Osteoporotic fractures, resulting from reduced bone mineral density (BMD), are a major public health problem. Worldwide, osteoporosis is associated with >8.9 million fractures annually, resulting in an osteoporotic fracture every 3 seconds.1 The incidence of hip fracture is estimated to reach 2.6 million by 2025, and will double by 2050 to 4.5 million.2 One year after a hip fracture, 40% of patients are still unable to walk independently and 60% have difficulty with at least one essential activity of daily living.3 Moreover, the overall 1-year mortality after a hip fracture in an elderly population was reported as 21.2%.4
Even though soft drink consumption rates are dropping in the United States,5 they remain high. In 2013, 30% of respondents to the Behavioral Risk Factor Surveillance System questionnaire reported daily intake of sugar-sweetened beverages.6 In 2017, the United States had the second highest soda consumption in the world, after Argentina, with 154 liters consumed per person per year.7 Increased consumption of carbonated soft drinks has been associated with lower BMD and increased rates of bone fractures in studies of children and young adults.8-11 However, some clinical trials and cross-sectional studies have not found significant associations between carbonated beverages and osteoporosis,12-14 and others have suggested that risk for reduced BMD is only increased for certain kinds of beverages, such as colas, but not for all soft drinks.15,16
A large cross-sectional study of 2,500 men and women aged 30 to 86 years from the Framingham Offspring Cohort reported that cola intake was associated with significantly lower BMD at each of three hip sites, but not the spine, and among women, but not men.16 Similar associations were observed for low calorie colas but no associations were found between non-cola soft drinks and BMD. A prospective study of 73,572 postmenopausal women participating in the Nurses’ Health Study reported a 14% increased risk for hip fracture for each additional daily serving of any kind of soda including regular and diet sodas, caffeinated sodas, and non-colas.17 Thus, existing evidence in postmenopausal women is sparse and conflicting for types of soft drinks. To our knowledge, no previous study has examined the association between soda intake and both BMD and fractures.
The aims of this study were to examine associations between soft drink consumption (total, caffeinated, and caffeine-free) and levels of hip and spine BMD, and between soft drink consumption and the risk of incident hip fracture over a period of up to 16 years of follow-up among postmenopausal women.
The Women's Health Initiative (WHI) is an ongoing national study that enrolled 161,808 postmenopausal women recruited from 40 US clinical centers. It consists of three overlapping clinical trials including 68,132 women, as well as an observational study (OS) including 93,676 postmenopausal women aged 50 to 79 at baseline, enrolled during 1993 to 1998.18,19 Data collection at screening and enrollment included questionnaires completed by self-report or interview, physical examination, and blood specimen collection. Information on soda intake (total, caffeinated, and caffeine-free) was collected at the sixth follow-up year in the OS among 79,885 women who form the baseline study population for this analysis. After exclusion of participants who had incomplete exposure information (n = 5,413), had previous history of hip fracture (n = 471), or <1 day of follow-up after year 6 (n = 1,659), a cohort of 72,342 participants contributing 700,388 person/years of follow-up was used for the prospective analysis. BMD measurements were available for a subgroup of women at follow-up year 6 and were used to conduct a cross-sectional analysis of the relationship between soda consumption and BMD. All participants provided informed consent at baseline and subsequently for extended follow-up.
Women were sent questionnaires annually to report any hospitalization and a variety of outcomes, including clinical fractures. Hip fractures were confirmed by central review of radiology reports through 2010 for all WHI participants, and thereafter adjudicated only among women in the Medical Records Cohort of the WHI Extension Study (n = 22,316). A previous validation study in this cohort showed a high degree of agreement (approximately 80%) between self-reported hip fractures and physician-adjudicated fractures among WHI women.20
Participants at three clinical centers (Pittsburgh, PA; Birmingham, AL; and Phoenix and Tucson, AZ) underwent baseline BMD measurements (n = 6,108 women) at the hip, posterior-anterior spine, and total body. BMD was measured using dual-energy x-ray absorptiometry with a Hologic QDR densitometer (Hologic Inc, Marlborough, MA) by technicians who were trained and certified by the University of California, San Francisco, Bone Density Coordinating Center. Standard protocols for positioning and analysis, routine spine and hip phantoms, and a random-sample review were used. Hardware and software changes were centralized, and calibration phantoms, which were scanned across instruments and clinical sites, were in close agreement (interscanner variability: 1.5% for the spine, 4.8% for the hip, and 1.7% for linearity). T scores and Z scores were also recorded and reported.
Dietary information was collected using food frequency questionnaires (FFQ), which were designed by an ad hoc dietary assessment working group composed of WHI scientists and implemented by the WHI Clinical Coordinating Center. The questionnaires were the subject of specific research and considered valid and reliable.21,22 The main exposure was total soft drink intake measured in 12 oz servings per day, week, or month. The term soft drink refers in the literature to nonalcoholic beverages, and the term sodas refers to carbonation. The questionnaire specifically asked about consumption of commonly used carbonated nonalcoholic beverages, by their brand names. Hence, although the technically correct term in this case would be carbonated nonalcoholic beverages, the terms soft drinks and sodas are used indistinctly in this research. Caffeinated and caffeine-free soft drinks were examined separately, and intake was measured in terms of current consumption with nine progressively higher increasing categories ranging from never or <1 serving per month, to 6 or more servings per day. A continuous measure of intake was derived by imputing the midpoint for each category and then defining new variables composed of five intake categories (none, up to 2, 2.1-5, 5.1-14, and >14 servings/wk). The denominators were all converted into weeks to obtain a uniform time unit.
Covariates were selected for analysis based on their established associations with BMD or osteoporotic hip fractures,23-25 and their potential relationship with soft drink consumption. The most recent measure of each covariate was used if not available at study baseline, that is, follow-up year 6. Demographic covariates ascertained by self-report included age, race/ethnicity (White, African American, Asian, Hispanic/Latino, and others), education (<8th grade to postgraduate education), and annual family income (<US$ 12,000->US$ 50,000). Body mass index (BMI) was measured as weight in kilograms divided by squared height in meters, and categorized into underweight, normal, overweight and obesity using WHO criteria.26 Use of hormone therapy was categorized as never, past, or current. Information about use of oral contraceptives (ever), past history of falls (yes/no), and diabetes mellitus (yes/no) were also obtained from baseline questionnaires. Diabetes mellitus was defined as a self-report of physician diagnosis and treatment with insulin or oral antidiabetic drugs. Additional covariates included maternal history of hip fracture (yes/no), smoking and alcohol consumption (never, past, current), and coffee intake. FFQs were used for ascertaining dietary supplements including calcium and vitamin D (in mg/day). Dietary micronutrients including phosphorus, sodium, calcium, and magnesium were estimated from the FFQ by using a database derived from the University of Minnesota's Nutrition Coordinating Center. Participants were asked to bring all current prescription medications to their initial interviews. Selected medications (thyroid hormone, glucocorticoids, and thiazide diuretics) were analyzed as dichotomous variables (yes/no). Self-reported physical activity was analyzed as total expenditure of energy from recreational physical activity (including walking, mild, moderate, and strenuous physical activity in kcal/wk/kg).27 Self-rated general health status was ascertained using the RAND 36-Item Health Survey 1.0 Questionnaire, RAND Corp, and categorized as excellent, very good, good, fair, or poor.28
Baseline characteristics were examined by categories of soft drink consumption using chi-square tests, and analysis of variance tests with Scheffe post hoc analysis (for categorical and continuous variables, respectively). Linear regression analyses were used to examine the association of soda consumption with total hip and lumbar spine t scores. Progressively adjusted models ranged from minimal adjustment for age and race/ethnicity to full adjustment adding BMI, hormone therapy, smoking status, alcohol status, use of oral contraceptives, self-rated general health status, diabetes mellitus, osteoporosis therapy, thiazides, glucocorticoids, thyroid hormone, calcium and vitamin D supplements, coffee intake, total energy expenditure, income, and maternal hip fracture. P values for trend were calculated using soda consumption variables as continuous variables in the models.
Cox proportional hazards models were used to examine associations between soda consumption and time to hip fracture, which was defined from the date of the baseline OS year 6 visit to the date of hip fracture, death, loss to follow-up, or the final follow-up date (September 30, 2016), whichever came first. Hazard ratios (HRs) and 95% confidence intervals were computed from progressively adjusted models that ranged from minimal adjustment for age and race/ethnicity to full adjustment as described above. P values for trend were calculated using soda consumption variables as continuous variables in the models. To examine consistency of associations across categories of the strongest hip fracture risk factors, namely age, race/ethnicity, and BMI, crossproduct terms were added to the models to test the statistical significance of interactions for those variables; no significant interactions were found. A sensitivity analysis was conducted to examine associations between soda consumption and adjudicated hip fracture.
Exact P values for two-sided tests are shown for multivariable analysis; P < 0.05 was considered to be statistically significant for all analyses including interaction terms. Stata software (Ver 13.0; Statacorp Lp) was used in all analyses.
The final analytic sample included 72,342 women, who contributed to 700,388 person/years of follow-up, with a median of 11.9 years (range: 1 month-16 years) (Fig. 1). A total of 2,578 hip fractures were ascertained (3.5% of the final sample).
At baseline analysis (follow-up year 6), women in the highest category of total soda consumption (>14 servings/wk) were younger, reported lower education and lower self-reported general health status, and were much more likely to be obese than those in the lowest drinking category (40.4% vs 16.7%, respectively in the nondrinker category); 14.4% reported diabetes compared with 4.9% of nondrinkers (Table 1). With regards to bone health, women in higher consumption categories presented less demineralization, as reflected in T scores. However, none of the category's mean t scores were in abnormal ranges. A higher proportion of women in the highest consumption category reported use of hormone therapy and a history of oral contraceptive use. They were also more likely to smoke, but less likely to consume alcohol or coffee. Finally, fewer women in the high consumption category reported maternal history of hip fracture, whereas a higher proportion reported a history of two or more falls.
No associations were observed between amount of soda consumption and total hip or lumbar spine t scores after adjustment for age, ethnicity, BMI, and dietary micronutrients (Table 2). Results did not differ for caffeinated and caffeine-free soda drinks. In addition, BMD did not significantly differ among women in the highest soda consumption category compared to nondrinkers (data not shown).
For total soda consumption, both minimally and fully adjusted survival models showed a 26% increased risk of hip fracture among women who drank on average 14 servings per week or more compared with no servings (HR 1.26; 95% CI 1.01-1.56 fully adjusted) (Table 3). The association between high soda consumption and hip fracture was also observed for caffeine-free soda consumers, with a 32% (HR 1.32; 95% CI 1.00-1.75) increased risk of hip fracture among the women consuming >14 servings per week on average versus nondrinkers. No significant associations were observed for the caffeinated soda group, although the HRs were in the direction of increased risk. There was no apparent linear trend in the risk of hip fracture across categories of soda consumption in the fully adjusted models (P values for trend >0.10).
In a sensitivity analysis restricted to 1,275 women with adjudicated hip fractures (Table 4), higher HRs were observed in fully adjusted models among women in the highest intake categories of total soda (HR 1.45, 95% CI 1.07-1.95), and caffeinated soda (HR 1.48, 95% CI 1.00-2.26). The risk of hip fracture for caffeine-free soda consumption was significantly increased in the progressively adjusted models (eg, HR 1.52, 95% CI 1.04-2.24 in model III), although the association in the fully adjusted model (model IV) was not statistically significant (HR 1.43, 95% CI 0.97-2.11).
In this prospective study of soft drink consumption and incident hip fractures in postmenopausal women, we observed a significant association of high soda consumption (at least 14 soda drinks/wk) with incident hip fracture. The association was statistically significant for caffeine-free sodas; however, the risk was smaller and not statistically significant for caffeinated sodas. No associations were observed among women consuming less than the highest amount for any type of soda. In addition, we found no linear associations between soda consumption and reduced BMD in adjusted cross-sectional analysis.
Previous studies in other population-based cohorts showed inconsistent results. The only comparable large, long-term, prospective study analyzing the relation between soda intake and hip fractures was conducted in the Nurses’ Health Study cohort. This study incorporated repeated measures of soda consumption, and reported a linear and significant increase in hip fracture risk with higher soda intake in postmenopausal women.17 Although the linear trend for increased risk of hip fracture was significant for all types of soda, and the risk of hip fracture associated with total soda intake was 42% higher (HR 1.42; 95% CI 1.15-1.74) in the highest consumption category (≥10 servings/wk), the association for the highest category was not significant for colas (HR 1.18; 95% CI 0.81-1.71) or caffeinated sodas (HR 1.18; 95% CI 0.82-1.70). The linear relation was not evident for caffeine-free sodas, but there was a significantly elevated risk for the highest intake category (HR 1.56; 95% CI 1.16-2.09), suggesting a threshold effect. In partial agreement with the Nurses’ Health Study, our results showed no significant risks if the intake was <14 servings per week, suggesting a threshold effect rather than a dose-response relationship.
Two other population-based studies examined the relation between sodas and differences in BMD. A cross-sectional analysis conducted in 1,000 women from the Rancho Bernardo cohort (mean age 72 years, range 44-98) showed no associations between any carbonated beverage and BMD.14 This study defined high consumption using a lower threshold (≥1 servings/day), and controlled for a smaller set of covariates compared to our study. The other cross-sectional study, conducted on 2,500 middle aged men and women from the Framingham Offspring Cohort reported a significant linear association between increased soda intake and reduced hip BMD in women, but the association was only evident for colas (with and without caffeine).16 In this study, the differences in BMD became evident for women who drank more than three servings per week.
Considering these conflicting results, there appears to be no soda ingredient that can be reliably identified as responsible for the observed associations and thus the biological mechanism underlying the association between sodas and bone health still needs to be explored. Previous hypotheses have advanced colas as responsible considered that caffeine, and/or phosphoric acid, and/or cola extracts, all three commonly present in cola beverages, are involved in BMD reductions. We could not analyze colas and non-colas as exposures, but did examine caffeinated and caffeine-free sodas. Although the correlation is not exact because of the existence of a few caffeine containing sodas that are not colas, our results do not support the cola hypothesis: high consumption of caffeine-free sodas showed the highest risk for hip fractures.
Another potential explanation focuses on added sugars as having a negative impact on mineral homeostasis and calcium balance.29,30 We could not control for this factor or examine diet sodas as a separate exposure. However, similar risks were observed for sweetened and diet soda in the Nurses’ Health Study. Furthermore, although previous research questions the role of adults’ milk intake in the risk of osteoporosis and hip fractures, previous research proposed the displacement of milk by sodas as another potential mechanism for reduced BMD.31 We found no significant differences in the mean calcium dietary intake across soda consumption categories, but the study performed on the Framingham cohort reported lower calcium-to-phosphorus ratios in daily soda consumers. Finally, carbonation, the process of dissolving carbon dioxide in water, is the common characteristic of all sodas. It results in the formation of carbonic acid that might alter gastric acidity and, consequently, nutrient absorption. Whether this factor plays a role in these findings32 is yet to be explored.
In our cross-sectional analysis of a subcohort of WHI OS women, although women at higher soda intake categories presented slightly better mean t scores than those with fewer intake, no association was observed between soda consumption as a continuous variable and t scores at the hip or lumbar spine. These null findings argue against a linear relation between soda consumption and BMD. We also failed to find any association when we repeated the analysis comparing the highest drinking category to nondrinkers. In addition, these null findings suggest the need to consider and further investigate whether a reduction in BMD is the causal mechanism in the observed association between high soda consumption and incident hip fracture risk. For instance, a causal path in which soda consumption indirectly increases the likelihood of falls could be explored.
Strengths of this study include up to 16 years of follow-up in a large, diverse cohort of postmenopausal women and availability of information to adjust for a large number of covariates. Although generalizability of these findings is restricted to the universe covered by the Women's Health Initiative in terms of age ranges, education, social status, and other factors, this is the only large prospective study that included examination of both BMD and incident hip fracture. Among the limitations, exposure ascertainment was limited to self-reported soda consumption early in the follow-up period, without repeated measures later in follow-up. This fact implies the impossibility to study cumulative exposure and forces the assumption that the pattern of consumption reported at baseline remained stable over time. In addition, the imputation of average intake data points at each category involves potential distortions in the ascertainment of the exposure. After 2010, hip fractures were self-reported in a large portion of the cohort, but sensitivity analyses restricted to adjudicated hip fractures showed stronger associations with high soda consumption. The reinforcement of the association for the subgroup with adjudicated hip fractures may relate to some extent of outcome misclassification, even though, overall, the correlation between self-reported and adjudicated hip fractures in this cohort was reported to be 78%.20 This result supports the robustness of the finding of increased risk of hip fractures inclusive of self-reported events. In addition to being potentially explained by less misclassification of adjudicated hip fractures, the stronger association in this subgroup could be associated to the fact that a greater proportion of adjudicated hip fractures were measured closer to the baseline when soda consumption was measured and thus, would be less affected by changes in consumption that occurred later in life.
In this prospective study, high soda consumption was associated with a modest increased risk of incident hip fracture but no difference in hip or spine t scores. Contemporary cohorts of US women may drink more soda over their lifetimes than WHI women, for whom the heaviest consumption category was rare (4.9% for total soda consumption over 14 servings/wk). Future research is needed to quantify risks of soda consumption over adulthood, separate out diet from sugar-containing soda consumption, colas from non-colas and carbonated from noncarbonated beverages. More research is also needed to identify the biological mechanisms that might be involved in these findings.
1. Johnell O, Kanis JA. An estimate of the worldwide prevalence, mortality and disability associated with hip fracture. Osteoporos Int
2. Gullberg B, Johnell O, Kanis JA. World-wide projections for hip fracture. Osteoporos Int
3. Cooper C. The crippling consequences of fractures
and their impact on quality of life. Am J Med
4. Schnell S, Friedman SM, Mendelson DA, Bingham KW, Kates SL. The 1-year mortality of patients treated in a hip fracture program for elders. Geriatr Orthop Surg Rehabil
5. Welsh JA, Sharma AJ, Grellinger L, Vos MB. Consumption of added sugars is decreasing in the United States. Am J Clin Nutr
6. Park S, Xu F, Town M, Blanck HM. Prevalence of sugar-sweetened beverage intake among adults—23 States and the District of Columbia, 2013. MMWR Morb Mortal Wkly Rep
8. Wyshak G. Teenaged girls, carbonated beverage consumption, and bone fractures
. Arch Pediatr Adolesc Med
9. Wyshak G, Frisch RE, Albright TE, Albright NL, Schiff I, Witschi J. Nonalcoholic carbonated beverage consumption and bone fractures
among women former college athletes. J Orthop Res
10. Wyshak G, Frisch RE. Carbonated beverages, dietary calcium, the dietary calcium/phosphorus ratio, and bone fractures
in girls and boys. J Adolesc Health
11. Sámano R, Rodríguez Ventura AL, Godínez Martínez EY, et al. Association of consumption of carbonated beverages and decalcification in woman on reproductive and non-reproductive age of Mexico City. Nutr Hosp
12. Schoppen S, Pérez-Granados AM, Carbajal A, De la Piedra C, Pilar Vaquero M. Bone remodelling is not affected by consumption of a sodium-rich carbonated mineral water in healthy postmenopausal women. Br J Nutr
13. Supplee JD, Duncan GE, Bruemmer B, Goldberg J, Wen Y, Henderson JA. Soda intake and osteoporosis
risk in postmenopausal American-Indian women. Public Health Nutr
14. Kim SH, Morton DJ, Barrett-Connor EL. Carbonated beverage consumption and bone mineral density among older women: the Rancho Bernardo study. Am J Public Health
15. Ogur R, Uysal B, Ogur T, et al. Evaluation of the effect of cola drinks on bone mineral density and associated factors. Basic Clin Pharmacol Toxicol
16. Tucker KL, Morita K, Qiao N, Hannan MT, Cupples LA, Kiel DP. Colas, but not other carbonated beverages, are associated with low bone mineral density in older women: the Framingham osteoporosis
study. Am J Clin Nutr
17. Fung TT, Arasaratnam MH, Grodstein F, et al. Soda consumption and risk of hip fractures
in postmenopausal women in the Nurses’ Health Study. Am J Clin Nutr
18. Anderson GL, Manson J, Wallace R, et al. Implementation of the Women's Health Initiative
study design. Ann Epidemiol
2003; 13 (9 suppl):S5–S17.
19. Anderson GL, Cummings SR, Freedman LS, et al. Design of the Women's Health Initiative
clinical trial and observational study. Control Clin Trials
20. Chen Z, Kooperberg C, Pettinger MB, et al. Validity of self-report for fractures
among a multiethnic cohort of postmenopausal women: results from the Women's Health Initiative
observational study and clinical trials. Menopause
21. Block G, Woods M, Potosky A, Clifford C. Validation of a self-administered diet history questionnaire using multiple diet records. J Clin Epidemiol
22. Patterson RE, Kristal AR, Tinker LF, Carter RA, Bolton MP, Agurs-Collins T. Measurement characteristics of the Women's Health Initiative
food frequency questionnaire. Ann Epidemiol
23. Kanis JA, McCloskey EV. Risk factors in osteoporosis
24. Dontas IA, Yiannakopoulos CK. Risk factors and prevention of osteoporosis
. J Musculoskelet Neuronal Interact
25. Lane NE. Epidemiology, etiology, and diagnosis of osteoporosis
. Am J Obstet Gynecol
2006; 194 (2 suppl):S3–S11.
26. World Health Organisation. WHO global strategy on diet, physical activity and health. World Heal Organ Geneva
27. Ainsworth BE, Haskell WIL, Whitt MC, et al. Compendium of physical activities: an update of activity codes and MET intensities. Med Sci Sports Exerc
2000; 32 (9 suppl):S498–S504.
28. Hays RD, Sherbourne CD, Mazel RM. The rand 36-item health survey 1.0. Health Econ
29. Milne DB, Nielsen FH. The interaction between dietary fructose and magnesium adversely affects macromineral homeostasis in men. J Am Coll Nutr
30. Tsanzi E, Fitch CW, Tou JC. Effect of consuming different caloric sweeteners on bone health and possible mechanisms. Nutr Rev
31. Fitzpatrick L, Heaney RP. Got Soda?
32. Cuomo R, Sarnelli G, Savarese MF, Buyckx M. Carbonated beverages and gastrointestinal system: between myth and reality. Nutr Metab Cardiovasc Dis