cow

laniea The Brattle Group, Cambridge, MAbNorth Carolina State University, Raleigh, NC

a r t i c l e i n f o

Article history:Received 27 January 2015Accepted 3 February 2015

Inc. All rights reserved.

government guidance suggested that the reduction of dietary fatwould be accompanied by a concurrent increase in the dietaryshare of carbohydrate. Taken together, these recommendationswere also considered to be benecial for the prevention of

tes, cancer, and other

ry recommendations,dramatically [2]. Thelained in a number ofthe supposition that

changed in response to advice to reduce their intake of fat,saturated fat, and cholesterol [5]. To our knowledge, nocomprehensive analysis of consumption levels both before andafter the endorsement of reduced-fat diets has been undertaken.

In this study, we use a set of government surveys between1965 and 2011 to document patterns and changes in the con-sumption of the average adult since the earliest recommenda-tions. We have three major ndings. First, on average, adults in

The authors have no conicts of interest to report.

Contents lists availab

Nutrit

ww

Nutrition 31 (2015) 727732* Corresponding author. Tel.: 1 617 864 7900; fax: 1 617 864 1576.cholesterol, based on a theoretical link between those foodcomponents and higher risk for coronary heart disease [1]. This

Americans have failed to adhere to dietary recommendations [4].However, others suggest that dietary patterns of Americans havesubstantial increase in obesity. 2015 Elsevier

Introduction

Since the late 1970s, the US government, following theAmerican Heart Association (AHA) and much of academia, hasconsistently recommended lowering the dietary percentage offat and saturated fat, as well as the absolute levels of dietary

overweight and obesity, along with diabechronic diseases [1].

Since the introduction of these dietarates of obesity and diabetes have increasedrapid increase in obesity rates has been expways [3], but is frequently attributed toto an increase in absolute consumption of carbohydrate as opposed to a change in total fat con-sumption. General adherence to recommendations to reduce fat consumption has coincided with aKeywords:NHANESFat shareCarbohydrate shareBMIUSDA dietary guidelinesE-mail address: evan.cohen@brattle.com (E. Cohen

http://dx.doi.org/10.1016/j.nut.2015.02.0070899-9007/ 2015 Elsevier Inc. All rights reserved.a b s t r a c t

Background: For almost 50 y, the US National Health and Nutrition Examination Survey (NHANES)has measured the caloric consumption, and body heights and weights of Americans. The aim of thisstudy was to determine, based on that data, how macronutrient consumption patterns and theweight and body mass index in the US adult population have evolved since the 1960s.Methods: We conducted the rst comprehensive analysis of the NHANES data, documenting howmacronutrient consumption patterns and the weight and body mass index in the US adult pop-ulation have evolved since the 1960s.Results: Americans in general have been following the nutrition advice that the American HeartAssociation and the US Departments of Agriculture and Health and Human Services have beenissuing for more than 40 y: Consumption of fats has dropped from 45% to 34% with a correspondingincrease in carbohydrate consumption from 39% to 51% of total caloric intake. In addition, from1971 to 2011, average weight and body mass index have increased dramatically, with the per-centage of overweight or obese Americans increasing from 42% in 1971 to 66% in 2011.Conclusions: Since 1971, the shift in macronutrient share from fat to carbohydrate is primarily dueEvan Cohen M.B.A. , MichAdele Hite M.P.H., R.D. b, Meragg Ph.D. , Jehan deFonseka M.B.A. ,Rosenberg B.S. a, Bin Zhou Ph.D. aa,* ael C a aApplied nutritional investigation

Statistical review of US macronutrient19652011: Americans have been follocoincident with the rise in obesity

journal homepage:).nsumption data,ing dietary guidelines,

le at ScienceDirect

ion

w.nutr i t ionjrnl .comthe United States have adhered to the governments dietary

and sweets sparingly, and to consume between 6 and 11servings of bread, cereal, rice, and pasta [11]. The 30% dietary fatrecommendations were issued as part of the guidelines in1990, 1995, and 2000. In the 2005 and 2010 guidelines, fatconsumptionwas capped at 35% of calories. A new set of dietaryguidelines is expected in 2015.

It is notable that maintaining or achieving a desirable bodyweight is not addressed at all in the Senates initial DietaryGoals report. Although the recommendation to avoid over-weight was added to the second edition of the report and thesubsequent guidelines, it was not until 2005 and 2010 thatguidance regarding estimated caloric requirements wasincluded in the guidelines [12]. Instead, recommendationsfocused on altering the share of total calories attributable to fat,saturated fat, and carbohydrate as a way of achieving caloricbalance.

ition 31 (2015) 727732recommendations by shifting their proportional consumptionof total calories from fats to carbohydrates. Fat consumptiondropped from 44.7% in 1965 to 33.6% in 2011, whereas carbo-hydrate intake increased from 39% to 50.5% over the sameperiod. Second, the reduction in the proportion of fat andcholesterol consumption arose largely from an increase in thedaily intake of carbohydrate by about 260 calories from 1965 to2011. Third, there is a high correlation (close to 90%) betweenthe change in dietary pattern and the rise of obesity over thelast 40 y. From 1971 to 2011, average weight and body massindex (BMI) have increased dramatically, with the percentage ofoverweight Americans increasing from 42.3% in 1971 to 66.1%in 2011.

The sources of data for this study are 10 US National Healthand Nutrition Examination Surveys (NHANES) conducted bythe National Center for Health Statistics (NCHS) of the Centersfor Disease Control and Prevention (CDC) since the 1970s. Wealso analyzed their predecessor study, the 1965 US Depart-ment of Agriculture (USDA) Household Food ConsumptionSurvey [6]. A 1997 study [7] using the 1976 to 1991 NHANESsurveys showed that fat intake decreased and obesityincreased during that 15-y period. However, contrary to ourndings over a longer period, that study documented a con-current drop in total caloric consumption. To our knowledge,except for the aforementioned article [7], this study presentsthe rst comprehensive compilation of data from NHANES toexamine the long-term trends in consumer dietary behaviorsand concurrent obesity levels.

The second part of this study provides an overview of the USgovernmental dietary recommendations. We then describe thedata source and research methods and present the results onconsumer dietary patterns and correlation with signicant in-creases in obesity.

History of macronutrient intake recommendations

In 1961, spurred by emerging medical and epidemiologicresearch, the AHA issued dietary recommendations to reducethe intake of total fat, saturated fat, and cholesterol [8]. In 1977,the US Senate Select Committee on Nutrition and Human Needsissued Dietary Goals for the United States, which recommendedthat fat consumption be reduced to 30% of energy intake, andthat carbohydrate consumption be increased to account for 55%to 60% of energy intake [1].

Following this report, Dietary Guidelines for Americans, is-sued by the USDA and the US Department of Health, Educationand Welfare (now the Department of Health and Human Ser-vices; DHHS) in 1980, recommended a reduction in the con-sumption of the share of total macronutrients attributable to fatand saturated fat, and a reduction in the absolute consumptionof cholesterol. To compensate, the guidelines recommendedincreasing consumption of carbohydrate as a share of totalcalories because carbohydrates contain less than half thenumber of calories per ounce than fats [9]. During the 1980s,the federal government continued to issue reports and recom-mendations encouraging Americans to limit fat consumption. In1982, the Committee on Diet, Nutrition, and Cancer of the Na-tional Research Council issued Interim Dietary Guidelines thatrecommended fat intake be lowered from 40% to 30% of totalcalories in the diet, ofcially endorsing the AHAs recommen-dations from 1961 and the Senate committees recommenda-tions from 1977 [10]. The USDA and DHHS recommendations

E. Cohen et al. / Nutr728have remained largely unchanged since 1980. In 1992, the FoodGuide Pyramid was released, urging Americans to use fats, oils,NHANES data and research methods

To examine the long-term dietary consumption of the USpopulation, we compiled detailed individual-level consumptionfrom NHANES since the 1970s. Under the auspice of the NCHS, adivisionwithin the CDC (and ultimately, the DHHS), the NHANEShas been conducted periodically since 1971 to assess the healthand nutritional status of the US population by means of stan-dardized interviews and physical examinations. The datacollected include individual-level caloric consumption and bodyweight statistics, as well as detailed demographic information.These individual survey responses can be located on USDAswebsite. The caloric consumption data we relied on are callednutrient composition in the rst two surveys, total nutrientintake for surveys between 1988 and 2002, and total nutrientintakesdday 1 from 2003 onward. Starting in 2003, totalnutrient intake data have been collected over two separate dates,but for consistency with earlier years, we only relied on datacollected from the rst date. All of the data analysis was con-ducted using STATA version 13.

To date, 10 NHANES surveys have been conducted. The rstthree surveys: 19711974, 19761980, and 19881994, occurredirregularly and over uneven survey periods. Starting in 1999,NHANES was regularly conducted over each consecutive 2-yperiod. They include 19992000, 20012002, 20032004,20052006, 20072008, 20092010, and 20112012. The latestsurvey results (20112012) were released in September 2014. Ofthe earlier surveys, NHANES III was conducted in two phases,Fig. 1. Sex and race decomposition represented by NHANES surveys. Source: Au-thors analysis of the NHANES data. Horizontal axis indicates the start year of eachsurvey.

designed to sample larger numbers of certain subgroups ofparticular public health interest, with the planned over-sampling varying by year [13]. For example, in the late1960s and early 1970s, there was concern that very-low-income populations and women of childbearing age were atgreater risk for malnutrition than the general population.Therefore, these subgroups were oversampled in NHANES Ibetween 1971 and 1974. In the three surveys between 1999and 2004, the oversampled subgroups included blacks,Mexican-Americans, low-income white Americans (beginningin 2000), adolescents aged 12 to 19 y, and individuals aged60 y.

To correct for this potential demographic bias, we dividedobservations from each NHANES survey into 18 different de-mographic groups organized by two sexes (male, female), threeraces (white/Hispanic, black, other) and three poverty levels(poverty ratio 2). Figure 1 (six sex/race groups) and Figure 2 (three poverty-level groups) show the nonuniform decomposition of survey

Fig. 2. Poverty level decomposition represented by NHANES surveys. Source: Au-thors analysis of the NHANES data. Horizontal axis indicates the start year of eachsurvey. PR1, below poverty line (poverty ratio 2; PR, poverty ratio.

E. Cohen et al. / Nutrition 31 (2015) 727732 72919881990 and 19911994. In the following analyses, we dividedthe NHANES III survey results into two subperiods in order to bemore in line with the average survey periods and sample sizes inother NHANES surveys. We also referenced each survey by thestarting year. For example, survey 1999 stands for the surveyconducted between 1999 and 2000.

In addition to the NHANES data, we also included some datapoints from a 1965 USDA Household Food Consumption Survey[6]. Unlike the subsequent NHANES data releases, the 1965 in-dividual survey data are unavailable, with only summary tablesreleased. As a result, we did not conduct a detailed analysis of theunderlying data, nor follow our complete resampling method-ology discussed here to normalize the demographic compositionof the surveys over time. However, to the extent possible, weincorporated the 1965 summary data in our long-term trendanalyses.

Without statistical adjustments, the NHANES data havetwo biases for longitudinal analyses. First, the demographiccomposition of the US population, by race, age, and residence,changed over the 40-y time span. To the extent that differentgroups have different consumption and BMI characteristics,inconsistent shares of each group will tend to bias the agg-regated data. Second and more importantly, NHANES wasFig. 3. Average adult fat and carbohydrate caloric consumption. Source: Authors anaresponses for survey participants aged 18 to 64 y.In this study, we focused on the consumption patterns of

Americans aged 18 to 64 y. Given the demographic changes inthe country over the past 40 y and intentional over- andundersampling of the demographics in surveys, we calculatedthe diet and BMI statistics for each of the 18 cohort groups, andthen resampled the survey data to provide a more consistentdiet and BMI comparison of an average American over time.For each of the years, we applied the 19992011 average sampleweights to each cohort group, as the sample weights wererelatively stable during those years. This produced comparabledemographic groups for all survey years to the overall samplebetween 1999 and 2011 Additionally, pregnant women wereexcluded from our reported statistics, both to adjust for theoversampling in the NHANES during the early 1970s and toavoid consumption bias not representative of nonpregnantwomen.

The reported statistics for 1965 consumption and BMI pat-terns in that follow are only resampled for sex. In other words,the 1965 summary statistics for men and women have beenresampled to an average 19992011 sex weight. However,because race and poverty-level sample compositions were notreported with the requisite consumption and bodylysis of the NHANES data. Horizontal axis indicates the start year of each survey.

measurements, the 1965 data for an average American do notcontrol for sampling changes with regard to race and poverty.

Results

Americans caloric intake, 19652011

Trends in American caloric intake on average reect theUSDA/DHHSs recommendations regarding the consumption of fat andcarbohydrate as a share of total calories. According to the 1965USDA survey, fat consumption comprised 44.7% of adult Ameri-cans diets, compared with 39% for carbohydrate. Even by 1971,those numbers had changed signicantly, with fat down to 37.5%and carbohydrateup to 45.6%. By 1999, fat consumption reached a

disconnect that may explain these outcomes. At the same time,our analyses of the NHANES data indicate a correlation betweendietary shift and signicant increases in rates of obesity, which isassociated with increased risk for heart disease. For example,

E. Cohen et al. / Nutrition 31 (2015) 727732730trough of 32.4%, whereas carbohydrate consumption hit its peakof 52.1%. The latest data from 2011 shows these trends slightlyebbing, but fat consumption is still down by 10.4% since 1971, and24.9% since 1965, with fat comprising 33.6% of total caloric con-sumption for the average American adult. The pattern for carbo-hydrate consumption is the mirror opposite. Since 1971,carbohydrate consumption increased 10.9%, and 29.7% since1965. In 2011, on average, 50.5% of total calories were derivedfrom carbohydrate. The trend shown in Figure 3, which presentsfat and carbohydrate consumption as a share of total consump-tion over time, is consistent with the USDA/DHHSs recom-mended shift in share of consumption from fat to carbohydrate,although ultimately never reaching the 55% goal for carbohydrateconsumption, or the 30% ceiling for fat consumption. Proteinconsumption remained relatively constant throughout theperiod, ranging only from15% to 17% of total caloric consumption.

The Senates Dietary Goals report also included the objec-tives of lowering saturated fat consumption to about 10percent of total energy intake and reducing total cholesterolconsumption to about 300 mg per day [1]. Americans con-sumption patterns followed these guidelines. In 1971, saturatedfat comprised 13.5% of total calories. By 2011, Americans wereeating 10.7% of their calories as saturated fat, a 20.5% reductionsince 1971 (Fig. 4).

Likewise, as shown in Figure 5, Americans consumption pat-terns followed the guidelines on cholesterol consumption. Percapita consumption was down 400 mg/din 1971.

These patterns and changes in consumption are consistentwith the hypothesis that Americans on average adhered to thegovernment dietary recommendations regarding fat, saturatedfat, cholesterol, and carbohydrate. However, without specicFig. 4. Saturated fat consumed by adults as a portion of diet. Source: Authorsanalysis of the NHANES data. Horizontal axis indicates the start year of each survey.Saturated fat consumption data not available for 1965.recommendations from the AHA or the USDA/DHHS on totalcaloric intake on an absolute basis, the shift in the share of fatand carbohydrate is primarily due to an almost 65 g, or about a260 kcal, daily increase in Americans intake of carbohydratefrom 1965 to 2011. For fat, Americans consumption on an ab-solute basis fell between 1965 and 1971 by >25 g, but has sinceremained largely at. As shown in Figure 6, the average adultAmerican consumed about 109 g/d of fat and 213 g/d of carbo-hydrate in 1965. Daily fat consumption fell to 83 g in 1971, andremained at approximately the same level through 2011. Incontrast, carbohydrate consumption, although basically at from1965 to 1971 in terms of total calories, has risen to 278 g/d since1965, an increase of 30.6%.

Americans BMI, 19652011

The main policy objective for recommending lower fat,saturated fat, and cholesterol consumption was to reduce heartdisease. A number of recent studies showed that although theincidence of mortality has declined, the goal of reducing risk forheart disease and its effect on the health care system has notbeen achieved on an absolute or per-capita basis [14]. A review ofguidance from advisory committees, including the one for theUSDA/DHHS Dietary Guidelines, found conclusions about therelationship between saturated fat and heart disease reportedby these committees do not reect available science [15], a

Fig. 5. Average mg/d cholesterol consumed by adults. Source: Authors analysis ofthe NHANES data. Horizontal axis indicates the start year of each survey. Choles-terol consumption data not available for 1965.Fig. 6. Total daily consumption y adults. Source: Authors analysis of the NHANESdata. Horizontal axis indicates the start year of each survey.

evidence from the FraminghamHeart Study shows that obesity isa signicant independent predictor of cardiovascular disease

As indicated in Figure 8, in 1971, 42.3% of adult Americanswere overweight, 14.7% were obese, and 1.6% severely obese.Over the past 4 decades, the percent overweight increased by56.3% to 66.1%, the percent obese more than doubled to 35.2%,and the percent severely obese increased >4-fold to 7.2%.

To explore whether this shift in BMI distribution towardobesity, as well as its correlation with the change in dietarycomposition, can be explained by an increase in overall caloricintake, we conducted the same tests over two subsamples ofsurvey participants who consumed similar calories over time.First, we examined whether the pattern held true in a sub-sample of men and women who consumed around the averagecaloric levels between 1971 and 2011, namely men whoconsumed between 2400 and 2500 kcal/d and women whoconsumed between 1700 and 1800 kcal/d in each survey. (Thisgroup comprised 1.5% and 1.6% of the total sample of men in1971 and 2011, respectively, and 3.6% and 2.5% of the totalsample of women in those same years, respectively.) Our resultsand general conclusions of this study for the whole sample heldtrue for this subsample, demonstrating that the change in the

Fig. 7. Average adult male and female BMI. Source: Authors analysis of theNHANES Data. Horizontal axis indicates the start year of each survey. Overweight:BMI >25 kg/m2; obese: BMI >30 kg/m2. BMI, body mass index.

E. Cohen et al. / Nutrition 31 (2015) 727732 731[16,17].Our analysis of the NHANES shows that the distribution of

BMI has shifted. BMI is a well-accepted proxy for weightcompared with height, with BMI >25 kg/m2 representing over-weight, >30 kg/m2 representing obese, and >40 kg/m2 repre-senting severely obese. Figure 7 shows the signicant increasefrom 1965 in average adult BMI for both men and women.Average BMI increased from about 24 kg/m2 in 1965 to about 29kg/m2 in 2011. There is a strong relationship between the in-crease in carbohydrate share of total intake and obesity. This isreected by the high correlation at the individual level betweencaloric share from carbohydrate and adult BMI: 85.3% for menand 91.2% for women.

The large uptick in average BMI is reected in the large in-creases in the American population who are overweight (BMI>25 kg/m2), obese (BMI >30 kg/m2) and severely obese (BMI>40 kg/m2). Figure 8 shows that the BMI distribution was muchtighter around the median BMI 24.1 kg/m2 in 1971, with only aslight rightward skew in the higher BMI end of the tail. By 2011,the median BMI had risen steeply to 27.4 kg/m2, with the dis-tribution skewing much more toward the severely obese.Fig. 8. Adult BMI distribution for 1971 and 2011. Source: AuthBMI distribution was not caused by total caloric intake for thissubgroup.

Second, we examined the BMI change and dietary pattern formen and women who consumed signicantly higher total calo-ries per day. For men, we looked at the group consuming be-tween 3400 and 3600 kcal/d, whereas for women we looked atthe group consuming 2400 to 2600 kcal/d. The patterns for thissubsample are consistent with the full sample: Macronutrientshares start at different levels but trend consistently with the fullsample, and BMI also rises in concert with the full sample. Totalfat calories fell for both groups by 3 to 4 percentage points in2011, from 40% in 1971. BMI for men was just under 25 kg/m2 in1971, but grew to just under 29 kg/m2 in 2011. This represented asimilar starting point in 1971, with a slight growth from 28 kg/m2

for the full sample, to 29 kg/m2 for the high-calorie group. Forwomen, BMI started off at 23 kg/m2 in 1971, signicantly lowerthan the full samples starting BMI of about 25 kg/m2. However,both groups of women grew to BMIs just under 29 kg/m2 in 2011.

Taken together, the similar trends in BMI for both the fullsample and two subsamples with relatively constant caloric in-takes indicated that the increase in total caloric consumptionors analysis of the NHANES data. BMI, body mass index.

since 1971 was not likely to offer any signicant explanation forthe increase in BMI over the past 4 decades.

Conclusion

In the rst comprehensive statistical analysis using all avail-able NHANES surveys, this study demonstrated that generaladherence to government dietary recommendations to decreasefat share of total dietary intake has been accompanied by a rapidincrease in obesity rates. The results from this study compel a fullanalysis of the potential causal relationship between Americansseeming adherence to the governments nutrition recommen-dations and obesity.

Acknowledgments

The authors acknowledge Anna Blatz for her statistical workon this article, and Nina Teicholz, for her support and editorialcontributions.

References

[1] Select Committee on Nutrition and Human Needs of the United StatesSenate. Dietary goals for the United States [Internet]. 2nd ed. Washington,DC: U.S. Government Printing Ofce; 1977. Available at: http://catalog.hathitrust.org/Record/000325810. Accessed January 14, 2015.

[2] Hite AH, Feinman RD, Guzman GE, Satin M, Schoenfeld PA, Wood RJ. In theface of contradictory evidence: report of the Dietary Guidelines forAmericans Committee. Nutr Burbank Los Angel City Calif 2010;26:91524.

[3] Zinn AR. Unconventional wisdom about the obesity epidemic. Am J Med Sci2010;340:48191.

Dietary Guidelines for Americans, 2010. Washington DC: U.S. Departmentof Agriculture; 2010.

[5] Marantz PR, Bird ED, Alderman MH. A call for higher standards of evidencefor dietary guidelines. Am J Prev Med 2008;34:23440.

[6] U.S. Department of Agriculture. Food and nutrient intake of individualsin the United States. Report No. 11. Available at: http://www.ars.usda.gov/SP2UserFiles/Place/12355000/pdf/6566/hfcs6566_rep_11.pdf. AccessedJanuary 14, 2015.

[7] Heini AF, Weinsier RL. Divergent trends in obesity and fat intake patterns:the American paradox. Am J Med 1997;102:25964.

[8] Kritchevsky D. History of recommendations to the public about dietary fat.J Nutr 1998;128:449S52S.

[9] U.S. Department of Agriculture and U.S. Department of Health and HumanServices. Nutrition and your health: dietary guidelines for Americans. 1sted. Washington, DC: U.S. Government Printing Ofce; 1980. Available at:http://www.cnpp.usda.gov/Publications/DietaryGuidelines/1980/DG1980pub.pdf. Accessed January 14, 2015.

[10] National Research Council. Diet nutrition and cancer. Washington DC:National Academy Press; 1982. Available at: http://archive.org/details/dietnutritionand019332mbp. Accessed January 14, 2015.

[11] Davis C, Saltos E. Dietary recommendations and how they have changedover time. Americas eating habits: changes and consequences. WashingtonDC: U.S. Department of Agriculture, Economic Research Service; 1999.

[12] U.S. Department of Health and Human Services and U.S. Department ofAgriculture. Dietary guidelines for Americans. 6th ed. Washington DC: U.S.Government Printing Ofce; 2005.

[13] Centers for Disease Control and Prevention. Key concepts about NHANESsurvey design. Available at: http://www.cdc.gov/nchs/tutorials/nhanes/SurveyDesign/SampleDesign/Info1.htm. Accessed February 27, 2015.

[14] Go AS, Mozaffarian D, Roger VL, Benjamin EJ, Berry JD, Borden WB, et al.Heart disease and stroke statisticsd2013 update a report from the Amer-ican Heart Association. Circulation 2013;127:e6245.

[15] Hoenselaar R. Saturated fat and cardiovascular disease: the discrepancybetween the scientic literature and dietary advice. Nutr Burbank LosAngel Cty Calif 2012;28:11823.

[16] Hubert HB, Feinleib M, McNamara PM, Castelli WP. Obesity as an inde-pendent risk factor for cardiovascular disease: a 26-year follow-upof participants in the Framingham Heart Study. Circulation 1983;67:96877.

E. Cohen et al. / Nutrition 31 (2015) 727732732[4] U.S. Department of Agriculture, U.S. Department of Health and HumanServices. Report of the Dietary Guidelines Advisory Committee on the[17] Eckel RH, Krauss RM. American Heart Association call to action: obesity as amajor risk factor for coronary heart disease. Circulation 1998;97:2099100.

Statistical review of US macronutrient consumption data, 19652011: Americans have been following dietary guidelines, coinc ...IntroductionHistory of macronutrient intake recommendationsNHANES data and research methodsResultsAmericans' caloric intake, 19652011Americans' BMI, 19652011

ConclusionAcknowledgmentsReferences