Observing Protein And Energy Nutrition Research Articles

Estimating and testing interactions when explanatory variables are subject to non-classical measurement error

Assessing interactions in linear regression models when covariates have measurement error (ME) is complex.We previously described regression calibration (RC) methods that yield consistent estimators and standard errors for interaction coefficients of normally distributed covariates having classical ME. Here we extend normal based RC (NBRC) and linear RC (LRC) methods to a non-classical ME model, and describe more efficient versions that combine estimates from the main study and internal sub-study. We apply these methods to data from the Observing Protein and Energy Nutrition (OPEN) study. Using simulations we show that (i) for normally distributed covariates efficient NBRC and LRC were nearly unbiased and performed well with sub-study size ≥200; (ii) efficient NBRC had lower MSE than efficient LRC; (iii) the naïve test for a single interaction had type I error probability close to the nominal significance level, whereas efficient NBRC and LRC were slightly anti-conservative but more powerful; (iv) for markedly non-normal covariates, efficient LRC yielded less biased estimators with smaller variance than efficient NBRC. Our simulations suggest that it is preferable to use: (i) efficient NBRC for estimating and testing interaction effects of normally distributed covariates and (ii) efficient LRC for estimating and testing interactions for markedly non-normal covariates.

Differential association of sugar-sweetened beverages in men and women: is it the sugar or calories?

Dear Sir: The findings from the recent study by Tasevska et al (1) suggest that there is a difference in response to sugars between women and men. Similar to other cohort studies that found a sex difference for soda intake and stroke risk (2), Tasevska et al reported a significant positive cardiovascular mortality trend with greater intakes of sugar from beverages in women but not in men (1). Although the difference in sex physiology is a plausible explanation for these associations, the consideration for other differences in sex responses is necessary in the interpretation of these analyses. The authors adjusted for total energy intake in their models. Taking this approach reduces the effect of potential confounding but does not address the underlying limitation of energy intake measurements: underreporting, which may lead to overestimation of the association between exposure and outcome. Although the Diet History Questionnaire used in the study has been validated and shown to have moderate correlations with 24-h recalls for added sugar assessment (0.68 in men and 0.79 in women) (3), the OPEN (Observing Protein and Energy Nutrition) Biomarker Study reported that the Diet History Questionnaire is still significantly limited by underreporting of energy intake (4). The underreporting of energy intake is concerning because a recent assessment of the validity of self-report dietary intake from 24-h recalls found that more than two-thirds of women have energy estimates that are not physiologically plausible (5). Furthermore, women also tend to underreport their energy intake much more than men (by ∼365 kcal/d in women and ∼281 kcal/d in men) (5). This inability to provide accurate estimations of energy intake limits our ability to interpret data and may have a significant impact on clinical decisions. Therefore, even when energy intake is adjusted in association models, these association models may not necessarily be free from the confounding effect of energy. Other lines of high-quality evidence from systematic reviews and meta-analyses of controlled trials have shown excess energy to be an important mediating factor in the effects of fructose on cardiometabolic risk. In a series of Canadian Institutes of Health Research–funded (clinicaltrials.gov identifier NCT01363791) systematic reviews and meta-analyses of controlled feeding trials, we found that fructose in isocaloric exchange for other carbohydrates (energy-matched conditions between the fructose and carbohydrate comparator arms) showed no signal for harm in relation to body weight (6), fasting and postprandial lipids (7), glycemic control (8), insulin (8), blood pressure (9), and uric acid (9) and markers of nonalcoholic fatty liver disease (10). Although there may be a dose threshold for fasting lipids in some subgroup analyses (7), an overall lack of harm is seen even under conditions of fructose overfeeding (positive energy balance) at high doses, as long as the comparison with the carbohydrate comparator remains matched for the excess calories. A consistent signal for harm is only seen in imbalanced, hypercaloric comparisons, in which fructose supplements control diets with excess calories compared with the same control diets alone without the excess calories. In the absence of a clear effect on cardiometabolic risk factors in isocaloric comparisons (especially under conditions of positive energy balance), fructose does not appear to be any worse than other refined carbohydrates. The implication is that the adverse effects seen in the hypercaloric comparisons relate to the excess energy rather than the fructose. Consideration for total energy intake would therefore seem to be essential in understanding whether an association with fructose exists beyond the energy it contributes. In conclusion, underreporting of energy complicates the interpretation of the association between sugar intake and cardiovascular mortality risk. This issue is especially important when considering the response observed in women because they are more likely to underreport their energy intake than men. To address the issue of energy as a confounding factor, isocaloric randomized controlled trials are needed to isolate the true effect of sugar intake, independent of energy intake, on cardiovascular health.

A Measurement Error Model for Physical Activity Level as Measured by a Questionnaire With Application to the 1999-2006 NHANES Questionnaire

Systematic investigations into the structure of measurement error of physical activity questionnaires are lacking. We propose a measurement error model for a physical activity questionnaire that uses physical activity level (the ratio of total energy expenditure to basal energy expenditure) to relate questionnaire-based reports of physical activity level to true physical activity levels. The 1999-2006 National Health and Nutrition Examination Survey physical activity questionnaire was administered to 433 participants aged 40-69 years in the Observing Protein and Energy Nutrition (OPEN) Study (Maryland, 1999-2000). Valid estimates of participants' total energy expenditure were also available from doubly labeled water, and basal energy expenditure was estimated from an equation; the ratio of those measures estimated true physical activity level ("truth"). We present a measurement error model that accommodates the mixture of errors that arise from assuming a classical measurement error model for doubly labeled water and a Berkson error model for the equation used to estimate basal energy expenditure. The method was then applied to the OPEN Study. Correlations between the questionnaire-based physical activity level and truth were modest (r = 0.32-0.41); attenuation factors (0.43-0.73) indicate that the use of questionnaire-based physical activity level would lead to attenuated estimates of effect size. Results suggest that sample sizes for estimating relationships between physical activity level and disease should be inflated, and that regression calibration can be used to provide measurement error-adjusted estimates of relationships between physical activity and disease.

Checking for completeness of 24-h urine collection using para-amino benzoic acid not necessary in the Observing Protein and Energy Nutrition study

The orally administered para-amino benzoic acid (PABA) is known to have near 100% excretion in urine and is used as a measure of 24-h urine collection completeness (referred to as PABAcheck). The purpose was to examine the effect of including urine collections deemed incomplete based on PABAcheck in a dietary measurement error study. The Observing Protein and Energy Nutrition (OPEN) study was conducted in 1999-2000 and included 484 men and women aged 40-69 years. A food frequency questionnaire and 24-h dietary recalls were evaluated using recovery biomarkers that included urinary nitrogen and potassium from two 24-h urine collections. Statistical modeling determined the measurement error properties of dietary assessment instruments. In the original analyses, PABAcheck was used as a measure of complete urine collection; incomplete collections were either excluded or adjusted to acceptable levels. The OPEN data were reanalyzed including all urine collections and by using criteria based on self-reported missing voids to assess the differences. Means and coefficients of variation for biomarker-based protein and potassium intakes, and measurement error model-based correlations and attenuation factors were similar regardless of whether PABAcheck or missed voids were considered. PABAcheck may not be required in large population-based biomarker studies. However, until there are more analyses evaluating the necessity of a PABAcheck, it is recommended that PABA be given to all participants, but not necessarily analyzed. Then, PABAcheck could be used selectively as a marker of completeness among the collections in which low levels of biomarker are detected or for which noncompliance is suspected.

Correction: Use of the Predictive Sugars Biomarker to Evaluate Self-Reported Total Sugars Intake in the Observing Protein and Energy Nutrition (OPEN) Study

Correction: Use of the Predictive Sugars Biomarker to Evaluate Self-Reported Total Sugars Intake in the Observing Protein and Energy Nutrition (OPEN) Study

Re: "Application of a repeat-measure biomarker measurement error model to 2 validation studies: Examination of the effect of within-person variation in biomarker measurements"

The paper by Preis et al. (1) contains potentially misleading statements concerning the impact of within-person biomarker variation on the Observing Protein and Energy Nutrition (OPEN) Study results. The major theme of the paper is estimation of deattenuation factors and correlations between intakes reported on a food frequency questionnaire or 24-hour diet recall and true usual intake. A second theme is estimating correlations between person-specific systematic errors in the food frequency questionnaire and 24-hour diet recall. Two models, labeled “(1)” in the paper and another, unlabeled, that we refer to as “model 2,” are considered. Model 1 assumes that 24-hour diet recalls provide unbiased measures of usual intake, whereas the biomarkers are biased. Model 2 assumes the reverse. We see no reason to consider model 1 for the cases of doubly labeled water and urinary nitrogen. Previous feeding studies with urinary nitrogen (2–4) and indirect calorimetry studies with doubly labeled water (3, 5–7) have found no appreciable bias in these recovery biomarkers (8). Although model 1 has been previously used (8, 9), it was only in studies with concentration biomarkers known to be biased. Claiming that doubly labeled water within-person variation is underestimated in the OPEN Study, Preis et al. (1) use an alternative estimate from the Automated Multiple-Pass Method (AMPM) Validation Study to reanalyze the OPEN Study’s deattenuation factors in their Table 5. However, as shown in the Appendix (1), the deattenuation factor does not depend on biomarker within-person variation under model 2. As a result, the authors’ estimated deattenuation factors for the OPEN Study food frequency questionnaire under model 2 (0.07 for energy, 0.16 for protein, and 0.33 for protein density) are similar to those reported by Kipnis et al. (10) (0.080 and 0.039 for energy, 0.156 and 0.137 for protein, and 0.404 and 0.316 for protein density (men and women, respectively)). Small differences are probably due to minor analytical differences from Kipnis et al. (10) (use of the second vs. the first food frequency questionnaire and the combining of analyses for men and women). We agree that the level of biomarker within-person variation does affect correlations between self-report and true usual intake and also that the estimated within-person variation could be sensitive to time between repeats. However, we see little evidence that this affected the OPEN Study results. The correlations for the OPEN Study reported in Table 4 (using the AMPM Validation Study within-person variation) were 0.25 for energy, 0.30 for protein, and 0.36 for protein density. Kipnis et al. (10) reported 0.199 and 0.098 for energy (men and women), 0.323 and 0.298 for protein, and 0.431 and 0.356 for protein density. The energy estimates of Kipnis et al. are slightly lower than the authors’ estimate of 0.25. However, epidemiologic analyses rarely include energy alone. Therefore, protein density is more relevant, and the larger AMPM Validation Study doubly labeled water within-person variance appears to have negligible impact on the protein density correlation. Thus, the results do not support the conclusion that the 2-week period between doubly labeled water repeats in the OPEN Study led to “underestimation of the FFQ’s [food frequency questionnaire’s] validity” (1, p. 684). Finally, Preis et al. (1) suggest in their Figure 1 that the short period between doubly labeled water repeats exaggerated the correlation between food frequency questionnaire and 24-hour diet recall systematic errors. Recalculation using the AMPM Validation Study biomarker within-person variation gave correlations of 0.31 for energy, 0.25 for protein, and 0.59 for protein density. Kipnis et al. (10) reported 0.45 and 0.28 for energy (men and women), 0.18 and 0.24 for protein, and 0.40 and 0.94 for protein density. Averaging over men and women gives values of 0.365, 0.21, and 0.67—not materially different from those reported by Preis et al. Although the timing of repeated biomarker measurements in a validation study deserves careful consideration, we think Preis et al. greatly exaggerated its potential impact with regard to previously reported estimates from the OPEN Study.

Use of the Predictive Sugars Biomarker to Evaluate Self-Reported Total Sugars Intake in the Observing Protein and Energy Nutrition (OPEN) Study

A predictive biomarker for intake of total sugars was recently developed under controlled conditions. We used this biomarker to assess measurement error (ME) structure in self-reported intake of total sugars in free-living individuals. The Observing Protein and Energy Nutrition (OPEN) study involved 484 participants aged 40 to 69 years. Diet was assessed using two administrations of a food frequency questionnaire (FFQ) and two nonconsecutive 24-hour dietary recalls (24HDR). Two 24-hour urine samples checked for completeness were analyzed on sucrose and fructose. We applied the biomarker calibrated in a feeding study to OPEN data to assess the ME structure and the attenuation factors (AF) for intakes of absolute total sugars and sugars density for the FFQ and 24HDR. The AFs for absolute sugars were similar for a single FFQ and 24HDR, but attenuation decreased with repeated 24HDRs. For sugars density, the AFs for FFQ (men: 0.39; women: 0.33) were greater than for single 24HDR (men: 0.30; women: 0.24), and similar to two 24HDRs (men: 0.41; women: 0.35). The attenuation associated with both instruments was greater in women than in men. Both the FFQ and 24HDR were found to be biased; hence, incorporation of the sugars biomarker in calibration studies within the cohorts may be necessary to more reliably estimate associations of sugars and disease. In this article, we propose a new dietary reference instrument based on the recently defined class of predictive biomarkers. Using sugars biomarker, we quantify ME in the FFQ- and 24HDR-reported absolute total sugars and total sugars density.

OPEN about obesity: recovery biomarkers, dietary reporting errors and BMI

Obesity-related under-reporting of usual dietary intake is one of the most persistent sources of bias in nutrition research. The aim of this paper is to characterize obese and non-obese individuals with respect to reporting errors observed with two common dietary instruments, using energy and protein recovery biomarkers as reference measures. This report employs data from the Observing Protein and Energy Nutrition (OPEN) study. Analyses are based on stratified samples of 211 (57 obese) men and 179 (50 obese) women who completed 24-h recalls (24HR), food frequency questionnaires (FFQ), doubly labelled water (DLW) and urinary nitrogen (UN) assessments. In obese and non-obese subgroups, FFQ yielded lower energy and protein intake estimates than 24HR, although biomarker-based information indicated under-reporting with both dietary instruments. Gender differences in obesity-related bias were noted. Among women, the DLW-based energy requirement was 378 kcal greater in obese than in non-obese groups; the FFQ was able to detect a statistically significant portion of this extra energy, while the 24HR was not. Among men, the DLW-based energy requirement was 485 kcal greater in the obese group; however, neither FFQ nor 24HR detected this difference in energy requirement. Combining protein and energy estimates, obese men significantly over-reported the proportion of energy from protein using the 24HR, but not with the FFQ. In obese women, no significant reporting error for energy percent protein was observed by either method. At the individual level, correlations between energy expenditure and reported energy intake tended to be weaker in obese than non-obese groups, particularly with the 24HR. Correlations between true and reported protein density were consistently higher than for protein or energy alone, and did not vary significantly with obesity. This work adds to existing evidence that neither of these commonly used dietary reporting methods adequately measures energy or protein intake in obese groups. The 24HR, while capturing more realistic energy distributions for usual intake, may be particularly problematic in the obese.

Precision of the doubly labeled water method in a large-scale application: evaluation of a streamlined-dosing protocol in the Observing Protein and Energy Nutrition (OPEN) study.

To evaluate whether the doubly labeled water (DLW) method is precise under conditions required for a large-scale evaluation of dietary intake instruments. Energy expenditure was measured in 484 subjects (main study). Subjects received one of five different weight DLW dose bottles prepared in advance of the study. A repeat energy expenditure measure was obtained in a subset of 24 subjects (substudy). DLW measures of energy expenditure were performed over a 2-week interval with urine collection at the beginning and end. Free-living environment with three clinic visits in the Maryland suburban area of Washington, DC. A total of 484 subjects (261 men and 223 women) aged 40-69 y, 24 of whom (13 men and 11 women) participated in a substudy in which DLW was administered a second time. The coefficient of variation of the DLW energy expenditure measurement was 5.1%. This included a 2.9% analytical and a 4.2% physiologic variation. Based on observed initial isotopic enrichment, the preweighed dosages were optimal in 70% of the main study subjects, and 9% received a dose that was less than optimal. Only six subjects (1%) were excluded because the final isotopic enrichment was too low to conduct precise measurement. Use of preweighed DLW dosages did not compromise the precision of the DLW method. The DLW method is a reliable measure of energy expenditure for large-scale evaluations of dietary intake instruments.

Comparison of estimated renal net acid excretion from dietary intake and body size with urine pH

Objective The aim of this study was to examine whether estimated net acid excretion (NAE) from two dietary assessment instruments can predict urine pH in healthy volunteers. Design The Observing Protein and Energy Nutrition (OPEN) study, which was designed to assess dietary measurement error, included men and women aged 40 to 69 years. Participants were asked to complete two food frequency questionnaires (FFQs) and two 24-hour recalls at approximately three-month intervals, and to return two 24-hour urine samples during the same period. NAE was estimated using a previously described equation. The estimates obtained from the different dietary assessment measures were correlated with actual urine pH. Subjects/Setting Four hundred eighty-four subjects were enrolled in the OPEN study. All participants were included in this study on urine pH, given that they had all completed at least one dietary assessment and one urine sample. The study took place in Montgomery County, MD, between September 1999 and March 2000. Results Correlation coefficients between NAE estimates using the FFQ data and the average of two urine pH measures were −0.34 and −0.39 for women and men, respectively. Removing within-person variance in the urine pH levels resulted in a deattenuated overall correlation of −0.50 for the FFQ. Correlations between the 24-hour recall estimates of NAE were −0.41 and −0.32 for women and men, respectively. For the 24-hour recalls, the deattenuated overall correlation was −0.50. Application/Conclusions Our findings suggest that FFQs provide adequate dietary data to estimate renal NAE and could be used as a surrogate for urinary pH levels in large epidemiologic studies. J Am Diet Assoc. 2003;103:1001-1007.

Invited Commentary: OPEN Questions

The papers by Subar et al. (1) and Kipnis et al. (2) in this issue of the Journal provide a unique set of data and a useful contribution to nutritional epidemiology. The Observing Protein and Energy Nutrition (OPEN) Study they report on was based on a reasonable sample of participants, had a remarkably high completion rate, and is the largest known study to use doubly labeled water as the standard for comparison. In addition, the measurements of doubly labeled water, which can be highly error prone (3), appear to have been made with a high degree of technical precision. The study also has several limitations inherent to using this approach, including that it enabled evaluation of only total energy and one specific nutrient, protein. Furthermore, the results cannot be readily generalized to other food frequency questionnaires because the values for total energy intake estimated by use of their questionnaire were substantially lower than typically provided by other full-length questionnaires in current use. In addition, although great expense was devoted to biomarker measurements, only 2 days of 24-hour diet recalls were used, many fewer days of assessment than are included in usual questionnaire validation studies. A more serious flaw in the design of the OPEN Study (1, 2) is failure to include a realistic measure of the withinperson variability of their standard, doubly labeled water plus urinary nitrogen. This measure should have been obtained for a sample of participants by collecting a replicate measure a number of months apart, say roughly 6 to 12, which would be consistent with the 1-year time frame of their questionnaire. In their study, the only measure of reproducibility was for 25 participants, without any interval between the replicates; the authors report a coefficient of variation of 5.1 percent total energy expenditure, suggesting little within-person variation. However, this is likely to underestimate biologic variability substantially in relation to the time frame of the dietary questionnaire. Increasing variability given a greater time interval would be expected, for example, because persons are likely to change their physical activity with time because of differences in seasons and other reasons (the fact that the authors note no average seasonal variation in total energy expenditure is not relevant to individual variability). This phenomenon is well documented in a quantitative review of the reproducibility of total energy expenditure measured by using doubly labeled water in 25 studies with repeated measurements (4). In this report, Black and Cole found that, overall, the within-person coefficient of variation was of similar magnitude to the betweenperson variation whether or not the studies involved an intervention, and it clearly increased with increasing interval between repeated measurements. They estimated that the within-person coefficient of variation was approximately twice as great at an interval of 1 year compared with having no interval between replicates. The lack of a realistic measure of variability in their standard method (1, 2), which is surprising for a study focusing on error, seriously limits the value of the findings and is likely to have led to misleading conclusions. As one example, Subar et al. (1) emphasize the importance of “intake-related bias” associated with their dietary questionnaire or 24-hour recalls, meaning that the ratio of intake measures divided by biomarker measures is inversely correlated with the biomarker measures. Because any unaccounted-for variation (random within-person error) in the biomarker would appear in the denominator of the ratio and the numerator of the biomarker variable, the inverse correlation could just be all or part artifact.

Structure of dietary measurement error: results of the OPEN biomarker study.

Multiple-day food records or 24-hour dietary recalls (24HRs) are commonly used as "reference" instruments to calibrate food frequency questionnaires (FFQs) and to adjust findings from nutritional epidemiologic studies for measurement error. Correct adjustment requires that the errors in the adopted reference instrument be independent of those in the FFQ and of true intake. The authors report data from the Observing Protein and Energy Nutrition (OPEN) Study, conducted from September 1999 to March 2000, in which valid reference biomarkers for energy (doubly labeled water) and protein (urinary nitrogen), together with a FFQ and 24HR, were observed in 484 healthy volunteers from Montgomery County, Maryland. Accounting for the reference biomarkers, the data suggest that the FFQ leads to severe attenuation in estimated disease relative risks for absolute protein or energy intake (a true relative risk of 2 would appear as 1.1 or smaller). For protein adjusted for energy intake by using either nutrient density or nutrient residuals, the attenuation is less severe (a relative risk of 2 would appear as approximately 1.3), lending weight to the use of energy adjustment. Using the 24HR as a reference instrument can seriously underestimate true attenuation (up to 60% for energy-adjusted protein). Results suggest that the interpretation of findings from FFQ-based epidemiologic studies of diet-disease associations needs to be reevaluated.