diabetes ( 0.074, 0.014) ( 0.049, 0.029) +Smoking status, physical activity ( 0.063, 0.024) ( 0.038, 0.041) +Total energy, alcohol, fruit and vegetable intakes ( 0.058, 0.031) ( 0.028, 0.051) +BMI, WHR ( 0.057, 0.032) ( 0.033, 0.046) Oily fish consumption Oily fish ( 0.034, 0.054) ( 0.10, 0.019) +Age ( 0.042, 0.044) ( 0.086, ) +Family history of diabetes ( 0.042, 0.043) ( 0.086, 0.006) +Smoking status, physical activity ( 0.028, 0.057) ( 0.076, ) +Total energy, alcohol, fruit and vegetable intakes ( 0.022, 0.065) ( 0.074, ) +BMI, WHR ( 0.017, 0.070) ( 0.072, ) Data are regression coefficients for fish intake of at least one portion per week, with 95% confidence intervals in parentheses, for (i) the basic model including only fish intake, and (ii) the more complex models developed by including potential confounding variables sequentially. Intake reference category: less than one portion of. nondietary factors, and may be a marker for different lifestyles. The associations observed suggest that, among women, oily fish consumption may be beneficial for HbA 1c levels when it is part of a healthy lifestyle characterised by increased physical activity, higher levels of fruit and vegetable intake, moderate alcohol consumption and no smoking. References Adler AI, Boyko EJ, Schraer CD & Murphy NJ (1994): Lower prevalence of impaired glucose tolerance and diabetes associated with daily seal oil or salmon consumption among Alaska Natives. Diabetes Care 17, Bingham SA, Margetts BM, Gill C, Welch A, Cassidy A, Runswick SA, Oakes S, Lubin R, Thurnham DI, Key TJ, Roe L, Khaw KT & Day NE (1997): Validation of dietary assessment methods in the UK arm of EPIC using weighed records, and 24-hour urinary nitrogen and potassium and serum vitamin C and carotenoids as biomarkers. Int. J. Epidemiol. 26 (Suppl 1), S137 S151. Bingham SA & Nelson M (1991): Assessment of food consumption and nutrient intake. In Design concepts in Nutritional Epidemiology, BM Margetts & M Nelson (eds)., pp Oxford: Oxford University Press. Bjerregaard P, Pedersen HS & Mulvad G (2000): The associations of a marine diet with plasma lipids, blood glucose, blood pressure and obesity among the Inuit in Greenland. Eur. J. Clin. Nutr. 54, Borkman M, Storlien LH, Pan DA, Jenkins AB, Chisholm DJ & Campbell LV (1993): The relation between insulin sensitivity and the fatty-acid composition of skeletal-muscle phospholipids. N. Engl. J. Med. 328, Chan DC, Watts GF, Barrett PHR, Beilin LJ & Mori TA (2002): Effect of atorvastatin and fish oil on plasma high-sensitivity C-reactive protein concentrations in individuals with visceral obesity. Clin. Chem. 48, Day N, Oakes S, Luben R, Khaw KT, Bingham S, Welch A & Wareham N (1999): EPIC-Norfolk: study design and characteristics of the cohort. European Prospective Investigation of Cancer. Br. J. Cancer 80 (Suppl 1), de Deckere EA, Korver O, Verschuren PM & Katan MB (1998): Health aspects of fish and n-3 polyunsaturated fatty acids from plant and marine origin. Eur. J. Clin. Nutr. 52, Dewailly E, Blanchet C, Gingras S, Lemieux S, Sauve L, Bergeron J & Holub BJ (2001a): Relations between n-3 fatty acid status and cardiovascular disease risk factors among Quebecers. Am. J. Clin. Nutr. 74, Dewailly E, Blanchet C, Lemieux S, Sauve L, Gingras S, Ayotte P & Holub BJ (2001b): n-3 Fatty acids and cardiovascular disease risk factors among the Inuit of Nunavik. Am. J. Clin. Nutr. 74, Feskens EJ, Bowles CH & Kromhout D (1991): Inverse association between habitual fish intake and risk of glucose intolerance in normoglycemic elderly men and women. Diabetes Care 14, Grundt H, Nilsen DW, Hetland O, Aarsland T, Baksaas I, Grande T & Woie L (1995): Improvement of serum lipids and blood pressure during intervention with n-3 fatty acids was not associated with changes in insulin levels in subjects with combined hyperlipidaemia. J. Intern. Med. 237, Harding AH, Sargeant LA, Khaw KT, Welch A, Oakes S, Luben RN, Bingham S, Day NE & Wareham NJ (2002): Cross-sectional association between total level and type of alcohol consumption and glycosylated haemoglobin level: the EPIC-Norfolk Study. Eur. J. Clin. Nutr. 56, Khaw K-T, Wareham N, Luben R, Bingham S, Oakes S, Welch A & Day N (2001): Glycated haemoglobin, diabetes, and mortality in men

8 284 in Norfolk cohort of European Prospective Investigation of Cancer and Nutrition (EPIC-Norfolk). BMJ 322, Kromann N & Green A (1980): Epidemiological studies in the Upernavik District, Greenland. Acta. Med. Scand. 208, Lavigne C, Tremblay F, Asselin G, Jacques H & Marette A (2001): Prevention of skeletal muscle insulin resistance by dietary cod protein in high fat-fed rats. Am. J. Physiol. Endocrinol. Metab. 281, E62 E71. Lovegrove JA, Brooks CN, Murphy MC, Gould BJ & Williams CM (1997): Use of manufactured foods enriched with fish oils as a means of increasing long-chain n-3 polyunsaturated fatty acid intake. Br. J. Nutr. 78, Mori TA, Bao DQ, Burke V, Puddey IB, Watts GF & Beilin LJ (1999): Dietary fish as a major component of a weight-loss diet: effect on serum lipids, glucose, and insulin metabolism in overweight hypertensive subjects. Am. J. Clin. Nutr. 70, Mori TA, Burke V, Puddey IB, Watts GF, O Neal DN, Best JD & Beilin LJ (2000): Purified eicosapentaenoic and docosahexaenoic acids have differential effects on serum lipids and lipoproteins, LDL particle size, glucose, and insulin in mildly hyperlipidemic men. Am. J. Clin. Nutr. 71, National Diabetes Data Group (1996): Diabetes in America, 2nd Edition. Bethesda, MD: National Institutes of Health. National Food Survey Committee (1982): Annual Report of Household Food Consumption and Expenditure London: Her Majesty s Stationery Office. Riboli E & Kaaks R (1997): The EPIC Project: rationale and study design. European Prospective Investigation into Cancer and Nutrition. Int. J. Epidemiol. 26 (Suppl 1), S6 S14. Sargeant LA, Khaw KT, Bingham S, Day NE, Luben RN, Oakes S, Welch A & Wareham NJ (2001a): Cigarette smoking and glycaemia: the EPIC-Norfolk Study. Int. J. Epidemiol. 30, Sargeant LA, Khaw KT, Bingham S, Day NE, Luben RN, Oakes S, Welch A & Wareham NJ (2001b): Fruit and vegetable intake and population glycosylated haemoglobin levels: the EPIC-Norfolk Study. Eur. J. Clin. Nutr. 55, Singer DE, Nathan DM, Anderson KM, Wilson PW & Evans JC (1992): Association of HbA1c with prevalent cardiovascular disease in the original cohort of the Framingham Heart Study. Diabetes 41, Soriguer F, Serna S, Valverde E, Hernando J, Martin Reyes A, Soriguer M, Pareja A, Tinahones F & Esteva I (1997): Lipid, protein, and calorie content of different Atlantic and Mediterranean fish, shellfish, and molluscs commonly eaten in the south of Spain. Eur. J. Epidemiol. 13, Standing SJ & Taylor RP (1992): Glycated haemoglobin: an assessment of high capacity liquid chromatographic and immunoassay methods. Ann. Clin. Biochem. 29, Storlien LH, Baur LA, Kriketos AD, Pan DA, Cooney GJ, Jenkins AB, Calvert GD & Campbell LV (1996): Dietary fats and insulin action. Diabetologia 39, Toft I, Bonaa KH, Ingebretsen OC, Nordoy A & Jenssen T (1995): Effects of n-3 polyunsaturated fatty acids on glucose homeostasis and blood pressure in essential hypertension. A randomized, controlled trial. Ann. Intern. Med. 123, Tremblay F, Lavigne C, Jacques H & Marette A (2003): Dietary cod protein restores insulin-induced activation of phosphatidylinositol 3-kinase/Akt and GLUT4 translocation to the T-tubules in skeletal muscle of high-fat-fed obese rats. Diabetes 52, Unwin N, Alberti K, Bhopal R, Harland J, Watson W & White M (1998): Comparison of the current WHO and new ADA criteria for the diagnosis of diabetes mellitus in three ethnic groups in the UK. Diabet. Med. 15, Vessby B (2000): Dietary fat and insulin action in humans. Br. J. Nutr. 83, 1s91 1s96. Wareham NJ, Jakes RW, Rennie KL, Schuit J, Mitchell J, Hennings S & Day NE (2003): Validity and repeatability of a simple index derived from the short physical activity questionnaire used in the European Prospective Investigation into Cancer (EPIC) study. Pub. Health Nutr. (in press). Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J, Hennekens CH & Speizer FE (1985): Reproducibility and validity of a semiquantitative food frequency questionnaire. Am. J. Epidemiol. 122, Young TK, Schraer CD, Shubnikoff EV, Szathmary EJ & Nikitin YP (1992): Prevalence of diagnosed diabetes in circumpolar indigenous populations. Int. J. Epidemiol. 21,