An insufficient folate intake is common in many populations, and is a strong risk factor of neural tube defects and other congenital anomalies when it occurs in pregnant women(Reference Eichholzer, Tonz and Zimmermann1), which has led to mandatory fortification policies in several countries(Reference Eichholzer, Tonz and Zimmermann1). In adult or elderly people, a low folate intake has also been associated with an increased risk of CVD and of stroke(Reference Eichholzer, Tonz and Zimmermann1, Reference Clarke, Lewington, Sherliker and Armitage2), as well as of Alzheimer's disease and of vascular dementias(Reference Clarke, Lewington, Sherliker and Armitage2, Reference Reynolds3). Due to its involvement in one-carbon metabolism, folate appears to have a fundamental role in brain function at all ages(Reference Reynolds3). Depression, one of the most widespread psychiatric disorders in the human population, is one of the neuropsychiatric symptoms of folate deficiency in human subjects(Reference Reynolds3, Reference Bottiglieri4). Moreover, a low folate status has been often observed in depressive patients, and adjunct treatment with folate improves the patients' response to antidepressants(Reference Bottiglieri4, Reference Coppen and Bolander-Gouaille5). In population observational studies, a low folate intake or a low folate status have been found to be associated with a higher risk of depression, which suggests that an increased folate intake could reduce depression risk in the general population(Reference Bjelland, Tell, Vollset, Refsum and Ueland6–Reference Tolmunen, Hintikka, Ruusunen, Voutilainen, Tanskanen, Valkonen, Viinamaki, Kaplan and Salonen11). However, only few of these studies have a prospective design, from which one could infer a causal link between folate intake or status and the subsequent onset of depression. In the present study, we have used the data of the French SU.VI.MAX cohort to look for the association between the level of folate intake, measured at the beginning of the follow-up, and the onset of depressive episode during the follow-up.
Subjects and methods
The SU.VI.MAX study
The SU.VI.MAX study is a randomised, double-blind, placebo-controlled primary prevention trial designed to test the effects of daily supplementation with a mixture of vitamin C, vitamin E, β-carotene, zinc and selenium on the incidence of ischaemic heart diseases and of cancers in a population of adult men and women(Reference Hercberg, Galan, Preziosi, Bertrais, Mennen, Malvy, Roussel, Favier and Briancon12). Another objective of the study was to evaluate food consumption in a national sample of middle-aged subjects from all over France. The cohort consisted of 13 017 subjects (5141 men and 7876 women) who were included in 1994–1995 with a planned follow-up of 8 years. Men were aged 45–60 and women 35–60 at enrolment. Details on study design, recruitment, and baseline characteristics of the subjects have been reported previously(Reference Hercberg, Galan, Preziosi, Bertrais, Mennen, Malvy, Roussel, Favier and Briancon12). All subjects gave their informed written consent to the study. The study was approved by ad hoc ethical committees, i.e. the ‘Comité consultatif de protection des personnes dans la recherche biomédicale’ (CCPPRB no. 706, Cochin Hospital, Paris, France), and the ‘Commission nationale de l'informatique et des libertés’ (CNIL no. 334641).
Definition of controls and cases
In the baseline questionnaire, the subjects had to mention if they had a history of depressive episodes. At baseline and every month during the follow-up, the subjects had to report health/disease events, in particular each consultation with a health professional, with the mention of the drugs prescribed. Among the subjects for whom complete prescription data were available for the whole study follow-up, including at baseline (8 years, until 2002), we selected: all subjects having declared at least one antidepressant or lithium prescription during the follow-up (cases, n 664); all subjects with no declarations of antidepressant or lithium prescription and with no depression history (controls, n 3084), i.e. a total of 3748 subjects. Within cases, subjects having declared one or more than one (two to five) antidepressant or lithium prescriptions during follow-up were distinguished. We used antidepressant or lithium prescriptions as markers of depressive episodes, antidepressants being defined as drugs classed as antidepressants in the 1994–2002 editions of the Vidal French drug repertory.
Dietary assessment
At inclusion and every 2 months until the end of the study, subjects were asked to complete a 24 h dietary record questionnaire, i.e. a total of six questionnaires per year. A large choice of foods and drinks (about 900 items) was displayed for each of three meals (breakfast, lunch and dinner) and of four other food-intake occurrences. For each food or drink mentioned, the subjects were asked to indicate the portion size consumed. They were helped by an instruction manual given to them at the start of the study, including photographs of three portion sizes of 236 foods and drinks, and the use of which has been validated. Six records had been shown to be sufficient to estimate the individual intakes of macronutrients with a good accuracy(Reference Mennen, Bertrais, Galan, Arnault, Potier de Courcy and Hercberg13). Among the 3748 subjects (1509 men and 2239 women) initially selected, we retained those having reported at least six 24 h dietary records within a period of 2 years from inclusion: 1864 subjects, 809 men and 1055 women, on whom this study is based. Other details on questionnaires have been published previously(Reference Hercberg, Galan, Preziosi, Bertrais, Mennen, Malvy, Roussel, Favier and Briancon12, Reference Astorg, Arnault, Czernichow, Noisette, Galan and Hercberg14). A food composition table adapted to the analysis of the dietary data collected in the SU.VI.MAX study was developed on the basis of existing data(Reference Arnault, Astorg, Aubert and Bertrais15).
Statistical analyses
Statistical analyses were performed using SAS software (SAS Institute, Cary, NC, USA). Frequencies were compared by the Fisher's exact test. Means were compared using the t test. Logistic regression was used to calculate OR for the 2nd and 3rd tertiles of folate intake relative to the first one, their 95 % CI and the test of linear trend, separately for men and women. Compared to linear regression, which could also be used in this design, logistic regression is better adapted to detect always possible non-linear relations, and can also test for linearity. In addition, it has the advantage of yielding OR estimates, representing variations of the risk according to the independent (nutritional) variable. Age, intervention group (intervention or placebo), marital status (single or living in couple), education level (three levels), socio-professional category (five categories) and total energy intake were included in the model as adjustment variables. Since depression history is a determinant of later depression, we also carried out the same analyses keeping only the cases having declared no depression history: sixty-one men instead of seventy-eight (78 %) and 192 women instead of 226 (85 %) (control subjects were initially selected with no depression history).
Results
Some characteristics of cases and controls are shown in Table 1, where cases with either one or more than one antidepressant prescription (Nad) during follow-up are distinguished. Women having declared one antidepressant prescription (Nad = 1) were living alone more often than controls (P = 0·003), but there was no significant difference in marital status between cases and controls among men. Men having declared two or more antidepressant prescriptions (Nad >1) had a decreased folate intake (P = 0·0005). Within each sex, age at baseline, education level, total energy intake and alcohol intake did not differ between cases and controls. Table 2 presents the OR of antidepressant prescription according to tertiles of folate intake, adjusted for age, intervention group, education level, marital status, socio-professional category, and total energy intake (OR1). No significant association was observed between folate intake and the risk of any depressive episode (Nad ≥ 1) or of a single depressive episode (Nad = 1) during the follow-up, in both men and women. In contrast, the risk of recurrent (two or more) depressive episodes during follow-up (Nad >1) was significantly and strongly decreased in men with increasing folate intake, with a 75 % reduced risk in the 3rd tertile of intake relative to the first one (OR1 = 0·25 (95 % CI 0·06, 0·98), P for trend 0·046). When the analysis was restricted to the cases with no depression history, the OR values (OR2) were essentially unaltered, but the association found in men between recurrent depression and folate intake was no longer significant (OR2 = 0·24 (95 % CI 0·05, 1·25) for the 3rd tertile v. the 1st one, P for trend = 0·10). This association between folate intake and recurrent depression was not observed in women.
Table 1 Characteristics of controls and cases
Nad, number of antidepressant prescriptions.
* Mean value was significantly different from controls (t test) P = 0·0005.
† Mean value was significantly different from controls (Fisher's exact test) P = 0·003.
Table 2 Association between occurrence of depressive episodes (marked by number of antidepressant prescriptions (Nad)) during follow-up and folate intake: OR by tertiles of folate intake
OR1, OR adjusted for age, intervention group (vitamin/mineral supplement or placebo), marital status (living alone or in couple), education level (three levels), socio-professional category (five categories) and total energy intake; OR2, OR for cases restricted to subjects with no depression history, adjusted as in OR1 model.
Discussion
In this study, we have found that a low folate intake was associated with an increased risk of recurrent depression, but not of single-episode depression, in men. This association was not found in women. The strengths of this study are the detailed dietary assessment at the beginning of the follow-up, and the recording of depressive episodes during the 8-year follow-up, resulting in a quasi-prospective design. The use of six 24 h records is an adapted method for the estimation of folate intake, provided the questionnaire is sufficiently detailed. Recent works have measured individual folate intake using only two or three 24 h recalls(Reference Han, Yon and Hyun16–Reference Verkleij-Hagoort, de Vries, Stegers, Lindemans, Ursem and Steegers-Theunissen18). In this study, the six 24 h records allow considerable reduction of the within-subject variance(Reference Mennen, Bertrais, Galan, Arnault, Potier de Courcy and Hercberg13), and the questionnaire with more than 900 items (and a recent table adapted to this questionnaire(Reference Arnault, Astorg, Aubert and Bertrais15)) reasonably covers all folate sources. The six 24 h records method has already been used for folate intake estimation in the SU.VI.MAX study(Reference Mennen, de Courcy and Guilland19). The weaknesses of this study are the selection of a sample from the initial population, a lack of power, and the ‘soft’ nature of the endpoint used. In our final sample, the 8-year prevalence of antidepressant prescription is 10 % in men and 21 % in women. These values are consistent with prevalence values of major depression in a lifetime (11–12 % in men and 22–25 % in women)(Reference Kovess20) and of antidepressant use for 1 year (4 % in men and 8 % in women)(Reference Gasquet, Nègre-Pagès, Fourrier, Nachbaur, El Hasnaoui, Kovess and Lépine21) found in the French adult population. Moreover, the values in our sample have not been substantially modified by the selection of subjects with six 24 h dietary records (about 50 % of the initial sample). Therefore, the sample selection did not appear to introduce a significant selection bias by excluding preferentially subjects having had depressive episodes. Antidepressant prescription is an imperfect proxy for depressive episodes since a significant proportion of depressed subjects does not seek treatment(Reference Lépine, Gastpar, Mendlewicz and Tylee22) and thus many cases are missed. However, this proportion is expected to be lesser in the SU.VI.MAX cohort than in the general population, because the volunteers are likely to be more health-conscious(Reference Hercberg, Galan, Preziosi, Bertrais, Mennen, Malvy, Roussel, Favier and Briancon12, Reference Astorg, Arnault, Czernichow, Noisette, Galan and Hercberg14). This is still more true for the sample used, consisting of the most compliant volunteers. In a cohort not designed to detect depression, antidepressant prescription is thus a useful marker of depressive episodes. It has the advantage on cross-sectional assessment of depressive symptoms of recording depressive episodes with time, not only current depression. Moreover, the misclassification bias that unavoidably results from the use of such a proxy would always lead to weaken an existing association, not to create it or to strengthen it artificially. The association found in men with recurrent episodes was not attenuated by adjustment for marital status, education level or socio-professional category. When the analysis was restricted to the cases with no depression history, the OR value remained identical, but it was no longer significant (although the linear trend is not far from significance (P = 0·10)), suggesting a loss of power. With a reserve on the latter point due to the lack of significance, these results overall suggest that the association found between folate intake and recurrent depression in men was not due to confounding by these known risk factors of depression.
Among the case–control or cross-sectional studies which have looked for an association between folate intake or status and depression or depressive symptoms in adult populations, some failed to find an association(Reference Penninx, Guralnik, Ferrucci, Fried, Allen and Stabler23, Reference Tiemeier, van Tuijl, Hofman, Meijer, Kiliaan and Breteler24), but others found lower folate intake or status in depressed subjects(Reference Bjelland, Tell, Vollset, Refsum and Ueland6–Reference Tolmunen, Voutilainen, Hintikka, Rissanen, Tanskanen, Viinamaki, Kaplan and Salonen10), this association being found only in women in some studies(Reference Bjelland, Tell, Vollset, Refsum and Ueland6, Reference Ramos, Allen, Haan, Green and Miller8). A prospective study in Japanese pregnant women did not find any association between folate intake in pregnancy and the risk of post-partum depression(Reference Miyake, Sasaki, Tanaka, Yokoyama, Ohya, Fukushima, Saito, Ohfuji, Kiyohara and Hirota25). In a cohort of Finnish men where depressive symptoms have been assessed cross-sectionally, men in the lowest tertile of folate intake had an increased risk of depression (OR = 1·67 relative to men in the highest tertile)(Reference Tolmunen, Voutilainen, Hintikka, Rissanen, Tanskanen, Viinamaki, Kaplan and Salonen10). In the same cohort, men with a baseline folate intake below the median have a three times greater risk of being hospitalised for depression during the 13-year follow-up than men with a folate intake above the median(Reference Tolmunen, Hintikka, Ruusunen, Voutilainen, Tanskanen, Valkonen, Viinamaki, Kaplan and Salonen11): the association with folate intake appears much stronger for later occurrence of severe depression than for current depressive symptoms. In agreement with the study by Tolmunen et al. (Reference Tolmunen, Hintikka, Ruusunen, Voutilainen, Tanskanen, Valkonen, Viinamaki, Kaplan and Salonen11), this study shows a strongly increased risk of recurrent depression in men with a low folate intake, whereas the risk of experiencing only one depressive episode was unchanged. Both studies suggest that a low folate intake is a risk factor mainly for severe or recurrent depression in men. There is no evident explanation for this observation. A possible hypothesis is that socio-demographic factors such as marital status and life events, strong determinants of first-episode depression, are not so, or much less, for recurrent depression, which tends to become independent of psycho-social circumstances(Reference Burcusa and Iacono26). Biological factors could be more important for recurrences, including genetic factors, as also external influences such as diet and drugs. This difference between single-episode and recurrent depression could also explain why depressive symptoms measured cross-sectionally have not always been found to be associated with folate intake: only current symptoms are measured, not past episodes. Depression often reduces food intake or changes food choices, and it is possible that the association observed in cross-sectional studies between current depression and folate intake is, at least in part, a consequence of the effects of the depressive state on eating behaviour. However, this seems less likely in our study, and in another prospective study(Reference Tolmunen, Hintikka, Ruusunen, Voutilainen, Tanskanen, Valkonen, Viinamaki, Kaplan and Salonen11), since food intake has been measured at the beginning of the follow-up, whereas depressive episodes were recorded throughout the whole follow-up. Moreover, the association we have observed was adjusted for total energy intake, thus taking food intake differences into account.
In cross-sectional studies, a high homocysteine plasma level has often been associated with depressive symptoms(Reference Bjelland, Tell, Vollset, Refsum and Ueland6, Reference Sachdev, Parslow, Lux, Salonikas, Wen, Naidoo, Christensen and Jorm9, Reference Tolmunen, Hintikka, Voutilainen, Ruusunen, Alfthan, Nyyssonen, Viinamaki, Kaplan and Salonen27), and homocysteine level has been found to increase with depression score in men(Reference Sachdev, Parslow, Lux, Salonikas, Wen, Naidoo, Christensen and Jorm9). In the SU.VI.MAX cohort, women have on average a higher folate status than men, and lower plasma levels of homocysteine(Reference Chango and Potier de Courcy28). Moreover, homocysteine level has been found to be inversely related to folate intake or status in men, but not in women(Reference Mennen, de Courcy and Guilland19). Together, these results would suggest that the increased risk of depression due to a low folate intake could be mediated by an increased level of plasma homocysteine in men. This would also explain why the association between folate intake and recurrent depression has been observed only in men. However, this hypothesis is not in agreement with studies which found, on the contrary, an association of depressive symptoms with a low folate plasma level in women, but not in men(Reference Bjelland, Tell, Vollset, Refsum and Ueland6, Reference Ramos, Allen, Haan, Green and Miller8). Considering that depression rate is about twice as high in women as in men, further studies are needed to clarify the relations between folate, homocysteine, gender and depression in populations.
In summary, this study suggests that a low folate intake increases the risk of recurrent depression, but not of single-episode depression, in men. This association was not found in women. This study and other studies also suggest that recurrence and gravity are items of importance in the research of nutritional risk factors of depression, and that prospective studies allowing the recording of depressive episodes are especially needed for this purpose.
Acknowledgements
Authors' contributions are acknowledged as follows: P. A. conceived the study and wrote the first draft of the manuscript. N. A., S. B. and P. M. assessed the food intake and drug prescription data. A. C., S. B. and P. A. performed the statistical analyses. G. P. C. helped with interpretation of the results. S. H. and P. G. managed the SU.VI.MAX study. All authors contributed to the final version of the manuscript.
Financial support was provided by the Direction Générale de la Santé (Ministère de la Santé, de la Jeunesse et des Sports). None of the authors had any conflict of interest.
