American Journal of Eptderrtology Copyright C 1992 by The Johns Hopkins University School of Hygiene and Pubfc Health Al rights reserved
Vol 135, No 3 Printed in U.S.A.
Is Gallstone Disease caused by Obesity or by Dieting?
The effects of dieting and obesity on the risk of acute gallstone disease were evaluated in a case-control study in Maastricht, The Netherlands, during 1983-1986. The study comprised 151 cases with acutely symptomatic gallstone disease and 451 population controls. The effects of dieting and obesity as measured by body mass index (weight (kg)/height (mf) were disentangled in a multivariate logistic regression analysis. Both dieting and body mass index were positively associated with the rate of gallstone disease. The association with dieting largely disappeared when initial body mass index was controlled for (rate ratio = 1.4, 95% confidence interval 0.8-2.4). Conversely, controlling for dieting did not affect the association between body mass index and gallstone disease. Analysis by risk period (year 1 and 2-5, 6-10, and 11-15 years prior to interview) did not show consistent risk period-specific effects when initial body mass index was controlled for. The authors conclude that dieting does not account for the association between obesity and gallstone disease. In obese persons, dieting does not increase the risk of acute gallstone disease over the long term. Am J Epidemiol 1992:135:274-80. body weight; cholelithiasis; diet; gallbladder diseases; obesity; risk factors; weight loss
Obesity is an established risk factor for gallstones (1-4). It does not automatically follow that weight reduction prevents the development of gallstone disease. In metabolic studies, half of the subjects who followed a weight-reduction diet had a transient increase in bile lithogenicity (1). Rapid weight loss after gastric bypass has been reported to result in a high rate of gallstone formation and associated symptoms (5-9). Therefore, it has been suggested that weight reduction might increase the risk of gallstone disease, especially if weight loss is not maintained (1, 10). Consequently, the higher risk of gallstones in obese persons may partly be explained by the greater instability of higher body mass.
In a case-control study, we attempted to disentangle the effects of dieting and body mass on the risk of acutely symptomatic gallstone disease. MATERIALS AND METHODS Subjects
Admission criteria for both cases and controls were being aged 30-76 years for males and aged 20-76 years for females. Gallstone cases were from a series of patients who were referred to Academic Hospital Maastricht, Maastricht, The Netherlands, in 1983-1985 with newly diagnosed acute gallstone disease (biliary colic of recent onset, acute cholecystitis, cholangitis, pancreatitis, or obstructive jaundice, as reported in the hospital records of patients who were admitted for cholecystectomy or in the referral forms of the patients at the Department of Radiology) and who had surgically or radiologically confirmed gallstones in the
Received fcx publication December 7,1990, and in final form September 12, 1991. Abbreviations: a , confidence interval; RR, rate ratio. From the Department of Epidemiology and BiostatJstics, Ri^sunivefsiteit Umburg, P.O. Box 616, NL-6200 MD Maastricht, The Netherlands. (Reprint requests to Dr. Caret Thijs at this address).
274
Downloaded from https://academic.oup.com/aje/article-abstract/135/3/274/97316 by university of winnipeg user on 20 January 2019
Care) Thijs, Paul KnipschikJ, and Pieter Letters
Gallstone Disease and Dieting
Measurements
Subjects were interviewed by trained interviewers using a structured questionnaire. They were asked whether they had ever dieted. If they had, the number of diet classes taken and the maximal weight loss were recorded. This was asked for the year prior to interview and for the time intervals between 1, 5, 10, and 15 years earlier. In addition, the subjects were asked whether their weight had generally remained lower after dieting or was rapidly regained. Current body weight and height were asked about and were measured by the interviewer in doubtful cases. Body weight was also sought for 5, 10, and 15 years before interview. Analysis
Rate ratios for dieting and body mass index (weight (kg)/height (m)2) were computed in categorical logistic regression analysis (11). Body mass index categories were based on quartiles (all subjects combined), and the resulting upper category (>26.3 kg/ m2) was further divided into two because of its informativeness. Categories were (kg/m2): 14.2-21.7, 21.8-23.9, 24.0-26.2, 26.3-28.9, and 29.0-44.4. To obtain estimates of risk period-specific effects of dieting, we separated the effects of dieting in the risk periods from each other by including all dieting variables for the risk periods in the regression models (12).
Primarily, dieting was analyzed as a dichotomous variable (yes/no). Dose-effect relations were evaluated using indicator variables for the number of diet classes and the maximum amount of weight loss (as continuous variables) in nested models (11). Rapid weight regain was evaluated likewise. Body mass index was separated into six categories, including one representing missing data (mainly young subjects whose body mass index in the past pertained to ages below 15). Each analysis was carried out both with and without control for body mass index at the beginning of the period in question ("initial body mass index"). A similar analysis was performed for the entire period from 15 years earlier up to the time of diagnosis. Our method of controlling for confounding needs explanation. People who diet are likely to be heavier at the outset than people who do not. For this reason, initial body mass index should be controlled for. Controlling for body mass index at the end of the risk period of interest would be inadequate if the effect of dieting is mediated by change in body mass index, or even if body mass index is only an indicator of some factor in the pathogenetic mechanism of gallstone formation, e.g., serum lipids (13). For these reasons, each analysis was conducted with control for body mass index at the beginning of the relevant risk period ("initial body mass index"). The same argument holds for confounding by some other factors. Life-style habits that are risk factors for gallstone disease may differ among people who diet and people who do not. Insofar as these habits precede dieting, they can be considered potential confounders which should be controlled. Such potential confounders are physical activity, skipping breakfast, and alcohol use. By contrast, changes in these risk factors may be part of the weight-reducing activities (intermediate factors), which should not be controlled for. For this reason, we controlled for these factors at the beginning of each risk period only (sedentary life-style (yes/no)), skipping breakfast (three categories), and alcohol use (three categories denoting patterns
Downloaded from https://academic.oup.com/aje/article-abstract/135/3/274/97316 by university of winnipeg user on 20 January 2019
gallbladder. This yielded 151 cases (51 men and 100 women; response rate = 86 percent). A random sample, stratified for age and sex, from the municipal population registries of the catchment area of the hospital (the only hospital serving that region) was chosen to serve as the control group. After exclusion of subjects with a history of gallstone disease (radiologic gallstone diagnosis (n = 7), biliary colic (n = 17), or cholecystectomy (« = 26)), the control group comprised 451 subjects (146 men and 305 women; response rate = 72 percent).
275
276
Thijs et a).
of use—never/seldom, irregular, and regular—with indicators of the number of drinks consumed per week (14)). Other confounders controlled for in all models were (number of categories in parentheses): sex; age (six); interviewer (three); pregnancy (two) and oral contraceptive use (two) in four risk periods; use of cholesterollowering drugs at any time (two) or longterm use of analgesics (two); medical diagnosis of diabetes (two); and history of gallstones in brothers (two), sisters (two), and parents (five). Evaluation of interactions in the logistic regression models showed no major differences in the results for men and women. For
RESULTS
Dieting (n = 166 (28 percent)) was associated with a higher rate of gallstone disease (rate ratio (RR) = 2.0 for ever dieting vs. never dieting; 95 percent confidence interval (CI) 1.24-3.22). When initial body mass index was controlled for in the analysis, the rate ratio decreased to 1.4 (95 percent CI 0.84-2.40). Conversely, the relation between gallstone disease and initial body mass index 15 years before interview (figure 1) was similar whether dieting was controlled for or not
64 32
Men
1
16 8
A
/ f
4 /
2 1 1/2 20
25
30
body mass index (kg/mJ)
RR
64
Women 32 16 8 1
1/2 2O
30
body mass Index (kg/m1) FIGURE 1. Incidence of acute gallstone disease as a function of body mass index 15 years before Interview, expressed as a ratio to the rate in the lowest body mass index category: Maastricht, The Netherlands, 1983-1986. Points in the graph are set at the midpoints (means) of these categories. Vertical bars represent 95% confidence intervals. Solid lines represent rate ratios (RR) from categorical logistic regression analysis, controlling for age and interviewer. (No other variables confounded the relation between body mass index and gallstone risk.) Broken fines represent body mass index as a continuous variable, controlling for age and interviewer. (No other variables confounded the relation between body mass index and gallstone risk.)
Downloaded from https://academic.oup.com/aje/article-abstract/135/3/274/97316 by university of winnipeg user on 20 January 2019
RR
this reason, the results are shown for men and women combined.
Gallstone Disease and Dieting
No dose-effect relations were found between the risk of gallstone disease and the number of diet classes or the maximal amount of weight ever lost. Neither was an effect found for extreme dieting (dieting more than once, with weight loss of 10 kg or more at least once, and rapid weight regain). Among women, there were 10 cases and 13 controls with extreme dieting (mean number of diet classes = 8.4 (range, 2-49); mean weight loss = 17.0 kg (range, 10-42 kg)). In this group, the rate ratio was 2.4 (95 percent CI 0.79-7.02) compared with never dieting, controlling for initial body mass index; it was 1.2 (95 percent CI 0.39-4.00) when we additionally controlled for weight gain (mean = 7.3 kg in extreme dieters in the period from 15 to 5 years before diagnosis). In men, only one control had reported extreme dieting. In the risk period-specific analyses, a positive relation between dieting and gallstone disease was found only in the 6- to 10-year risk period (RR = 3.5 (95 percent CI 1.52-
TABLE 1. Relation between acute gallstone disease and dieting during the 15 years prior to interview, expressed as a rate ratio and stratified by initial body mass Index (15 years before Interview): Maastricht, The Netherlands, 1983-1986 Initial body mass index*
Men
Women
Dieted
RRt
95% a t
11 88
5.2 1.0§
1.33-20.5*
4 11
14 56
1.0 1.0§
0.31-3.07*
5 31
5 14
18 28
1.4 1.0§
0.44-4.41*
8 4
12 15
9 7
15 15
1.7 1.0§
0.54-5.40*
2 4
3 2
12 12
13 5
0.2 1.0§
0.04-1.20*
3
6 9
11 31 1.4 1.0§
0.84-2.40D
Cases
Cases
Controls
Yes No
4
1 30
5 6
21.8-23.9
Yes No
2 13
6 38
24.0-26.2
Yes No
4 10
26.3-28.9
Yes No
29.0-44.4
Yes No
Unknown
Yes No
14.2-21.7
Total
Controls
Yes No
16 35
27 119
41 59
82 223
Total
51
146
100
305
• Weight (kg)/height (m)». t RR, rate ratio; CI, confidence Interval. $ Controfing tor age, sex, and Interviewer only, because of the sma« numbers in each stratum erf initial body mass index. § Referent | ControHng for initial body mass Index (15 years before Interview; six categories); physical activity, skipping breakfast, and alcohol use, 15 years before diagnosis; and aO other contounders.
Downloaded from https://academic.oup.com/aje/article-abstract/135/3/274/97316 by university of winnipeg user on 20 January 2019
(rate ratios for body mass index in men and women were 1.19 (90 percent CI 1.11 -1.26) and 1.20 (90 percent CI 1.12-1.27), respectively, for an increment in body mass index of 1 kg/m2). The results did not differ essentially for rapid weight regain (n = 75; RR = 1.3 (95 percent CI 0.67-2.64)) and more enduring weight loss (« = 91; RR = 1.5 (95 percent CI 0.79-2.84)); both were compared with never dieting and controlled for initial body mass index. The relation between dieting and the occurrence of gallstone disease appeared to be modified by initial body mass index (table 1; trend of decreasing rate ratio with increasing initial body mass index, two-tailed p = 0.023 (test of the interaction between dieting and initial body mass index, assigning scores of 1 to 5 for the respective body mass index categories)). An increased rate of gallstone disease was found in thin persons; in obese subjects, dieting seemed to protect against gallstone disease.
277
278
Thijs et al.
DISCUSSION
We emphasize that our study focused on a clinical problem (acutely symptomatic gallstone disease), not merely on a biologic problem (formation of gallstones). Consequently, the study base from which the cases were derived was defined as the experience over time of people who have not already
had acutely symptomatic gallstone disease, regardless of whether they have gallstones or not, and who would appear as cases in our study if they were to get acutely symptomatic gallstone disease. The choice of the reference group (a general population control group) logically follows from this research question, since all people in the hospital catchment region who get acute gallstone disease will indeed appear at Academic Hospital Maastricht. This should preclude bias due to selection with respect to factors related to dieting history and obesity. The exclusion of subjects with a history of symptomatic gallstone disease but not with asymptomatic gallstones also followed from the same principle (11). Therefore, even a high prevalence of asymptomatic gallstones in the general population (10 percent in men and 20 in women in our population (15)) cannot have led to misclassification bias. The following findings make it less likely that obese persons have a higher risk of acutely symptomatic gallstone disease because they experience unsuccessful weight reduction more often than nonobese persons. First, the association between dieting and gallstone disease was much weaker than
TABLE 2. Logistic regression analysis of the relation between acute gallstone disease and dieting in specified risk periods during the 15 years prior to interview: Maastricht, The Netherlands, 1983-1986* Risk period (years before Interview)
Not controlltng for Initial body mass indexf Dieted (no.)
RR§
95%CI§
11-15
Yes (25) No (577)
0.4 1.0|
0.13-1.48
6-10
Yes (47) No (555)
3.5 10|
1.52-6.31
2-5
Yes (86) No (516)
1.2 1.0|
1
Yes (80) No (522)
1.2 10|
ControHng for initial body mass Indexf, t Dieted (no.)
RR
95% CI
Yes (25) No (577)
0.3 1.0|
0.09-1.12
Yes (47) No (555)
2.6
1.05-6.31
0.61-2.39
Yes (86) No (516)
0.8 1.0Q
0.35-1.59
0.61-2.57
Yes (80) No (522)
0.9 1.0Q
0.42-2.11
1.01
•Afl models contained the variables dieting (yes/no) 1 year before Interview and 1-5, 6-10, and 11-15 years before diagnosis at trie same time, and controlled for the fotowtng variables: physical activity, skipping breakfast, and alcohol use, al three at the beginning of each risk period; sex; age; Interviewer; pregnancy and oral contraceptive use in al four risk periods; use of cholesterollowering drugs at any time and tang-term use of analgesics; medical diagnosis of diabetes; and history of gallstones in brothers, sisters, father, and mother. t Weight (kgyheignKm)1. t Control of Initial body mass Index was achieved by inclusion of body mass Index 15 years before diagnosis for the 11-to 15year risk period; body mass Index 10 and 15 years before dagnosls for the 6- to 10-year risk period; and body mass Index, 5,10, and 15 years before dtegnosis for the 2- to 5-year and 1-year risk periods. § RR, rate ratio; CI, confidence Interval. | Referent.
Downloaded from https://academic.oup.com/aje/article-abstract/135/3/274/97316 by university of winnipeg user on 20 January 2019
6.31); table 2). After we controlled for initial body mass index, the rate ratios for gallstone disease decreased in all risk periods. In the 6- to 10-year risk period, the rate ratio remained substantial (RR = 2.6 (95 percent CI 1.05-6.31)). However, a reversed doseeffect relation was found in that period (negative trend with maximum amount of weight lost (as a continuous variable), twotailed p = 0.005; for weight loss of 1-8 kg (n = 14), RR = 12.3 (95 percent CI 2.85-53.2); for >9 kg (n = 32), RR = 1.2 (95 percent CI 0.40-3.60) (referent, no dieting (n = 555)); one subject had missing data for amount of weight lost in that period). Consistent doseeffect relations or a relation with rapid weight regain was not found in any other risk period.
Gallstone Disease and Dieting
Various gastric and jejunoileal operations for the treatment of morbid obesity have been found to be followed by a high rate of gallstone formation and development of symptomatic gallbladder disease (5-9). One study has shown that symptomatic gallstone disease occurred during the first 2 years after surgery in the majority of cases (6). Another study demonstrated that dieting (a mean weight reduction of 16.5 kg) led to gallstone formation in 13 of 51 subjects after 8 weeks, of whom three underwent cholecystectomy for symptomatic gallstone disease. In half of the subjects with asymptomatic gallstones, the gallstones had disappeared at 6 months of follow-up (16). It has been shown experimentally that bile lithogenicity increases during weight reduction but decreases later to a level lower than the initial level (17,18). During a lifetime, gallstones develop in about one out of every six men and two out of every five women (15). In the obese, the incidence is likely to be much higher. There-
fore, the question remains as to what the net, long-term effect of dieting is in obese people (19). Experimental studies so far have had too short follow-up periods or insufficient numbers of subjects to answer this question, so that knowledge of the long-term risk must be based on the results available from nonexperimental studies. Two earlier nonexperimental studies have been reported, one on asymptomatic gallstone disease and one on symptomatic gallstone disease. In the first study, the prevalence of gallstones upon ultrasound screening was higher in men when a history of dieting was present, but not in women (20). In the second, a prospective cohort study in women, a higher risk of symptomatic gallstone disease was found for weight gain but not for weight loss, as computed from body weight at various points during follow-up and in the past (21). We found only a slightly and not statistically significantly increased rate of acutely symptomatic gallstone disease associated with dieting in obese persons when the data were considered for the entire 15-year risk period. At first glance, this suggests that dieting is not deleterious for the obese in the long run. However, a rate for a long risk period should be interpreted with caution (12). It is especially important to exclude a short-term increased rate of gallstone disease of sufficient magnitude to overrule a decreased rate afterwards. In any case, our findings appear to rule out a highly increased rate during the first 5 years after self-reported dieting. The results for subsequent risk periods do not indicate a long-term increased rate. Therefore, the result of the risk periodspecific analysis suggests that dieting does not cause acutely symptomatic gallstone disease over the long term.
REFERENCES 1. Bennion LJ, Grundy SM. Risk factors for the development of cholelithiasis in men. (Second of two parts). N Engl J Med 1978^299:1221-7. 2. Strom BL, West SL. The epidemiology of gallstone disease. In: Cohen S, Soloway RD, eds. Gallstones.
Downloaded from https://academic.oup.com/aje/article-abstract/135/3/274/97316 by university of winnipeg user on 20 January 2019
the association between body mass index and gallstone disease. Second, the effect of dieting always decreased substantially whenever initial body mass was controlled for, whereas the results for body mass did not change when dieting was controlled for. Inasmuch as the relation between extreme dieting and the rate of gallstone disease in the entire 15-year risk period did not disappear fully when initial body mass index was controlled for, the relation was largely accounted for by weight gain in this group of extreme dieters. Third, in obese persons, dieting seemed to protect against acute gallstone disease rather than cause it. Furthermore, within each risk period, dieting was negatively related to the rate of gallstone disease when initial body mass index was controlled for. An exception is that dieting was associated with a doubled rate for the 6to 10-year risk period. However, no consistent positive dose-effect relation with the amount of weight loss was found. Fourth, the relation between body mass index and gallstone disease persisted when the analysis was restricted to subjects with constant body mass (within limits of 3 kg) over the last 15 years before diagnosis (results not shown).
279
280
3. 4.
6. 7. 8. 9.
10. 11. 12.
(Contemporary issues in gastroentcrology, vol 4). New York: Churchill Livingstone, Inc, 1985:1-26. Pixley F. Epidemiology. In: Bateson MC, ed. Gallstone disease and its management Lancaster, England: MTP Press, 1986:1-23. Scragg RKR. Aetiology of gallstones. In: Bateson MC, ed. Gallstone disease and its management. Lancaster, England: MTP Press, 1986:25-55. Wattchow DA, Hall JC, Whiting MJ, et al. Prevalence and treatment of gall stones after gastric bypass surgery for morbid obesity. BMJ 1983^286: 763. Deitel M, Petrov I. Incidence of symptomatic gallstones after bariatric operations. Surg Gynecol Obstet 1987;164:549-52. Knecht BH. Experience with gastric bypass for massive obesity. Am Surg 1978;44:496-504. Amaral JF, Thompson WR. Gallbladder disease in the morbidly obese. Am J Surg 1985;149:551-7. Yang HY, Petersen GM, Marks JW, et al. Risk factors for gallstone formation during rapid weight loss. (Abstract). Gastroenterology 199O;98(suppl): A266. Atilli AF. Natural history. In: Bateson MC, ed. Gallstone disease and its management Lancaster, England: MTP Press, 1986:57-70. Miettinen OS. Theoretical epidemiology: principles of occurrence research in medicine. New York: John Wiley & Sons, Inc, Publishers, 1985. Thijs C, Knipschild P, Leffers P. Pregnancy and gallstone disease: an empiric demonstration of the importance of specification of risk periods. Am J
Epidemiol 1991;134:186-95. 13. Thijs CT, Knipschild PG, Brombacher P. Serum lipids and gallstones: a case-control study. Gastroenterology 1990;99:843-9. 14. Thijs CT, Knipschild PG, Leffers P. Does alcohol protect against the formation of gallstones? A demonstration of protopathic bias. J Clin Epidemiol 1991;44:941-6. 15. Thijs C, Knipschild P, van Engelshoven J. The prevalence of gallstone disease in a Dutch population. Scand J Gastroenterol 1990^25:155-60. 16. Iiddle RA, Goldstein RB, Saxton J. Gallstone formation during weight-reduction dieting. Arch Intern Med 1989; 149:1750-3. 17. Bennion LJ, Grundy SM. Effects of obesity and caloric intake on biliary lipid metabolism in man. J Clin Invest 1975;56:996-1011. 18. Janowitz P, Swobodnik W, Wechsler JG, et al. Veranderungen der biliaren Lithogenitat waJirend einer vierwochigen Gewichtsreduktion. (In German). Leber Magen Darm 1989;4:190-5. 19. Faloon WW. Hepatobiliary effects of obesity and weight-reducing surgery. Semin liver Dis 1988;8: 229-36. 20. Jorgensen T. Gall stones in a Danish population: relation to weight, physical activity, smoking, coffee consumption, and diabetes mellitus. Gut 1989;30: 528-34. 21. Maclure KM, Hayes KC, Colditz GA, et al. Weight diet, and the risk of symptomatic gallstone disease in middle-aged women. N Engl J Med 1989;321: 563-9.
Downloaded from https://academic.oup.com/aje/article-abstract/135/3/274/97316 by university of winnipeg user on 20 January 2019
5.
Thijs et al.
Vol 135, No 3 Printed in U.S.A.
Is Gallstone Disease caused by Obesity or by Dieting?
The effects of dieting and obesity on the risk of acute gallstone disease were evaluated in a case-control study in Maastricht, The Netherlands, during 1983-1986. The study comprised 151 cases with acutely symptomatic gallstone disease and 451 population controls. The effects of dieting and obesity as measured by body mass index (weight (kg)/height (mf) were disentangled in a multivariate logistic regression analysis. Both dieting and body mass index were positively associated with the rate of gallstone disease. The association with dieting largely disappeared when initial body mass index was controlled for (rate ratio = 1.4, 95% confidence interval 0.8-2.4). Conversely, controlling for dieting did not affect the association between body mass index and gallstone disease. Analysis by risk period (year 1 and 2-5, 6-10, and 11-15 years prior to interview) did not show consistent risk period-specific effects when initial body mass index was controlled for. The authors conclude that dieting does not account for the association between obesity and gallstone disease. In obese persons, dieting does not increase the risk of acute gallstone disease over the long term. Am J Epidemiol 1992:135:274-80. body weight; cholelithiasis; diet; gallbladder diseases; obesity; risk factors; weight loss
Obesity is an established risk factor for gallstones (1-4). It does not automatically follow that weight reduction prevents the development of gallstone disease. In metabolic studies, half of the subjects who followed a weight-reduction diet had a transient increase in bile lithogenicity (1). Rapid weight loss after gastric bypass has been reported to result in a high rate of gallstone formation and associated symptoms (5-9). Therefore, it has been suggested that weight reduction might increase the risk of gallstone disease, especially if weight loss is not maintained (1, 10). Consequently, the higher risk of gallstones in obese persons may partly be explained by the greater instability of higher body mass.
In a case-control study, we attempted to disentangle the effects of dieting and body mass on the risk of acutely symptomatic gallstone disease. MATERIALS AND METHODS Subjects
Admission criteria for both cases and controls were being aged 30-76 years for males and aged 20-76 years for females. Gallstone cases were from a series of patients who were referred to Academic Hospital Maastricht, Maastricht, The Netherlands, in 1983-1985 with newly diagnosed acute gallstone disease (biliary colic of recent onset, acute cholecystitis, cholangitis, pancreatitis, or obstructive jaundice, as reported in the hospital records of patients who were admitted for cholecystectomy or in the referral forms of the patients at the Department of Radiology) and who had surgically or radiologically confirmed gallstones in the
Received fcx publication December 7,1990, and in final form September 12, 1991. Abbreviations: a , confidence interval; RR, rate ratio. From the Department of Epidemiology and BiostatJstics, Ri^sunivefsiteit Umburg, P.O. Box 616, NL-6200 MD Maastricht, The Netherlands. (Reprint requests to Dr. Caret Thijs at this address).
274
Downloaded from https://academic.oup.com/aje/article-abstract/135/3/274/97316 by university of winnipeg user on 20 January 2019
Care) Thijs, Paul KnipschikJ, and Pieter Letters
Gallstone Disease and Dieting
Measurements
Subjects were interviewed by trained interviewers using a structured questionnaire. They were asked whether they had ever dieted. If they had, the number of diet classes taken and the maximal weight loss were recorded. This was asked for the year prior to interview and for the time intervals between 1, 5, 10, and 15 years earlier. In addition, the subjects were asked whether their weight had generally remained lower after dieting or was rapidly regained. Current body weight and height were asked about and were measured by the interviewer in doubtful cases. Body weight was also sought for 5, 10, and 15 years before interview. Analysis
Rate ratios for dieting and body mass index (weight (kg)/height (m)2) were computed in categorical logistic regression analysis (11). Body mass index categories were based on quartiles (all subjects combined), and the resulting upper category (>26.3 kg/ m2) was further divided into two because of its informativeness. Categories were (kg/m2): 14.2-21.7, 21.8-23.9, 24.0-26.2, 26.3-28.9, and 29.0-44.4. To obtain estimates of risk period-specific effects of dieting, we separated the effects of dieting in the risk periods from each other by including all dieting variables for the risk periods in the regression models (12).
Primarily, dieting was analyzed as a dichotomous variable (yes/no). Dose-effect relations were evaluated using indicator variables for the number of diet classes and the maximum amount of weight loss (as continuous variables) in nested models (11). Rapid weight regain was evaluated likewise. Body mass index was separated into six categories, including one representing missing data (mainly young subjects whose body mass index in the past pertained to ages below 15). Each analysis was carried out both with and without control for body mass index at the beginning of the period in question ("initial body mass index"). A similar analysis was performed for the entire period from 15 years earlier up to the time of diagnosis. Our method of controlling for confounding needs explanation. People who diet are likely to be heavier at the outset than people who do not. For this reason, initial body mass index should be controlled for. Controlling for body mass index at the end of the risk period of interest would be inadequate if the effect of dieting is mediated by change in body mass index, or even if body mass index is only an indicator of some factor in the pathogenetic mechanism of gallstone formation, e.g., serum lipids (13). For these reasons, each analysis was conducted with control for body mass index at the beginning of the relevant risk period ("initial body mass index"). The same argument holds for confounding by some other factors. Life-style habits that are risk factors for gallstone disease may differ among people who diet and people who do not. Insofar as these habits precede dieting, they can be considered potential confounders which should be controlled. Such potential confounders are physical activity, skipping breakfast, and alcohol use. By contrast, changes in these risk factors may be part of the weight-reducing activities (intermediate factors), which should not be controlled for. For this reason, we controlled for these factors at the beginning of each risk period only (sedentary life-style (yes/no)), skipping breakfast (three categories), and alcohol use (three categories denoting patterns
Downloaded from https://academic.oup.com/aje/article-abstract/135/3/274/97316 by university of winnipeg user on 20 January 2019
gallbladder. This yielded 151 cases (51 men and 100 women; response rate = 86 percent). A random sample, stratified for age and sex, from the municipal population registries of the catchment area of the hospital (the only hospital serving that region) was chosen to serve as the control group. After exclusion of subjects with a history of gallstone disease (radiologic gallstone diagnosis (n = 7), biliary colic (n = 17), or cholecystectomy (« = 26)), the control group comprised 451 subjects (146 men and 305 women; response rate = 72 percent).
275
276
Thijs et a).
of use—never/seldom, irregular, and regular—with indicators of the number of drinks consumed per week (14)). Other confounders controlled for in all models were (number of categories in parentheses): sex; age (six); interviewer (three); pregnancy (two) and oral contraceptive use (two) in four risk periods; use of cholesterollowering drugs at any time (two) or longterm use of analgesics (two); medical diagnosis of diabetes (two); and history of gallstones in brothers (two), sisters (two), and parents (five). Evaluation of interactions in the logistic regression models showed no major differences in the results for men and women. For
RESULTS
Dieting (n = 166 (28 percent)) was associated with a higher rate of gallstone disease (rate ratio (RR) = 2.0 for ever dieting vs. never dieting; 95 percent confidence interval (CI) 1.24-3.22). When initial body mass index was controlled for in the analysis, the rate ratio decreased to 1.4 (95 percent CI 0.84-2.40). Conversely, the relation between gallstone disease and initial body mass index 15 years before interview (figure 1) was similar whether dieting was controlled for or not
64 32
Men
1
16 8
A
/ f
4 /
2 1 1/2 20
25
30
body mass index (kg/mJ)
RR
64
Women 32 16 8 1
1/2 2O
30
body mass Index (kg/m1) FIGURE 1. Incidence of acute gallstone disease as a function of body mass index 15 years before Interview, expressed as a ratio to the rate in the lowest body mass index category: Maastricht, The Netherlands, 1983-1986. Points in the graph are set at the midpoints (means) of these categories. Vertical bars represent 95% confidence intervals. Solid lines represent rate ratios (RR) from categorical logistic regression analysis, controlling for age and interviewer. (No other variables confounded the relation between body mass index and gallstone risk.) Broken fines represent body mass index as a continuous variable, controlling for age and interviewer. (No other variables confounded the relation between body mass index and gallstone risk.)
Downloaded from https://academic.oup.com/aje/article-abstract/135/3/274/97316 by university of winnipeg user on 20 January 2019
RR
this reason, the results are shown for men and women combined.
Gallstone Disease and Dieting
No dose-effect relations were found between the risk of gallstone disease and the number of diet classes or the maximal amount of weight ever lost. Neither was an effect found for extreme dieting (dieting more than once, with weight loss of 10 kg or more at least once, and rapid weight regain). Among women, there were 10 cases and 13 controls with extreme dieting (mean number of diet classes = 8.4 (range, 2-49); mean weight loss = 17.0 kg (range, 10-42 kg)). In this group, the rate ratio was 2.4 (95 percent CI 0.79-7.02) compared with never dieting, controlling for initial body mass index; it was 1.2 (95 percent CI 0.39-4.00) when we additionally controlled for weight gain (mean = 7.3 kg in extreme dieters in the period from 15 to 5 years before diagnosis). In men, only one control had reported extreme dieting. In the risk period-specific analyses, a positive relation between dieting and gallstone disease was found only in the 6- to 10-year risk period (RR = 3.5 (95 percent CI 1.52-
TABLE 1. Relation between acute gallstone disease and dieting during the 15 years prior to interview, expressed as a rate ratio and stratified by initial body mass Index (15 years before Interview): Maastricht, The Netherlands, 1983-1986 Initial body mass index*
Men
Women
Dieted
RRt
95% a t
11 88
5.2 1.0§
1.33-20.5*
4 11
14 56
1.0 1.0§
0.31-3.07*
5 31
5 14
18 28
1.4 1.0§
0.44-4.41*
8 4
12 15
9 7
15 15
1.7 1.0§
0.54-5.40*
2 4
3 2
12 12
13 5
0.2 1.0§
0.04-1.20*
3
6 9
11 31 1.4 1.0§
0.84-2.40D
Cases
Cases
Controls
Yes No
4
1 30
5 6
21.8-23.9
Yes No
2 13
6 38
24.0-26.2
Yes No
4 10
26.3-28.9
Yes No
29.0-44.4
Yes No
Unknown
Yes No
14.2-21.7
Total
Controls
Yes No
16 35
27 119
41 59
82 223
Total
51
146
100
305
• Weight (kg)/height (m)». t RR, rate ratio; CI, confidence Interval. $ Controfing tor age, sex, and Interviewer only, because of the sma« numbers in each stratum erf initial body mass index. § Referent | ControHng for initial body mass Index (15 years before Interview; six categories); physical activity, skipping breakfast, and alcohol use, 15 years before diagnosis; and aO other contounders.
Downloaded from https://academic.oup.com/aje/article-abstract/135/3/274/97316 by university of winnipeg user on 20 January 2019
(rate ratios for body mass index in men and women were 1.19 (90 percent CI 1.11 -1.26) and 1.20 (90 percent CI 1.12-1.27), respectively, for an increment in body mass index of 1 kg/m2). The results did not differ essentially for rapid weight regain (n = 75; RR = 1.3 (95 percent CI 0.67-2.64)) and more enduring weight loss (« = 91; RR = 1.5 (95 percent CI 0.79-2.84)); both were compared with never dieting and controlled for initial body mass index. The relation between dieting and the occurrence of gallstone disease appeared to be modified by initial body mass index (table 1; trend of decreasing rate ratio with increasing initial body mass index, two-tailed p = 0.023 (test of the interaction between dieting and initial body mass index, assigning scores of 1 to 5 for the respective body mass index categories)). An increased rate of gallstone disease was found in thin persons; in obese subjects, dieting seemed to protect against gallstone disease.
277
278
Thijs et al.
DISCUSSION
We emphasize that our study focused on a clinical problem (acutely symptomatic gallstone disease), not merely on a biologic problem (formation of gallstones). Consequently, the study base from which the cases were derived was defined as the experience over time of people who have not already
had acutely symptomatic gallstone disease, regardless of whether they have gallstones or not, and who would appear as cases in our study if they were to get acutely symptomatic gallstone disease. The choice of the reference group (a general population control group) logically follows from this research question, since all people in the hospital catchment region who get acute gallstone disease will indeed appear at Academic Hospital Maastricht. This should preclude bias due to selection with respect to factors related to dieting history and obesity. The exclusion of subjects with a history of symptomatic gallstone disease but not with asymptomatic gallstones also followed from the same principle (11). Therefore, even a high prevalence of asymptomatic gallstones in the general population (10 percent in men and 20 in women in our population (15)) cannot have led to misclassification bias. The following findings make it less likely that obese persons have a higher risk of acutely symptomatic gallstone disease because they experience unsuccessful weight reduction more often than nonobese persons. First, the association between dieting and gallstone disease was much weaker than
TABLE 2. Logistic regression analysis of the relation between acute gallstone disease and dieting in specified risk periods during the 15 years prior to interview: Maastricht, The Netherlands, 1983-1986* Risk period (years before Interview)
Not controlltng for Initial body mass indexf Dieted (no.)
RR§
95%CI§
11-15
Yes (25) No (577)
0.4 1.0|
0.13-1.48
6-10
Yes (47) No (555)
3.5 10|
1.52-6.31
2-5
Yes (86) No (516)
1.2 1.0|
1
Yes (80) No (522)
1.2 10|
ControHng for initial body mass Indexf, t Dieted (no.)
RR
95% CI
Yes (25) No (577)
0.3 1.0|
0.09-1.12
Yes (47) No (555)
2.6
1.05-6.31
0.61-2.39
Yes (86) No (516)
0.8 1.0Q
0.35-1.59
0.61-2.57
Yes (80) No (522)
0.9 1.0Q
0.42-2.11
1.01
•Afl models contained the variables dieting (yes/no) 1 year before Interview and 1-5, 6-10, and 11-15 years before diagnosis at trie same time, and controlled for the fotowtng variables: physical activity, skipping breakfast, and alcohol use, al three at the beginning of each risk period; sex; age; Interviewer; pregnancy and oral contraceptive use in al four risk periods; use of cholesterollowering drugs at any time and tang-term use of analgesics; medical diagnosis of diabetes; and history of gallstones in brothers, sisters, father, and mother. t Weight (kgyheignKm)1. t Control of Initial body mass Index was achieved by inclusion of body mass Index 15 years before diagnosis for the 11-to 15year risk period; body mass Index 10 and 15 years before dagnosls for the 6- to 10-year risk period; and body mass Index, 5,10, and 15 years before dtegnosis for the 2- to 5-year and 1-year risk periods. § RR, rate ratio; CI, confidence Interval. | Referent.
Downloaded from https://academic.oup.com/aje/article-abstract/135/3/274/97316 by university of winnipeg user on 20 January 2019
6.31); table 2). After we controlled for initial body mass index, the rate ratios for gallstone disease decreased in all risk periods. In the 6- to 10-year risk period, the rate ratio remained substantial (RR = 2.6 (95 percent CI 1.05-6.31)). However, a reversed doseeffect relation was found in that period (negative trend with maximum amount of weight lost (as a continuous variable), twotailed p = 0.005; for weight loss of 1-8 kg (n = 14), RR = 12.3 (95 percent CI 2.85-53.2); for >9 kg (n = 32), RR = 1.2 (95 percent CI 0.40-3.60) (referent, no dieting (n = 555)); one subject had missing data for amount of weight lost in that period). Consistent doseeffect relations or a relation with rapid weight regain was not found in any other risk period.
Gallstone Disease and Dieting
Various gastric and jejunoileal operations for the treatment of morbid obesity have been found to be followed by a high rate of gallstone formation and development of symptomatic gallbladder disease (5-9). One study has shown that symptomatic gallstone disease occurred during the first 2 years after surgery in the majority of cases (6). Another study demonstrated that dieting (a mean weight reduction of 16.5 kg) led to gallstone formation in 13 of 51 subjects after 8 weeks, of whom three underwent cholecystectomy for symptomatic gallstone disease. In half of the subjects with asymptomatic gallstones, the gallstones had disappeared at 6 months of follow-up (16). It has been shown experimentally that bile lithogenicity increases during weight reduction but decreases later to a level lower than the initial level (17,18). During a lifetime, gallstones develop in about one out of every six men and two out of every five women (15). In the obese, the incidence is likely to be much higher. There-
fore, the question remains as to what the net, long-term effect of dieting is in obese people (19). Experimental studies so far have had too short follow-up periods or insufficient numbers of subjects to answer this question, so that knowledge of the long-term risk must be based on the results available from nonexperimental studies. Two earlier nonexperimental studies have been reported, one on asymptomatic gallstone disease and one on symptomatic gallstone disease. In the first study, the prevalence of gallstones upon ultrasound screening was higher in men when a history of dieting was present, but not in women (20). In the second, a prospective cohort study in women, a higher risk of symptomatic gallstone disease was found for weight gain but not for weight loss, as computed from body weight at various points during follow-up and in the past (21). We found only a slightly and not statistically significantly increased rate of acutely symptomatic gallstone disease associated with dieting in obese persons when the data were considered for the entire 15-year risk period. At first glance, this suggests that dieting is not deleterious for the obese in the long run. However, a rate for a long risk period should be interpreted with caution (12). It is especially important to exclude a short-term increased rate of gallstone disease of sufficient magnitude to overrule a decreased rate afterwards. In any case, our findings appear to rule out a highly increased rate during the first 5 years after self-reported dieting. The results for subsequent risk periods do not indicate a long-term increased rate. Therefore, the result of the risk periodspecific analysis suggests that dieting does not cause acutely symptomatic gallstone disease over the long term.
REFERENCES 1. Bennion LJ, Grundy SM. Risk factors for the development of cholelithiasis in men. (Second of two parts). N Engl J Med 1978^299:1221-7. 2. Strom BL, West SL. The epidemiology of gallstone disease. In: Cohen S, Soloway RD, eds. Gallstones.
Downloaded from https://academic.oup.com/aje/article-abstract/135/3/274/97316 by university of winnipeg user on 20 January 2019
the association between body mass index and gallstone disease. Second, the effect of dieting always decreased substantially whenever initial body mass was controlled for, whereas the results for body mass did not change when dieting was controlled for. Inasmuch as the relation between extreme dieting and the rate of gallstone disease in the entire 15-year risk period did not disappear fully when initial body mass index was controlled for, the relation was largely accounted for by weight gain in this group of extreme dieters. Third, in obese persons, dieting seemed to protect against acute gallstone disease rather than cause it. Furthermore, within each risk period, dieting was negatively related to the rate of gallstone disease when initial body mass index was controlled for. An exception is that dieting was associated with a doubled rate for the 6to 10-year risk period. However, no consistent positive dose-effect relation with the amount of weight loss was found. Fourth, the relation between body mass index and gallstone disease persisted when the analysis was restricted to subjects with constant body mass (within limits of 3 kg) over the last 15 years before diagnosis (results not shown).
279
280
3. 4.
6. 7. 8. 9.
10. 11. 12.
(Contemporary issues in gastroentcrology, vol 4). New York: Churchill Livingstone, Inc, 1985:1-26. Pixley F. Epidemiology. In: Bateson MC, ed. Gallstone disease and its management Lancaster, England: MTP Press, 1986:1-23. Scragg RKR. Aetiology of gallstones. In: Bateson MC, ed. Gallstone disease and its management. Lancaster, England: MTP Press, 1986:25-55. Wattchow DA, Hall JC, Whiting MJ, et al. Prevalence and treatment of gall stones after gastric bypass surgery for morbid obesity. BMJ 1983^286: 763. Deitel M, Petrov I. Incidence of symptomatic gallstones after bariatric operations. Surg Gynecol Obstet 1987;164:549-52. Knecht BH. Experience with gastric bypass for massive obesity. Am Surg 1978;44:496-504. Amaral JF, Thompson WR. Gallbladder disease in the morbidly obese. Am J Surg 1985;149:551-7. Yang HY, Petersen GM, Marks JW, et al. Risk factors for gallstone formation during rapid weight loss. (Abstract). Gastroenterology 199O;98(suppl): A266. Atilli AF. Natural history. In: Bateson MC, ed. Gallstone disease and its management Lancaster, England: MTP Press, 1986:57-70. Miettinen OS. Theoretical epidemiology: principles of occurrence research in medicine. New York: John Wiley & Sons, Inc, Publishers, 1985. Thijs C, Knipschild P, Leffers P. Pregnancy and gallstone disease: an empiric demonstration of the importance of specification of risk periods. Am J
Epidemiol 1991;134:186-95. 13. Thijs CT, Knipschild PG, Brombacher P. Serum lipids and gallstones: a case-control study. Gastroenterology 1990;99:843-9. 14. Thijs CT, Knipschild PG, Leffers P. Does alcohol protect against the formation of gallstones? A demonstration of protopathic bias. J Clin Epidemiol 1991;44:941-6. 15. Thijs C, Knipschild P, van Engelshoven J. The prevalence of gallstone disease in a Dutch population. Scand J Gastroenterol 1990^25:155-60. 16. Iiddle RA, Goldstein RB, Saxton J. Gallstone formation during weight-reduction dieting. Arch Intern Med 1989; 149:1750-3. 17. Bennion LJ, Grundy SM. Effects of obesity and caloric intake on biliary lipid metabolism in man. J Clin Invest 1975;56:996-1011. 18. Janowitz P, Swobodnik W, Wechsler JG, et al. Veranderungen der biliaren Lithogenitat waJirend einer vierwochigen Gewichtsreduktion. (In German). Leber Magen Darm 1989;4:190-5. 19. Faloon WW. Hepatobiliary effects of obesity and weight-reducing surgery. Semin liver Dis 1988;8: 229-36. 20. Jorgensen T. Gall stones in a Danish population: relation to weight, physical activity, smoking, coffee consumption, and diabetes mellitus. Gut 1989;30: 528-34. 21. Maclure KM, Hayes KC, Colditz GA, et al. Weight diet, and the risk of symptomatic gallstone disease in middle-aged women. N Engl J Med 1989;321: 563-9.
Downloaded from https://academic.oup.com/aje/article-abstract/135/3/274/97316 by university of winnipeg user on 20 January 2019
5.
Thijs et al.
