The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Weight Change in Adult Life and Health Outcomes
Article first published online: 6 SEP 2012
2005 North American Association for the Study of Obesity (NAASO)
Volume 13, Issue 10, pages 1784–1792, October 2005
How to Cite
Elliott, A. M., Aucott, L. S., Hannaford, P. C. and Smith, W. C. (2005), Weight Change in Adult Life and Health Outcomes. Obesity Research, 13: 1784–1792. doi: 10.1038/oby.2005.217
- Issue published online: 6 SEP 2012
- Article first published online: 6 SEP 2012
- Received for review May 25, 2004; Accepted in final form July 12, 2005
- weight gain;
- weight loss;
- cancer incidence;
- Top of page
- Research Methods and Procedures
Objective: To investigate the relationship between weight change in adult life and subsequent mortality and cancer incidence in women.
Research Methods and Procedures: In 1994 to 1995, all women (age range, 42 to 81) still under general practitioner observation in the United Kingdom's Royal College of General Practitioners Oral Contraception Study (n = 12 303) were sent a health survey asking about health and lifestyle issues, including current weight and weight at age 30. The main outcome measures were 6-year all-cause mortality and cancer incidence among different weight change deciles. Cox regression was used to calculate hazard ratios that were adjusted for: social class at recruitment, BMI at age 30, and age group, parity, smoking status, and hormone replacement therapy status in 1995.
Results: Women who had been obese at age 30 were more likely to die and significantly more likely to develop cancer in the 6 years after the health survey than non-obese respondents. Women reporting weight gains between age 30 and 1995 were significantly less likely to die during the 6 years after the health survey than those with a stable weight, whereas those with weight loss did not fare any better than those in the stable-weight group.
Discussion: Although obesity at young age was associated with subsequent mortality and cancer incidence, weight gain over a time period of 12 to 51 years appeared to be beneficial when compared with women with stable weight over the same time period. Further research is needed to confirm or refute our findings and to allow detailed examination of potential explanations for them.
- Top of page
- Research Methods and Procedures
The World Health Organization has described the health problem of excess weight and obesity as a pandemic affecting over one-half of the adult population in many countries (1). The 2002 Health Survey for England indicated that 43% of men and 34% of women were overweight (BMI between 25 and 30 kg/m2), and 22% of men and 23% of women were obese (BMI ≥ 30 kg/m2) (2). The average percentage increase per annum in the prevalence of obesity in the United Kingdom between 1990 and 1998 was estimated at 0.74% (3), and current secular trends mean that the prevalence of obesity in the United Kingdom is likely to continue to increase.
Links between excess weight and adverse health outcomes have been well documented in prospective cohort studies, typically indicating U- or J-shaped relationships between BMI and adverse outcomes such as mortality (4, 5, 6, 7). Obesity induces multiple metabolic abnormalities that contribute to the pathogenesis of diabetes mellitus and cardiovascular disease and is associated with increased morbidity and mortality risk (8). The Framingham Heart Study showed increases in cardiovascular disease, particularly coronary heart disease, in overweight men and women (9).
The effect of weight change on health outcomes and, in particular, the long-term implications of weight change on mortality have not been so rigorously investigated. Although there is a large literature on weight loss and health benefits in obese populations (10, 11, 12, 13, 14, 15, 16, 17, 18, 19), the effect of weight change on mortality in this population is uncertain (12, 13, 14, 15, 16, 17, 18), although Christou et al. (19) claim some benefits of weight reduction on mortality. The evidence for the long-term implications of weight change on mortality in the general population is also unclear (7, 20, 21, 22, 23, 24, 25). Manson et al. (7) found that a weight gain of ≥10 kg increased mortality in middle adulthood. In a Danish study (20), both weight level and weight change had independent effects on total mortality, with both of these associations being U-shaped. A British study (21, 22, 23) found some evidence that weight loss and weight cycling had a detrimental effect on mortality for certain subgroups but reported that this was probably due to disadvantageous lifestyle and preexisting disease rather than weight change. Maru et al. (24) reported that mortality from all causes, cancer, and cardiovascular disease did not increase significantly in women with a weight gain of ≥15%. Weight loss was not significantly related to mortality either. Deyer et al. (25) found that weight loss and weight gain were significantly related to 15-year mortality but that there was no relationship for 16- to 25-year mortality.
Further work to investigate the effect of weight change on mortality and other health outcomes is required. A health survey of participants in the United Kingdom's Royal College of General Practitioners (RCGP)1 Oral Contraception Study (OCS) (26) presented an opportunity to investigate the relationship between weight change in adult life and subsequent mortality and cancer incidence in women. In particular, we wanted to assess the concept that weight gain is detrimental and that weight loss (among those suffering with obesity) is beneficial.
Research Methods and Procedures
- Top of page
- Research Methods and Procedures
Design and Sample
This was a prospective cohort study conducted within the RCGP OCS (26). Between November 1994 and July 1995, the practices of the 12, 303 women (26.5% of the original cohort) still under general practitioner observation in the OCS were asked to forward a health survey questionnaire to each woman on our behalf. Most agreed to do so, dispatching questionnaires to 11, 797 women (Figure 1). The questionnaires were sent through participating general practitioner practices because identifying information about the recruited women was not available to the organizers of the main study. To maintain patient confidentiality, each woman was assigned a unique study number, the key to which was held by the family doctor. Women receiving the questionnaire were asked to complete and return it directly to the main study office. Second and third mailings were sent, again through the general practitioner, if the questionnaire was not returned. By comparing key information already held in the OCS database with that given on the returned forms, we were able to check that the general practitioners had sent the questionnaire to the correct woman. The questionnaire asked about a number of health and lifestyle issues, including height, current weight (i.e., weight in 1995), weight at age 30 (as an approximate measure of weight at recruitment), lifetime smoking habits, and general health.
Death and Cancer Registration Notifications
In the mid 1970s, women still in the OCS were flagged at the National Health Service Central Registries in Southport and Edinburgh. Thus, all of the women in the health survey were flagged. At regular intervals, the Central Registries send notification of any deaths and/or cancer registrations occurring in flagged women. The notifications include each woman's unique study number, enabling this new information to be linked with the existing database while maintaining confidentiality. Death and cancer notifications were coded using the International Classification of Diseases (ICD), version 8 (27). This information enabled us to determine which women died or developed cancer after the health survey.
The women were divided into roughly equal sized deciles using self-reported weight change in kilograms between age 30 and 1995. Group 1 included only women who reported a weight loss (−116.58 to −1.81 kg); Group 2 included women with stable weight (−1.81 to +1.36 kg), and Groups 3 to 10 included women with progressively larger weight gains (1.36 to 66.68 kg). Weight change deciles were based on exact weight change (i.e., weight change in kilograms between age 30 and 1995) because it is intuitively easier to understand than relative weight change. However, weight change deciles based on relative weight change (i.e., the percentage weight change between age 30 and 1995) and average annual weight change (i.e., weight change between age 30 and 1995 divided by the number of years in that time period) were also analyzed to examine any differences in trends or effect.
The groups were compared with respect to key sociodemographic characteristics held in the OCS database: social class at recruitment; BMI at age 30 (an approximate measure of BMI at recruitment); and age, parity, oral contraceptive (OC) status, smoking status, and hormone replacement therapy (HRT) status in 1995. Differences between groups were investigated using χ2 tests.
Cox regression was used to calculate unadjusted hazard ratios for 6-year all-cause mortality and their associated 95% confidence intervals (CIs) and p values for a two-sided test for each decile, using the stable-weight group (Decile 2) as the referent group. The hazard ratios were then adjusted for: social class at recruitment, BMI at age 30, and age group, parity, smoking status, and HRT status in 1995. Hazard ratios were also calculated for each decile using data about all cancers (ICD8 140-208) excluding skin cancers (ICD8 172-173), hormone-related cancers [breast (ICD8 174), uterus (ICD8 182), and ovary (ICD8 183)], and all other non-hormone-related cancers. Death and cancer notifications were still being supplied to the study unit at the time of analysis; however, due to delays in the registering of cancers and delays in the notifications of deaths and cancers to the unit, only events occurring up to the end of 2001 that had been notified to us were included in the analysis. The analyses were carried out using SPSS for Windows version 11 (SPSS Inc. 2001, Chicago, IL).
- Top of page
- Research Methods and Procedures
Of the 11, 797 women who were mailed the health survey, 10, 073 returned a completed questionnaire, giving an overall corrected response rate of 85.4% (Figure 1). The age range of women in 1995 was 42 to 81, and the mean age was 56.
Based on self-reported weight at age 30, 16.6% of women were classified as overweight at recruitment (BMI between 25 and 30 kg/m2) and 4.5% as obese (BMI ≥ 30 kg/m2). Women who were obese at age 30 were more likely to die and significantly more likely to develop cancer (all cancers and hormone-dependent cancers) in the 6 years after the health survey than non-obese women at this age (Table 1). Based on self-reported weight in 1994 to 1995, 34.6% of the women were overweight at the time of the health survey, and 14.8% were obese. There was no significant association between being obese in 1994 to 1995 and subsequent all-cause mortality or cancer (Table 1).
|All-cause mortality||Incidence of all cancers||Incidence of hormone-related cancer (breast, uterus, ovary)||Incidence of non-hormone-related cancer|
|All women||Hazard ratio (95% CI)||p||Hazard ratio (95% CI)||p||Hazard ratio (95% CI)||p||Hazard ratio (95% CI)||p|
|BMI at age 30|
|≥30 kg/m2||447||1.48 (0.98 to 2.24)||0.063||1.93 (1.12 to 3.34)||0.018||2.52 (1.26 to 5.02)||0.009||1.36 (0.55 to 3.36)||0.503|
|BMI in 1994 to 1995|
|≥30 kg/m2||1469||0.98 (0.74 to 1.31)||0.911||1.22 (0.82 to 1.81)||0.326||1.53 (0.91 to 2.57)||0.111||0.94 (0.51 to 1.72)||0.829|
Weight change between age 30 and 1995 could be calculated for 9918 (98%) of all respondents and 100% of obese women. Women in the various actual weight change deciles differed significantly in a number of sociodemographic characteristics (Table 2). Proportionally more women in the manual group were in the lowest decile (lost weight) and highest decile (gained the largest amount of weight) compared with women in the non-manual group. Similarly, proportionally more women who were overweight or obese at recruitment were in the lowest and highest deciles than women with a BMI < 25. Those under 50 years of age in 1995 were least likely to be in the highest deciles (large weight gains). Those with no children in 1995 were less likely to have large weight gains, and those with more than four children tended to have large weight gains. Current smokers were more likely to be in the lower deciles, and proportionally more women who had never smoked or were ex-smokers were in the higher deciles. Proportionally more women who had never used HRT in 1995 were in the lowest and highest deciles compared with women who had ever used HRT preparations. None of these differences was unexpected.
|Actual weight change deciles(weight change in kg)†|
|Sociodemographic characteristics||n||1 (−116.58 to −1.81)||2 (−1.81 to 1.36)||3 (1.36 to 3.18)||4 (3.18 to 4.54)||5 (4.54 to 6.35)||6 (6.35 to 7.71)||7 (7.71 to 9.53)||8 (9.53 to 12.70)||9 (12.70 to 16.78)||10 (16.78 to 66.68)|
|Social class at recruitment*|
|BMI at age 30*|
|25 to 30||1633||19.5||6.8||4.5||6.1||7.8||9.6||9.3||8.0||15.4||13.0|
|Age group in 1995*|
|50 to 54||2692||9.2||10.5||10.1||8.2||10.1||13.5||10.1||9.2||9.7||9.4|
|55 to 59||2538||8.7||9.0||10.0||7.4||10.5||12.2||9.3||8.9||12.1||11.9|
|60 to 64||1815||10.5||8.4||8.7||6.2||8.5||12.7||10.0||11.0||12.2||11.9|
|Parity in 1995*|
|1 to 3||5344||9.9||11.0||11.3||8.1||10.0||13.2||9.2||9.0||9.9||8.5|
|Smoking status in 1995*|
|Current smoker (1 to 14 per day)||727||12.9||15.3||11.6||9.4||10.3||11.4||7.7||6.7||8.1||6.6|
|Current smoker (15+ per day)||1275||14.1||15.3||10.7||8.0||8.5||12.4||7.6||6.6||9.6||7.2|
|OC status in 1995|
|HRT status in 1995*|
A total of 379 deaths had occurred in the 9918 women for whom weight change data were available in the 6 years after the health survey. The unadjusted and adjusted hazard ratios for 6-year all-cause mortality in each actual weight change decile are shown in Table 3. Women in a number of the deciles reporting weight gain were significantly less likely to die during the 6 years after the postal health survey than those in decile 2 (the stable-weight group). Women in the first decile (weight losers) did not differ significantly from those in the weight-stable group (decile 2) and appeared to be closer to this referent group than any of the other groups. To eliminate bias from the possible effects of illness-related weight loss, we ran an additional Cox regression analysis excluding women who died in the first 2 years after completing the health survey. Similar results to those in the full analysis were found (data not shown). Analyses of the data using deciles of relative weight change and average annual weight change were also conducted to look for any differences in trends or effects. Both analyses showed a very similar pattern of results, with only slight variation in the risk estimates (Table 4).
|Actual weight change deciles‡ (weight change in kg)||All women (n)||Deaths (n)||Hazard ratio (95% CI)||p||Hazard ratio (95% CI)||p|
|1 (−116.58 to −1.81)||964||59||1.10 (0.77 to 1.59)||0.600||0.96 (0.65 to 1.43)||0.855|
|2 (−1.81 to 1.36)†||1042||57||1.00||1.00|
|3 (1.36 to 3.18)||1007||26||0.46 (0.29 to 0.74)||0.001||0.52 (0.33 to 0.83)||0.006|
|4 (3.18 to 4.54)||770||23||0.54 (0.33 to 0.88)||0.013||0.61 (0.37 to 0.99)||0.045|
|5 (4.54 to 6.35)||972||29||0.54 (0.34 to 0.84)||0.006||0.61 (0.39 to 0.96)||0.033|
|6 (6.35 to 7.71)||1248||36||0.52 (0.34 to 0.79)||0.002||0.50 (0.32 to 0.77)||0.002|
|7 (7.71 to 9.53)||929||31||0.60 (0.39 to 0.93)||0.023||0.59 (0.37 to 0.92)||0.021|
|8 (9.53 to 12.70)||911||35||0.64 (0.41 to 0.98)||0.041||0.63 (0.40 to 0.99)||0.045|
|9 (12.70 to 16.78)||1071||45||0.75 (0.50 to 1.10)||0.142||0.70 (0.46 to 1.05)||0.085|
|10 (16.78 to 66.68)||1004||38||0.67 (0.44 to 1.01)||0.054||0.61 (0.39 to 0.95)||0.028|
|Weight change deciles‡||All women (n)||Deaths (n)||Hazard ratio (95% CI)||p||Hazard ratio (95% CI)||p|
|Relative weight change deciles (% weight change)|
|1 (−57.14 to −3.01)||988||59||1.04 (0.72 to 1.50)||0.831||0.91 (0.61 to 1.35)||0.637|
|2 (−3.01 to 2.38)†||992||56||1.00||1.00|
|3 (2.38 to 5.41)||993||24||0.42 (0.26 to 0.68)||<0.001||0.50 (0.31 to 0.80)||0.004|
|4 (5.41 to 7.94)||995||29||0.51 (0.32 to 0.80)||0.003||0.56 (0.35 to 0.88)||0.011|
|5 (7.94 to 10.53)||1053||28||0.46 (0.29 to 0.73)||0.001||0.52 (0.33 to 0.82)||0.005|
|6 (10.53 to 12.86)||911||33||0.63 (0.41 to 0.97)||0.038||0.60 (0.38 to 0.95)||0.029|
|7 (12.86 to 16.22)||1015||28||0.48 (0.31 to 0.76)||0.002||0.50 (0.32 to 0.80)||0.004|
|8 (16.22 to 20.79)||988||41||0.69 (0.46 to 1.04)||0.079||0.70 (0.46 to 1.07)||0.700|
|9 (20.79 to 28.38)||993||43||0.72 (0.48 to 1.08)||0.116||0.68 (0.45 to 1.03)||0.065|
|10 (28.38 to 116.67)||990||38||0.65 (0.43 to 0.99)||0.045||0.56 (0.36 to 0.88)||0.011|
|Average annual weight change deciles‡ (weight change in kg)|
|1 (−4.66 to −0.07)||986||57||0.96 (0.66 to 1.38)||0.810||0.99 (0.67 to 1.48)||0.978|
|2 (−0.07 to 0.05)†||996||59||1.00||1.00|
|3 (0.05 to 0.13)||988||35||0.59 (0.39 to 0.90)||0.013||0.61 (0.40 to 0.93)||0.021|
|4 (0.13 to 0.18)||998||34||0.57 (0.37 to 0.86)||0.008||0.64 (0.41 to 0.98)||0.040|
|5 (0.18 to 0.24)||991||37||0.61 (0.40 to 0.92)||0.018||0.63 (0.41 to 0.97)||0.036|
|6 (0.24 to 0.30)||974||27||0.44 (0.28 to 0.70)||0.001||0.47 (0.29 to 0.76)||0.002|
|7 (0.30 to 0.38)||1004||29||0.45 (0.28 to 0.70)||0.001||0.47 (0.29 to 0.76)||0.002|
|8 (0.38 to 0.49)||998||41||0.68 (0.46 to 1.02)||0.062||0.84 (0.55 to 1.27)||0.407|
|9 (0.49 to 0.68)||985||34||0.57 (0.38 to 0.88)||0.010||0.76 (0.49 to 1.18)||0.217|
|10 (0.68 to 3.52)||998||26||0.41 (0.26 to 0.66)||<0.001||0.56 (0.33 to 0.92)||0.023|
A total of 175 cancer registrations had occurred in the 9918 women for whom weight change data were available in the 6 years after the health survey (including 62 breast, 13 uterus, 12 ovary, 12 lung, and 15 colon). Table 5 presents the adjusted hazard ratios for all cancers, hormone-related cancers, and all other cancers in each decile. There was a general trend of lower odds ratios with moderate weight gain. However, the risk estimates were less precise than for all-cause mortality because they were based on smaller numbers (and so had wider CIs), and none of the risk estimates or trends reached statistical significance.
|All cancers||Hormone-related cancer (breast, uterus, ovary)||Non-hormone-related cancer|
|Weight change deciles‡ (weight change in kg)||All women (n)||n||Hazard ratio (95% CI)*||p||n||Hazard ratio (95% CI)*||p||n||Hazard ratio (95% CI)*||p|
|1 (−116.58 to −1.81)||964||22||1.08 (0.57 to 2.05)||0.818||13||1.51 (0.60 to 3.80)||0.379||9||0.77 (0.31 to 1.92)||0.575|
|2 (−1.81 to 1.36)†||1042||20||1.00||9||1.00||11||1.00|
|3 (1.36 to 3.18)||1007||15||0.79 (0.40 to 1.58)||0.504||10||1.20 (0.46 to 3.13)||0.705||5||0.49 (0.17 to 1.42)||0.190|
|4 (3.18 to 4.54)||770||19||1.17 (0.60 to 2.28)||0.640||8||0.85 (0.28 to 2.60)||0.775||11||1.43 (0.62 to 3.31)||0.400|
|5 (4.54 to 6.35)||972||9||0.52 (0.23 to 1.15)||0.107||5||0.66 (0.21 to 2.02)||0.464||4||0.42 (0.13 to 1.31)||0.134|
|6 (6.35 to 7.71)||1248||16||0.58 (0.29 to 1.18)||0.133||7||0.63 (0.22 to 1.83)||0.400||9||0.55 (0.21 to 1.41)||0.212|
|7 (7.71 to 9.53)||929||16||0.95 (0.49 to 1.85)||0.880||9||1.28 (0.49 to 3.35)||0.610||7||0.71 (0.28 to 1.85)||0.712|
|8 (9.53 to 12.70)||911||13||0.78 (0.38 to 1.59)||0.493||7||1.01 (0.36 to 2.80)||0.993||6||0.62 (0.23 to 1.70)||0.356|
|9 (12.70 to 16.78)||1071||21||1.00 (0.53 to 1.89)||0.996||9||0.97 (0.36 to 2.61)||0.951||12||1.02 (0.45 to 2.34)||0.961|
|10 (16.78 to 66.68)||1004||19||1.02 (0.53 to 1.95)||0.950||9||1.15 (0.44 to 3.04)||0.775||10||0.93 (0.39 to 2.21)||0.861|
- Top of page
- Research Methods and Procedures
As expected, we found that women who were obese at age 30 were more likely to die and develop cancer in the 6 years after the health survey than women not obese at age 30 (a finding in line with previous research that has shown important associations between obesity and mortality or cancer) (4, 5, 6, 7, 8, 24, 28, 29, 30). However, a much weaker and non-significant association was found when examining the relationship between obesity in 1994 to 1995 and overall mortality and cancer incidence. Further examination of these two groups revealed that 71.8% of those obese at age 30 were still obese in 1995 and, therefore, may have been obese over a long time period. This long-term obesity may be associated with obesity risks. In contrast, only 21.9% of those obese in 1995 had been obese at age 30, showing that a number of women obese in 1994 to 1995 may have been obese for only a short period of time. These figures may explain the weaker relationship seen between obesity in 1994 to 1995 and subsequent mortality and cancer, suggesting that although the long term-effects of obesity can lead to increased mortality and cancer, the short-term effects may not be so clear.
Contrary to expectation, we found that women reporting weight gains between age 30 and 1995 were significantly less likely to die during the 6 years after the health survey than those with stable weight, whereas those with weight loss did not fare any better than those in the stable-weight group. This result remained statistically significant regardless of whether weight change deciles were based on actual weight change, relative weight change, or average annual weight change, after adjustment for potential confounders and after a sensitivity analysis, in which women were excluded who might have lost weight because of serious illness. Similarly weight gain did not seem to be detrimental for the development of cancer, although the risk estimates were imprecise because of small numbers, and none reached statistical significance. These results may be regarded as reassuring given the general trend among most populations in the developed world to increase in mean weight, by between 0.5 and 1 kg per year depending on age and gender (31, 32). The findings, however, must be interpreted with caution given the relationship between obesity at age 30 and increased mortality and cancer risk.
Our findings suggest that weight gain may not be detrimental to health and may actually be associated with a lower risk of cancer and mortality in middle age. These results are not consistent with the majority of previous studies that have found a significant association between weight gain and increased mortality (7, 20, 25), although some studies have found no association (24), and others have indicated that moderate weight gain may be beneficial, e.g., that weight bearing, intensified by obesity, increases bone density (33), whereas weight loss often leads to bone loss, particularly in women, unless controlled by calcium supplementation (34, 35).
Contrary to popular perception, weight loss in our study was not associated with reduced mortality risk. The relationship between weight loss and mortality has been conflicting in previous research. Although some studies have reported a significant increase in mortality (20, 21, 22, 23, 25, 36), some have found no significant association (24), and others have indicated that in certain circumstances, weight loss, even by small amounts, is associated with lower mortality risk (16). A recent trial by Christou et al. (19) indicated mortality benefits for surgical interventions on morbidly obese populations, a finding that Torgorsen et al. (15) (part of the Swedish Obesity Study group) also hope to illustrate, although as yet they do not have sufficient follow-up. These studies have allowed for intentionality of weight loss and the time period over which the weight is lost. There is debate about the best ways of losing weight, with suggestions that diet, exercise, and lifestyle changes are preferable. Recent literature has suggested that intentional weight loss is crucial to identify (16, 37), that the period of time taken to lose the weight may effect psychological and physical health (17, 38), and that strategies to induce fat free mass weight loss are required (37, 39). These also indicate the importance of weight loss methods and maintenance strategies.
Unlike other large prospective studies, our study achieved a high follow-up rate. All of the women were flagged using national registration systems, facilitating the supply of accurate and comprehensive follow-up information about subsequent deaths and cancers. The study also included women from a wide range of social classes and included a relatively large population of overweight or obese women at baseline.
There were, however, a number of potential limitations of the study. First, there were relatively few outcomes reported during the follow-up period, particularly cancers, due to the time delay in cancer registration notifications. This limited power made it difficult to draw any conclusions about the cancer data presented, particularly the hormone-dependent cancers. Small numbers meant that we could not meaningfully stratify for menopausal status in our cancer analysis. Such adjustments are important because estrogen is a known tumor promoter, and endogenous estrogens are increased in obese women. To account for this, we did adjust for menopausal status in our analysis. The numbers of deaths present in the cohort also meant that cause-specific mortality could not be examined. Second, weight was not measured objectively by participating general practitioners; instead, all analyses were based on self-reported weight change information. There may be systematic bias in self-reported data; e.g., obese women may have been more likely to report lighter weights than less heavy women. In addition, weight at age 30 was collected retrospectively as part of the 1994 to 1995 health survey as an approximate measure of weight at recruitment. There may, therefore, have been an element of information bias because the participants were required to recall their weight 25 to 30 years previously. Third, weight change was assessed over only two time-points; the effect and degree of weight fluctuations during the intervening period could not be examined. Fourth, we could adjust only for factors available to us, and residual confounding may exist. Comprehensive information on smoking status over the period of weight change was not available in this study; therefore, more precise adjustment for smoking history was not possible. For example, we did not know how long ago ex-smokers had stopped or for how long current smokers had been smoking. Failure to adjust for this completely might have resulted in residual confounding, whereby women who had gained weight had been less likely to be long-term smokers and, therefore, might have been healthier on this basis.
The analysis could include only women still under general practitioner observation in 1994 to 1995 who completed the health survey. This was 26.5% of women recruited to the original OCS. If women with a large weight gain had died before the health survey, we might be observing the effects of weight change in a healthy cohort of survivors. Previous analyses of mortality among the cohort have not revealed any material biases introduced by the large losses to follow-up (40). This suggests that internal analyses within the women remaining in the cohort are valid. Nonetheless, women in the study have better standardized mortality rates than national figures (40), indicating the need for caution when generalizing our findings to other populations.
This study has identified some interesting and unexpected findings regarding weight change and its association with mortality and cancer. Although obesity at young age was associated with subsequent mortality and cancer incidence, weight gain over a time period of 12 to 51 years appeared to be beneficial when compared with women with stable weight over the same time period. Further research looking at weight change and health outcomes is needed, either to confirm or to refute our findings and to allow detailed examination of potential explanations for the findings.
- Top of page
- Research Methods and Procedures
We thank the hundreds of doctors who have contributed to the OCS during the past 30 years and the thousands of women who completed the questionnaire. The postal questionnaire was funded by an unconditional grant from Wyeth-Ayerst International, Inc., Madison, NJ. Additional support for the study has been received from RCGP, Schering AG, and Schering Health Care.
Nonstandard abbreviations: RCGP, Royal College of General Practitioners; OCS, Oral Contraception Study; ICD, International Classification of Diseases; OC, oral contraceptive; HRT, hormone replacement therapy; CI, confidence interval.
- Top of page
- Research Methods and Procedures
- 1World Health Organization Obesity http:www.who.inttopicsobesityen (Accessed September 2005).
- 2Health Survey for England Department of Health Website http:www.publications.doh.gov.ukstatstrends1.htm (Accessed May 2004).
- 3British Heart Foundation Statistics Annual Increase in Prevalence of Obesity, 1980–1998, All Available Countries Worldwide http:www.heartstats.org (Accessed May 2004).
- 102001) Dieting to reduce body weight for controlling hypertension in adults. Nursing Times 97: 17–23., , , et al (
- 26Royal College of General Practitioners (1974) Oral Contraceptives and Health Pitman Medical London, United Kingdom.
- 27World Health Organization (1967) International Classification of Diseases (Eighth Revision) World Health Organization Geneva.
- 401996) Mortality in relation to method of follow-up in the Royal College of General Practitioners’ Oral Contraception Study. Hannaford PC Webb AMC eds. Evidence-Guided Prescribing of the Pill Parthenon Publishing Group Lancashire, United Kingdom., , (