Longitudinal studies have revealed that US and Australian mid-age women (42–55 years) gain on average between 0·2 and 1 kg per year( Reference Lee, Djousse and Sesso 1 – Reference Sternfeld, Wang and Quesenberry 3 ). The prevention of weight gain in premenopausal women has been identified as an important health goal( Reference Kremers, Reubsaet and Martens 4 ), since abdominal adiposity and cardiovascular risk increase markedly following menopause( Reference Franklin, Ploutz-Snyder and Kanaley 5 , Reference Rosano, Vitale and Marazzi 6 ).
Faster self-reported rates of eating have been shown in cross-sectional studies to be associated with excess body weight( Reference Leong, Madden and Gray 7 – Reference Sasaki, Katagiri and Tsuji 10 ), independent of self-reported energy intake( Reference Otsuka, Tamakoshi and Yatsuya 9 , Reference Sasaki, Katagiri and Tsuji 10 ). Results from our baseline survey of mid-age women( Reference Leong, Madden and Gray 7 ) showed that after adjusting for demographic, health and behavioural variables, every one-category increase in self-reported speed of eating (‘very slow’, relatively slow’, ‘medium’, ‘relatively fast’, ‘very fast’) was associated with a 2·8 % increase in BMI. Cross-sectional studies have also demonstrated positive associations between faster eating rate and CVD risk factors( Reference Lee, Kim and Jang 11 ) and insulin resistance( Reference Otsuka, Tamakoshi and Yatsuya 12 ).
Few longitudinal studies have examined the relationship between weight gain and speed of eating( Reference Gerace and George 13 – Reference Yamane, Ekuni and Mizutani 16 ). A retrospective longitudinal study of 529 male workers in Japan demonstrated that only among the 20–29 year age group, compared with the combined group of slow and medium-speed eaters, the fast-eating group had a higher average 8-year weight gain (1·9 kg v. 0·7 kg)( Reference Tanihara, Imatoh and Miyazaki 14 ). A 7-year prospective study of 438 male fire service personnel showed that those who at baseline reported faster eating rates at the fire station (compared with other locations) gained 1·4 kg more over 7 years than those who reported at baseline that their eating rate did not differ by location( Reference Gerace and George 13 ). Recently, a 3-year prospective study of 1396 healthy-weight first-year university students showed that among females and males, those who ate quickly at baseline were three times more likely to become overweight compared with those who ate slowly at baseline( Reference Yamane, Ekuni and Mizutani 16 ). Among healthy-weight children (aged 9–10 years), girls who reported eating quickly at baseline and 3 years later had significantly greater gains in BMI, waist circumference, percentage body fat and waist-to-height ratio, compared with those who were not eating quickly at baseline and at 3-year follow-up( Reference Ochiai, Shirasawa and Ohtsu 15 ).
Given the potential benefits of slower eating, the present study aimed to examine in mid-life women: (i) the associations between changes in speed of eating and weight gain over 3 years; and (ii) whether faster eating at baseline was related to healthy-weight women becoming overweight after 3 years.
Methods
Study design and participants
A cohort of 1601 women, aged 40–50 years and randomly selected from the nationwide general and Māori electoral rolls (all New Zealand residents eligible to vote), was recruited in May 2009 to participate in a prospective study. The baseline sample was reasonably representative of mid-age New Zealand women in terms of socio-economic status (as measured by the New Zealand Socioeconomic Index Score, NZSEI) and ethnicity( Reference Leong, Madden and Gray 7 , Reference Madden, Leong and Gray 17 ). Survey procedures for both baseline and 3-year surveys were similar and have been previously described( Reference Leong, Madden and Gray 7 , Reference Madden, Leong and Gray 17 ). A total of sixty-four respondents were excluded from analysis because they did not meet inclusion criteria (i.e. they were pregnant or breast-feeding, outside the age range or, in one case, male), there was reason to doubt the reliability of the answers (e.g. geometric patterns were made by circling answers, the respondent simultaneously answered opposite ends of a scale, the questionnaire was answered on behalf of someone else or the respondent indicated a poor understanding of English) or the potential responder was deceased. Of the 1318 women who were eligible to participate in the 3-year follow-up, 77·8 % (n 1025) responded in May 2012. Ten women reported having had bariatric surgery prior to the 3-year follow-up and were excluded, leaving 1015 participants for analyses.
Participants were informed that completing the questionnaires would imply informed consent.
Questionnaires
Women provided information on self-reported speed of eating, physical activity, menopause status, smoking status, and height and weight, both at baseline and 3 years. Demographic and health information was collected at baseline. Questionnaires are described in detail elsewhere( Reference Leong, Madden and Gray 7 , Reference Madden, Leong and Gray 17 ).
The speed of eating question was slightly modified from the original question ‘How fast is your rate of eating?’( Reference Sasaki, Katagiri and Tsuji 10 ) to ‘How would you describe your usual rate of eating?’ As in the study by Sasaki et al.( Reference Sasaki, Katagiri and Tsuji 10 ), participants were asked to choose one response from ‘very slow’, ‘relatively slow’, ‘medium’, ‘relatively fast’ or ‘very fast’. The original question by Sasaki et al.( Reference Sasaki, Katagiri and Tsuji 10 ) has been shown to have good repeatability over a 1-year period( Reference Maruyama, Sato and Ohira 8 ). This measure of self-reported speed of eating has shown a good level of agreement (κ=0·7) with speed of eating measured in the laboratory( Reference Petty, Melanson and Greene 18 ).
Women were classified as experiencing a ‘change in menopause status’ (premenopausal at baseline to perimenopausal at 3 years, or premenopausal at baseline to postmenopausal at 3 years) or ‘no change in menopause status’ (premenopausal at baseline and 3 years, or postmenopausal at baseline to peri-/postmenopausal at 3 years). In each survey, participants were classified as smoker or non-smoker. Their transition from baseline to 3 years was modelled into three categories: (i) ‘smoking ceased’ (smoker to non-smoker); (ii) ‘smoking initiated’ (non-smoker to smoker); and (iii) ‘smoking status remained unchanged’ (either remained non-smoker or remained smoker).
Self-reported physical activity levels were categorised into two categories: ‘underactive’ (answered ‘yes’ to either ‘I rarely or never do any physical activity’ or ‘I do some light activity every week’) and ‘active’ (answered ‘yes’ to ‘I do at least 30 min of moderate physical activity five or more days per week’ or ‘yes’ to ‘I do at least 20 min of vigorous physical activity on three or more days per week’). Using information at baseline and follow-up, four categories of change in physical activity level were modelled: (i) ‘remained active’; (ii) ‘becoming active’; (iii) ‘remained underactive’; and (iv) ‘becoming underactive’.
Participants’ height and body weight were used to compute BMI as [weight (kg)]/[height (m)]2. Healthy BMI was defined as at least 18·5 but less than 25·0 kg/m2.
Statistical analysis
As relatively few women reported eating ‘very slow’ (n 26 and 25 at baseline and 3 years, respectively) or ‘very fast’ (n 64 and 48 at baseline and 3 years, respectively), women were classified into two categories as follows: (i) ‘slower’ eaters (combining ‘very slow’, ‘relatively slow’ and ‘medium’); and (ii) ‘faster’ eaters (combining ‘relatively fast’ and ‘very fast’). This is consistent with previous studies( Reference Maruyama, Sato and Ohira 8 , Reference Yamane, Ekuni and Mizutani 16 ). For participants who completed both baseline and 3-year surveys, distributions of continuous variables were compared between the two speeds of eating categories using t tests and categorical variables were compared using χ 2 tests. Linear weighted κ statistics were used to determine the degree of concordance between self-reported speed of eating at baseline and 3 years later using all five levels listed above.
Partially adjusted and adjusted multiple linear regression models were used to examine the associations between 3-year BMI adjusting for baseline BMI and baseline speed of eating categories (‘slower’ and ‘faster’) as well as the four change in speed of eating categories (‘consistently faster eating’, ‘consistently slower eating’, ‘slower eating at baseline but not at 3 years’ and ‘faster eating at baseline but not at 3 years’).
Among a sub-sample of women whose baseline BMI was categorised as healthy, logistic regression models were used to examine the association between baseline speed of eating and BMI category change (‘remained healthy weight at 3 years’ and ‘had become overweight/obese at 3 years’), as well as between the four change in speed of eating categories and BMI category change. Where the Wald test for a categorical variable was statistically significant, post hoc pairwise comparisons between the levels of that variable were performed.
The adjusted linear regression model and the logistic regression model controlled for baseline BMI, age, NZSEI score, thyroid condition, ethnicity, change in physical activity, change in smoking status and change in menopause status.
Residuals of linear regression models were assessed for normality and homogeneity of variances. Variance inflation factors were used to check for excessive collinearity in all adjusted linear models. The Homer–Lemeshow goodness-of-fit tests were examined for all logistic regression models. Two-sided P values less than 0·05 were deemed statistically significant. Statistical analyses were conducted using the Stata statistical software package version 12·1 (2012).
Results
The geometric mean BMI of women was 25·7 (sd 1·2) kg/m2 at baseline and 25·8 (sd 1·2) kg/m2 at 3 years, with a mean 3-year weight gain of 0·9 (sd 6·1) kg. Baseline characteristics of participants according to baseline and 3-year self-reported speed of categories are presented in Table 1. At baseline and 3 years, self-reported ‘faster’ eaters were significantly heavier than ‘slower’ eaters (P<0·001 for weight, BMI and BMI category). Slower eaters were more likely to be smokers at both baseline and follow-up (both P<0·001). Other variables did not appear to be associated with speed of eating apart from socio-economic status (NZSEI) at follow-up, where faster eaters appeared to have higher NZSEI scores (P=0·005). Of the 488 healthy-weight women at baseline (BMI at least 18·5 but less than 25·0 kg/m2), seventy-seven (15·8 %) became overweight (BMI ≥25·0 kg/m2) after 3 years.
Characteristics of New Zealand female participants by baseline and 3-year self-reported speed of eating, May 2009–May 2012
NZSEI, New Zealand Socioeconomic Index Score.
Percentages may not add up to 100 % due to rounding.
* Values are expressed as n and %; percentages may not add to 100 % due to rounding.
† Multiple ethnicities could be reported and were prioritised as follows: Māori, Pacific People, Asian, Other and New Zealand Europeans. ‘Other’ and ‘New Zealand European’ were combined to form one category of classification due to small numbers in the ‘Other’ category.
‡ Values are expressed as geometric mean and geometric sd.
§ Differences in categorical variables by self-reported speed of eating assessed using χ 2 tests.
|| Differences in continuous variables by self-reported speed of eating assessed using t tests.
Table 2 describes the distribution of participants according to self-reported speed of eating at baseline and 3 years. Seventy-two per cent (n 727) of women were classified in the same category at both time points. Of the respondents, 14 % (n 149) increased speed of eating by one category and 0·3 % (n 3) increased by two categories from baseline to 3-year follow-up. Thirteen per cent (n 128) of women decreased their speed of eating by one category and 0·7 % (n 7) decreased their speed of eating by two categories from baseline to 3-year follow-up. The agreement between self-reported speed of eating at baseline and 3 years (five levels) was 92·7 % with a κ statistic of 0·7.
Distribution of New Zealand female participants according to self-reported speed of eating at baseline and at 3-year follow-up, May 2009–May 2012
Percentages may not add up to 100 % due to rounding.
Table 3 depicts the associations between each of baseline speed of eating and change in speed of eating, and each of BMI change over 3 years and BMI category change (remained healthy weight or had become overweight/obese at 3 years for baseline healthy BMI women). None of these associations were statistically significant (all P≥0·1).
Associations between each of baseline speed of eating and change in speed of eating, and each of BMI change over 3 years and BMI category change (remained healthy weight or had become overweight/obese at 3 years for baseline healthy BMI women), among New Zealand female participants, May 2009–May 2012
NZSEI, New Zealand Socioeconomic Index Score.
* Adjusted for baseline BMI, age, NZSEI score, thyroid condition, ethnicity, change in physical activity, change in smoking status and change in menopause status.
Discussion
We found no evidence of an association between changes in speed of eating and weight gain, nor between faster eating at baseline and healthy-weight women becoming overweight after 3 years. Speed of eating was stable over 3 years. These results were consistent with a recent retrospective longitudinal study among males reporting that, among the 40–49 and 50–59 year age groups, there was no evidence of an association between self-reported speed of eating (measured at follow-up) and weight gain over 8 years( Reference Tanihara, Imatoh and Miyazaki 14 ). Results from an earlier study by Sasaki et al.( Reference Sasaki, Katagiri and Tsuji 10 ) among adult men and women suggested that fast eating may be acquired in childhood and maintained up to and during adulthood. It is possible that faster eating may be more strongly associated with weight gain earlier in life( Reference Ochiai, Shirasawa and Ohtsu 15 , Reference Yamane, Ekuni and Mizutani 16 ) than during mid-life. In studies where faster eating has been linked with weight gain, excess energy intake in faster eaters presents a possible mechanism( Reference Otsuka, Tamakoshi and Yatsuya 12 ). The lack of association with weight gain in the present study may reflect that, in women of this age group, faster eating might not necessarily be associated with an increased energy intake. Faster eating may be a consequence of stress or a lack of time available for consuming meals. Further research is needed to explore contributors to weight gain in mid-life.
Despite findings suggesting that modifying speed of eating may be more beneficial for weight management early in life than in mid-life, recent cross-sectional studies among mid-aged men and women have shown that faster eating was associated with cardiovascular risk factors( Reference Lee, Kim and Jang 11 ) and insulin resistance( Reference Otsuka, Tamakoshi and Yatsuya 12 ) after controlling for BMI. If proved to be causal, interventions to reduce speed of eating might be of value even during mid-life for preventing the development of CVD and diabetes mellitus. Around mid-age, there is evidence of increasing body fatness( Reference Sowers, Zheng and Tomey 19 ) and worsening lipid profiles( Reference Matthews, Crawford and Chae 20 ) in women, with metabolic syndrome accounting for almost half of coronary events in postmenopausal women( Reference Kaaja 21 ). Factors that help to reduce cardiovascular risk and insulin resistance during this life stage may therefore be particularly valuable.
The current study presents evidence of a high degree of stability over 3 years in self-reported speed of eating among mid-age women. This may partly explain why a 1991 study found that although women’s eating speed decreased following an intervention intended to promote slower eating, relapse to faster eating was common( Reference Spiegel, Wadden and Foster 22 ).
The main strengths of the present study are the nationally representative sampling frame (New Zealand electoral rolls include 97 % of adults aged 40–49 years)( 23 ), good baseline response rate (66 %)( Reference Leong, Madden and Gray 7 ) and 3-year retention rate (79 %), and longitudinal design. A limitation is the reliance on self-reported height and weight. However, studies conducted in New Zealand( Reference Sharples, Crutchley and García 24 ) and overseas( Reference Stommel and Schoenborn 25 , Reference Dahl, Hassing and Fransson 26 ) have shown that among mid-aged women, the mean differences between self-reported and measured height and weight were small( Reference Sharples, Crutchley and García 24 – Reference Dahl, Hassing and Fransson 26 ) and stable over time( Reference Dahl, Hassing and Fransson 26 ). Nevertheless, there exists the possibility of under-reporting of body weight by obese women with faster eating( Reference Stommel and Schoenborn 25 ). A recent analysis of data from the US National Health and Nutrition Examination Survey, however, revealed that between 1988–1994 and 2005–2008, under-reporting of weight declined among obese adults leading to an improvement in the accuracy of BMI classifications based on self-reported height and weight( Reference Stommel and Osier 27 ). This may reflect changing social norms and greater acceptance of higher BMI values as ‘normal’. Another limitation of the study was that we did not control for total energy intake, which has been previously shown to have a strong positive correlation with speed of eating( Reference Maruyama, Sato and Ohira 8 , Reference Otsuka, Tamakoshi and Yatsuya 9 ).
Conclusions
Although mid-age women who eat fast are significantly heavier than slower eaters, our longitudinal results suggest that once women reach mid-age, eating rate may not significantly influence further weight change. Interventions promoting slower eating for weight management may be more effective in earlier life. However, since faster eating during mid-life has been associated with insulin resistance and cardiometabolic risk factors, independent of weight, intervention studies are required to elucidate the impact of slower eating on health outcomes.
Acknowledgements
Acknowledgements: The authors would like to thank Clara Madden, Heidi Sharples, Sara Boucher and Lindsay Bemelmans for contributing to designing and pre-testing the questionnaire and conducting the mail survey; and Fiona Hyland for data entry. Financial support: This work was supported by funding from the Department of Human Nutrition, University of Otago. Conflict of interest: None. Authorship: S.L.L., C.C.H. and A.G. had major roles in designing the questionnaire, interpreting results and writing the manuscript. S.L.L. carried out the mail survey. S.L.L. and A.G. conducted statistical analyses. C.C.H. is the Principal Investigator of the study. All authors had full access to the data and can take responsibility for the accuracy of data analyses. Ethics of human subject participation: This study was approved by the University of Otago Ethics Committee, Dunedin, New Zealand and the NgāiTahu Research Consultation Committee.
