Evidence for health effects of what we eat: how to integrate - - PowerPoint PPT Presentation

Evidence for health effects of what we eat: how to integrate findings from intervention and observational studies?

Prof Pieter van 't Veer, PhD Division of Human Nutrition Wageningen University Carla Dullemeijer, Olga W Souverein, Esmée L Doets, Hilko van der Voet, Janneke P van Wijngaarden, Waldo J de Boer, Maria Plada, Rosalie AM Dhonukshe-Rutten, Paulette in `t Veld, Adrienne EJM Cavelaars, Lisette CPGM de Groot

Domain of this presentation

• Knowledge of health effects of what we eat

(somehow) integrated in reference values

• Nutrient reference values serve as tools for

evaluating usual intake of individuals / populations, to provide advice / propose policies

• Focus on deriving nutrient reference values
• EURRECA NoE (2007-12): micronutrients

Reference values for populations and individuals

King et al. 2007

ANR, EAR (Estimated Average Requirement) A daily nutrient level estimated to meet the requirements of half the healthy individuals in a particular life stage and gender group. RDA, RDI, INL97.5 (Recom- mended Dietary Intake) The average daily dietary intake level that is sufficient to meet the nutrient requirements of nearly all (97–98 per cent) healthy individuals in a particular life stage and gender group.

*source: FNB:IOMFood and Nutrition Board: Institute

• f Medicine DRI process

Biological variation: susceptibility

Nutritional adequacy – the health criterion (Gibney)

Criterion of adequacy / health criterion determined by the level of intake that Prevents deficiency symptoms Optimizes body stores of a nutrient Optimizes some biochemical or physiological function Minimizes a risk factor for chronic disease Minimizes the incidence of disease Traditional Physiological / nutritional approach Epidemiology / public health perspective

(From Introduction to Human Nutrition, Gibney et al;(ed))

Methodological scope of presentation

(data from Eurreca on B12 preliminary & for illustration)

Two concepts Two applications Two study designs Dose response Balance Individual Population Intervention Observational

Three “dimensions” of studies on requirements

Study concept: “Dose-response” & “balance”

Dose-response data from one subject Health criterion (physiological requirement) (micro)nutrient intake Required intake Normal function Symptoms, adverse effects On each point on the line, a steady state (balance) is assumed. How to construct the line?

Dose-response: individ. requirements & popul. ref. intake

Marker of ‘status’ (physiological requirement) Health criterion value

Distribution

• f required

intake

P2.5 P97.5

ANR

Intake required to attain level of status marker

SDreq

Q1: Given an agreed value for the health criterion, how much should a (population of) individual(s) eat to attain that value?

Balance approach (“in = out”): example vit. B12

Health criterion

• Netherlands: Maintenance liver stores B12 > 500 g
• France: Compensation of losses via bile
• USA: maintenance haematological status in pernicious anaemia

patients (stable Hb, normal MCV & normal reticulocyte response)

Bioavailability factor

• All countries, assessed by faecal excretion, whole body counting

Add CV=10% to allow for variation of ANR between subjects Sum of losses (faeces, urine, skin, …., periods) + need for growth (fetus, child, pregnancy, lactation, ..) Bioavailability Factor

ANR ≡

Methodological scope of presentation

(data from Eurreca on B12 preliminary & for illustration)

Two concepts Two applications Two study designs Dose response Balance Individual Population Intervention Observational

Three “dimensions” of studies on requirements

Study design: Evidence pyramid (strength of...)

To be inserted

Evidence integrated per study type in pyramid (meta- & pooled analyses) Intervention studies,

• RCT-principle (“causality”)

(e.g., Cochrane approach)

Observational studies “epidemiology” (e.g. WCRF global report diet & cancer)

• Prospective cohort (-nested case

control)

• Case control studies (in dynamic

population)

• Cross-sectional studies (“descriptive”)
• Ecological studies (“aggregate”)

Methods relevant to assess requirements (Gibney)

Deprivation studies (‘depletion-repletion’; haem.status PA) Radioactive tracer studies (‘turnover’, ‘bioavailability’) Balance studies (input = loss in dynamic steady state) Factorial meth. (Σ losses + needs for growth: bile losses) Tissue levels (e.g. liver stores of vitamin B12) Biochem markers (serum/plasma B12, holoTC) Biological markers (early, interm, late endp., e.g. MMA) Animal experiments (all)

(From Introduction to Human Nutrition, Gibney et al;(ed))

Cross-tab “pyramid” by “data for requirements”

Methodology (Gibney) Interven- tion Prosp. cohort Case- control Cross- section Eco- logical Depletion-repletion + (no ran- dom alloc) Radio-active tracers

+

Balance studies

+

Factorial methods

+

Tissue levels

+

Biochemical markers + (RCT)

+

Biological markers + (RCT)

+

Animal experiments + (RCT)

Red: Mainly used in setting reference values (data limited) Box: Seldomly used (plenty of data available) OPPORTUNITY!! NB: “Nutritional status” subjects in RCT <represent. than X-sectional

Study methods by “concept” and “design”

Study concept Intervention design (manipulates the exposure) Observational design (assesses the exposure) Balance

i.e. input = output (under steady state condition)

Depletion-repletion (quasi experimental) Radioactive tracers (whole body count) Factorial methods (faeces, skin, etc) Tissue levels (liver) Example B12 Maintenance haematological status in pernicious anaemia patients (stable Hb, normal MCV & normal reticulocyte response) (US) Biliary stores > 500ug (NL) Biliary losses (FR) Dose-response

between intake and

• utcome (after reaching

steady state condition)

RCT, using supplements, enriched foods, whole diets Cross-sectional studies, intake by dietary assessment Examples B12 Relation to serum/plasma vitB12 levels, methylmalonic acid (MMA), serum/plasma holo-transcobalamin (holoTC)

Independent approaches Same results (ANRs)? Same markers, different designs: How to integrate?

Evidence base “balance approach” (B12)

Method Refs, yr Ntotal, sex Pop, remarks Conclusions Whole body counting 6 1960-68 13 male

Healthy subjects daily loss of total body pool = 0.15% (SE=0.01%)

Bile losses 6 1956-91 38 mixed, not specified

healthy, cholecyst- ectomised, post-

• perative, with(hout)

pernicious anaemia biliary loss roughly estimated as 1.4 ug/day

Maintenance haematological status 1 1958 7 sex not specified

pernicious anaemia patients At 1.4 ug/day 4/7 subjects reached a normal haematological status

Bioavailability 11 1959-01 ≈75 mixed, most unspecified

Healthy subjects. Foods raw/cooked, high/low B12 content BF = 50% range 5-65% High dose: less

Current estimates of ANR: error in estimates

Netherlands

• ANR = (minimum liver store) x (loss by whole body counting) / BF.
• ANR = (>500ug) * (0.15%) / (50%) = 2 ug/d
• CV(estimate) ≈7% / ≈ 10% totals ≈ 12%

France

• ANR = (biliary loss & account for reabsorption via IF) / BF = 2 ug/d
• CV(estimate) ≈ 30% (4 papers) / ≈ 10%, totals ≈ 32%

USA

• ANR = B12/day required to just restore normal haematological

status in PA patients = 1.40 (3 out of 7 normalized)

• “median” of 7 data points, very imprecise, CV = ??

Conclusion

• Liver size, bile loss, IF-binding considered fixed (“known”)
• Between-study variation is usually large (ignored)
• Formula’s not consistent (Fr, NL)
• Errors are lower limits: total error in estimate ANR is large!

Search strategy Eurreca NoE (example vit B12)

“Wide” (sensitive) search

• vitamin B12, “biomarkers”, “health outcomes”
• Aimed at associations “intake” – “status” – “health” (I-S-H)

Selection process

• 5913 papers, screened for in/exclusion criteria
• 903 full text papers screened for in/exclusion
• 84 full text papers screened for I-S, I-H, S-H dose response data

Results

• I/S-H: Cognition (19), Osteoporosis (10), Neurologic disease (7),

Anemia (2)

• I-S: 28 RCT, 21 observational studies (37 and 20 estimates)

Extended evidence base, incl. RCT & X-sectional

Method Refs, yr Ntotal, sex Pop, remarks Conclusions Whole body counting 6 1960-68 13 male

Healthy subjects daily loss of total body pool = 0.15% (SE=0.01%)

Bile losses 6 1956-91 38 mixed, not specified

healthy, cholecyst- ectomised, post-

• perative, with(hout)

pernicious anaemia biliary loss roughly estimated as 1.4 ug/day

Maintenance haematological status 1 1958 7 sex not specified

pernicious anaemia patients At 1.4 ug/day 4/7 subjects reached a normal haemato- logical status

Bioavailability 11 1959-01 ≈75 mixed, most unspecified

Healthy subjects. Foods raw/cooked, high/low B12 content BF = 50% range 5-65% High dose: less

RCTs serum/ plasma B12 conc. 37 3398 (21-217)

24 multivit capsules 13 diet / enriched food dose 2.1 – 1000ug ? Until now, not used

X-section serum/ plasma conc. 20 12,570 (64-2999)

17 X-sectional, 17 baseline from cohort ? Until now, not used

Study means of observational studies

Means of “status” and “intake” in all studies Intake: 7 ug/day Status 300 ug/L Much more interesting: associations, i.e. regression coeff. (or r), if available

NOTE: Weights are from random effects analysis Overall (I-squared = 98.3%, p = 0.000) Dhonukshe-Rutten et al. (caps) Van Vliet et al. (m, spread 1) Van Vliet et al. (f, spread 1) Eussen et al. (multivit caps) Al-Khatib et al. Winkels et al. Shuaibi et al. Weinstein et al. Eussen et al. (vitB12 caps) Garry et al. Manders et al. Telford et al. Tapola et al. De Jong et al. (enr. foods + exerc) Nath et al. Earnest et al. Seal et al. (high) Seal et al. (low) Church et al. Spiller et al. Study Tobin et al. Appel et al. Campbell et al. McMahon et al. Clarke et al. Weight et al. Woods et al. Lewerin et al. Dhonukshe-Rutten et al. (milk) Vogiatzoglou et al. (III) Yang et al. Vogiatzoglou et al. (II) Green et al. Cockle et al. (f) Vogiatzoglou et al. (IV) Grieger et al. Ubbink et al. (vitB12 caps) Wolters et al. Planells et al. De Jonget al (enr. foods) Cockle et al. (m) Waldmann et al. Wouters-Wesseling et al. Bates et al. Van Vliet et al. (f, spread 2) Van Asselt et al. McKay et al. Hoey et al. Tucker et al. Flicker et al. Tucker et al. Yajnik et al. Vogiatzoglou et al (I) Van Guelpen et al. Ubbink et al. (multivit caps) Van Vliet et al. (m, spread 2) 2005 2007 2007 2006 2006 2008 2008 2008 2006 1984 2009 1992 2004 2001 2006 2003 2002 2002 2003 1985 Year 2009 2000 2003 2006 2003 1988 2003 2003 2005 2009 2007 2009 2007 2000 2009 2009 1994 2005 2003 2001 2000 2005 2002 2003 2007 1998 2000 2007 2004 2008 2000 2007 2009 2005 1994 2007 RCT RCT RCT RCT Obs RCT Obs Obs RCT Obs RCT RCT RCT RCT Obs RCT RCT RCT RCT RCT Design RCT RCT Obs RCT RCT RCT Obs RCT RCT Obs Obs Obs RCT RCT Obs RCT RCT RCT Obs RCT RCT Obs RCT Obs RCT Obs RCT Obs RCT RCT Obs RCT Obs Obs RCT RCT 0.16 (0.12, 0.19) 0.15 (0.11, 0.19)

• 0.01 (-0.41, 0.40)

0.31 (-0.17, 0.78) 0.23 (0.21, 0.25) 0.02 (-0.02, 0.05) 0.41 (0.31, 0.51) 0.19 (0.03, 0.34) 0.15 (0.04, 0.25) 0.20 (0.17, 0.22) 0.29 (0.24, 0.33) 0.30 (-0.01, 0.60) 0.18 (0.12, 0.24) 0.03 (-0.11, 0.18) 1.58 (0.88, 2.28) 0.09 (0.02, 0.17) 0.11 (0.09, 0.12) 0.21 (0.15, 0.26) 0.17 (0.05, 0.29) 0.10 (0.06, 0.13) 0.06 (0.01, 0.11) Beta (95% CI) 0.12 (0.07, 0.17) 0.90 (-0.16, 1.96) 0.05 (0.01, 0.09) 0.15 (0.13, 0.16) 0.15 (0.13, 0.17) 0.01 (-0.04, 0.05) 0.00 (-0.06, 0.06) 0.12 (0.09, 0.14) 0.15 (0.11, 0.19) 0.07 (0.03, 0.12) 0.00 (-0.11, 0.12) 0.08 (0.05, 0.11) 0.17 (0.14, 0.20) 0.24 (0.14, 0.33) 0.15 (0.09, 0.20) 0.32 (0.22, 0.43) 0.12 (0.06, 0.19) 0.27 (0.19, 0.35) 0.09 (0.04, 0.13) 0.82 (0.53, 1.10) 0.06 (-0.04, 0.16) 0.00 (-0.07, 0.07) 0.90 (0.48, 1.33) 0.03 (-0.01, 0.07) 0.51 (0.10, 0.93) 0.54 (0.27, 0.81) 0.86 (0.54, 1.18) 0.00 (0.00, 0.00) 0.41 (0.21, 0.61) 0.17 (0.15, 0.19) 0.21 (0.19, 0.23) 0.13 (0.06, 0.20) 0.12 (0.08, 0.15)

• 0.01 (-0.11, 0.09)

0.14 (0.09, 0.19) 0.07 (-0.33, 0.48) 0.16 (0.12, 0.19) 0.15 (0.11, 0.19)

• 0.01 (-0.41, 0.40)

0.31 (-0.17, 0.78) 0.23 (0.21, 0.25) 0.02 (-0.02, 0.05) 0.41 (0.31, 0.51) 0.19 (0.03, 0.34) 0.15 (0.04, 0.25) 0.20 (0.17, 0.22) 0.29 (0.24, 0.33) 0.30 (-0.01, 0.60) 0.18 (0.12, 0.24) 0.03 (-0.11, 0.18) 1.58 (0.88, 2.28) 0.09 (0.02, 0.17) 0.11 (0.09, 0.12) 0.21 (0.15, 0.26) 0.17 (0.05, 0.29) 0.10 (0.06, 0.13) 0.06 (0.01, 0.11) Beta (95% CI) 0.12 (0.07, 0.17) 0.90 (-0.16, 1.96) 0.05 (0.01, 0.09) 0.15 (0.13, 0.16) 0.15 (0.13, 0.17) 0.01 (-0.04, 0.05) 0.00 (-0.06, 0.06) 0.12 (0.09, 0.14) 0.15 (0.11, 0.19) 0.07 (0.03, 0.12) 0.00 (-0.11, 0.12) 0.08 (0.05, 0.11) 0.17 (0.14, 0.20) 0.24 (0.14, 0.33) 0.15 (0.09, 0.20) 0.32 (0.22, 0.43) 0.12 (0.06, 0.19) 0.27 (0.19, 0.35) 0.09 (0.04, 0.13) 0.82 (0.53, 1.10) 0.06 (-0.04, 0.16) 0.00 (-0.07, 0.07) 0.90 (0.48, 1.33) 0.03 (-0.01, 0.07) 0.51 (0.10, 0.93) 0.54 (0.27, 0.81) 0.86 (0.54, 1.18) 0.00 (0.00, 0.00) 0.41 (0.21, 0.61) 0.17 (0.15, 0.19) 0.21 (0.19, 0.23) 0.13 (0.06, 0.20) 0.12 (0.08, 0.15)

• 0.01 (-0.11, 0.09)

0.14 (0.09, 0.19) 0.07 (-0.33, 0.48)

• 1

1 2

Effect size: Forest plots RCTs and observational

Cross- sectional RCT

Dose response meta-analysis B12 (“intake-status”)

RCTs Cross- sectional RCT & X- sectional combined

Preliminary regression slopes (stratified analysis)

Pooled βoverall (95%CI) Pooled βadults

1

(95%CI) Pooled βelderly

2

(95%CI) RCT’s 0.17 (0.15-0.19) 0.14 (0.10-0.18) 0.19 (0.16-0.22) Observational 0.09 (0.04-0.14) 0.08 (0.01-0.14) 0.13 (0.04-0.21) Combined 0.16 (0.12-0.19) 0.11 (0.07-0.15) 0.18 (0.15-0.20)

• Regression line loge(conc) = 5.31 + 0.16*loge(intake)
• All association (slopes) highly significant
• Intake 4 to 11 ug serum/plasma conc 253 to 297 pmol/L(i.e. +16%)
• Stratified analysis
• Observational weaker than RCTs; combined close to RCT
• Suggests tronger association in elderly

Dose response meta-analysis B12 (“intake-status”)

RCTs Cross- sectional RCT & X- sectional combined

Traditional ANR >> “projected ANR” based on regression line

Trad ANR±20%

How to derive pop ref intakes using RCT and

• bservational data?

Extended meta-analysis for Intake-Status Traditional meta-analysis: focus on one parameter (e.g. beta Intake-Status, “deterministic”) Extended meta-analysis: estimate the bivariate distribution in a target population (“stochastic”)

• Intake-status relation (βSI or ρSI)
• Intake distribution (µI, σI)
• “Status” distribution (µS, σS)
• These 5 define the bivariate distribution I-S

intake status

Correlations instead of slopes (95 % conf ellipses)

Heterogeneous results for r (or B,Sx,Sy) repre- sented by ellipses (as for individuals!) “meta-analysed” r = 0.38 (correlation

• f ln(intake) with

ln(‘status’)

Observational data (o) and RCT data (+)

Extended meta-analysis

• r=0.29
• byx = 0.16
• bxy
• 1 = 0.16 / 0.292

= 1.90

In a scenario with increased intake, which line to follow?

• Method 2: Predict

status | intake (public health scenarios)

• Method 1: Predict

required intake | clinical health endpoint

Extrapolation method 2: “status | intake”

data Health criterion (physiological requirement) ≈ 5.2% This mean intake is sufficient for 97.5% of subjects Projected “optimal scenario” INL2.5.....ANR .....INL97.5 Projected “requirement distribution”

Extrapolation method 2: “status | intake”

If we extrapolate to sufficient “status” for 97% of people

• Prevalence reduces but mean intake has to increase to ca 10 ug
• Extrapolation is in the range of observational and RCT data

If we extrapolate the line to “prevalence insufficiency”=50%

• this “projected ANR” is much less than the current ANR
• Although the observed CV is very large, the “projected RDA” will

probably be >>> 20% above the “projected ANR”

• Note: this extrapolation is outside the range of observed data, it is

very much to the extreme lower end of the observations (and this with large imprecision)

Extrapolation meth. 1 (intake | “borderline” status)

• Corresponds to question: What intake is

required to just prevent the adverse health

• utcome in 50% of the people:
• Given marginal health, what is the

corresponding intake of that population?

• Unlike extrapolation method 2, this method

would lead to an ANR and RDA above the current values

vitB12 (log scale) Intake vitB12 (log scale) Intake

Comparison “balance” and “dose response”

“Balance”

• based on very

limited data (24 studies, > 100 subj), close to clinical endpoint

• high imprecision of

estimate, CV- between persons of 10% pragmatic.

• evaluation of intake

beyond range of

• bservations

“Dose response”

• > 2 x the # of studies (57),

and 100 x the # subjects (>15,000), healthy people

• projected ANR (RDA) could

be lower or higher, depending on model assumptions

• evaluation of intake: predict
• prev. of inadequacy (in range
• f observ.)

need to reconsider the concept “intake|status” (clinical)

• r “status|intake” (PH) and “health criteria” ??

Further work / work in progress

• Statistical model: Refine estimates, try to reduce between-study

variation, understanding the extrapolation methods, individual data!

• Apply to Eurreca databases: folate, iodine (if possible iron, zinc).

Regarding the size of evidence base similar results anticipated. Findings on ANRs and scenario’s generalizable to other micronut’s?

• Discuss health criterion (150ug/L serum/plasma), more clinically

relevant endpoints (add cognition, osteoporosis, neurologic disease, to pernicious anemia & liver stores). What is “nutritional health”?

• Extend model I-S to I/S-H (trivariate), but so far lack of sufficient

data on hard health endpoints (disease risk, suboptimal function). Here imprecision is a big issue, as with the traditional approach.

• Discuss methodol. and prelim results in scientific meetings (here &

now) and with recommendation setting bodies (workshop < mid ’12)

Thanks for your attention

Carla Dullemeijer, Olga W Souverein, Esmée L Doets, Hilko van der Voet, Janneke P van Wijngaarden, Waldo J de Boer, Maria Plada, Rosalie AM Dhonukshe-Rutten, Paulette in `t Veld, Adrienne EJM Cavelaars, Lisette CPGM de Groot, Pieter van ’t Veer and all other co-workers in the EURRECA NoE