Expert nutrition insights, empowering your decisions with data
Guiding evidence-based decisions in nutrition and public health.
Nutrition Evidence Synthesis and Data Analysis (NESDA), led by Dr. Mei Chung, provides
research consulting services that empower nutrition and public health guideline and policy
decisions. With a robust background in nutritional epidemiology and biostatistics, NESDA
strengthens evidence synthesis through methodological research and conducts epidemiological
studies linking diet, lifestyle, and health outcomes.
Dr. Mei Chung is a highly esteemed researcher in the field of nutritional epidemiology
and data science, with a strong background in evidence synthesis.
She leads the way in developing nutrition guidelines and policies that are informed by
robust data. As an Adjunct Associate Professor at the Friedman School of Nutrition Science and
Policy of the Tufts University, she mentors graduate students in conducting
epidemiological research, and teaches evidence synthesis.
Dr. Chung is dedicated to mentoring the next generation of researchers, ensuring they are
well-equipped to tackle the complexities of nutrition and health outcomes. Through NESDA,
she brings that same rigor to consulting work for organizations that need defensible,
evidence-based answers.
Services
Research consulting for organizations developing nutrition guidelines, policies, and programs — from synthesizing what the evidence says, to improving the methods behind it, to generating new evidence from population data.
01
Evidence synthesis for nutrition guidelines
Our expertise in evidence synthesis empowers policymakers and organizations to develop
effective nutrition guidelines. We meticulously analyze and interpret a vast array of
nutritional studies to extract meaningful insights — synthesizing evidence from clinical
trials, observational studies, and epidemiological research into comprehensive reports
that support informed decisions in nutrition policy. Our approach combines rigorous
methodology with practical application, so your guidelines rest on the best available
evidence and ultimately enhance public health outcomes.
02
Methodological research (research on research)
We specialize in developing and refining the methodologies that make evidence synthesis
in nutrition research more effective. Using advanced statistical techniques and innovative
approaches to complex data sets, we ensure findings are robust and reliable. By focusing
on methodological advancement, we improve the quality of evidence synthesis applications,
contribute to a better understanding of dietary impacts on health, and support the
development of effective interventions.
03
Epidemiological studies on diet & health
We conduct in-depth epidemiological studies exploring the relationships between diet,
lifestyle choices, and health outcomes. Using large-scale data sets and advanced
analytical methods, we uncover insights that inform public health recommendations and
interventions — addressing critical questions about dietary patterns, chronic diseases,
and health behaviors, and helping stakeholders navigate the complex interplay between
diet and health.
New Publications
Recent work from NESDA and collaborators. Select a graphic abstract to view it full-size.
Tap to enlarge
Systematic ReviewMeta-AnalysisCognitive Health
The Effects of Carnosine on Cognitive Function and Mental Health: A Systematic Review and Meta-Analysis
Nutrients · 2026 · 18(9), 1385
This systematic review and meta-analysis synthesized 13 clinical trials on L-carnosine
supplementation and cognitive/mental-health outcomes. While most individual trials found
no significant effects on cognition, mood, or quality of life, pooled analyses showed
L-carnosine combined with anserine or antioxidants modestly improved global cognition
(MMSE) and delayed memory (WMS-LM2). Strength of evidence is low — a potential to slow
cognitive decline in older adults, but more high-quality trials are needed.
PFclassify: A New R Package Software for Standardized Classification of Lean and Processed Red Meat and Poultry
Pan Y, Hsiao Y, Wallace TC, Chung M · Journal of Food Composition and Analysis · 2026
The Dietary Guidelines recommend lean red meat and poultry while limiting processed
intake, yet most methods don't distinguish detailed subtypes. PFclassify is a new
open-source R package that classifies intake into detailed lean and processed subtypes
using USDA food codes, decomposing foods recursively to the ingredient level. Validated
with NHANES data, it matches current methods on processed-meat estimates while revealing
subtype patterns across subgroups and over time — enabling more precise, reproducible
nutrition research.
Notes on methods, tools, and the day-to-day craft of nutrition evidence synthesis.
Blog · November 2026
Two Apps for One Systematic Review
Data extraction and risk-of-bias tools built for the lutein and zeaxanthin review
What motivated me to build these apps is simple: I want to streamline and
modernize how I conduct systematic reviews. I used to build customized data
extraction forms in Google Sheets — single extraction with verification by
a second reviewer. It worked, but every tool I looked at had the same gaps:
none of them records a verification log, none accommodates results
extraction for more than two arms, and none collects all the quantitative
data that makes a meta-analysis possible. So I built my own.
For my current project — a systematic review and meta-analysis of dietary
lutein and zeaxanthin intake and visual function — I built two web apps to
do the work in: one for data extraction, one for risk-of-bias assessment.
This post is about why I built them, and how they are different by design.
The review
The review is Dietary Intake of Lutein or Zeaxanthin and Visual Function:
A Systematic Review and Meta-analysis, registered on PROSPERO as
CRD420261491120
(registered 5 October 2026). It asks whether lutein and zeaxanthin from foods
and dietary supplements affect visual function in adults — including
age-related macular degeneration, clinically accepted measures of visual
function, and biomarkers of exposure such as macular pigment optical density.
The evidence is intended to inform an expert panel that will establish
evidence-based recommended intake ranges for these nutrients.
After double independent abstract and full-text screening, the list of
publications was imported into the data extraction app. That is a lot of
structured information to extract, check, and appraise — and it is exactly
the kind of work where a spreadsheet starts to fight you: versions multiply,
and blinding is a matter of trust rather than something the tool enforces.
So I built two apps instead, one for each job.
SR Extractions
The first app is for data extraction. For each study, one reviewer extracts
the study characteristics, the arms, the outcomes, and the results, and a
checker verifies the extraction. That is a deliberate choice: extraction is
mostly transcription and arithmetic, and a careful check catches the errors
that matter. Every change made after a submission is recorded in a
verification log, so the record shows what changed, when, and by whom.
The forms are the review's own — study characteristics, arms with separate
lutein and zeaxanthin doses, outcomes, and results — not a generic template.
All the instructions for how to answer each question are shown on the form
itself, so reviewers see the guidance right where they need it. The app
exports the extracted data as CSV and Excel, ready for analysis.
NUQUEST ROB
The second app is for risk of bias, and it works differently on purpose.
Risk-of-bias assessment is a judgment call, so it gets two independent
assessors. Each assessor reads the paper and rates the 14 NUQUEST questions
across four domains — selection of participants, comparability of study
groups, ascertainment of outcomes, and nutrition-specific items — plus the
four domain ratings and an overall rating, each Good, Neutral, or Poor, by
their own judgment. The app does not calculate scores.
The two assessors cannot see each other's answers until both have submitted.
Then the app opens a compare-and-consolidate view: the two assessments side
by side, differences flagged, discussion threads on each question, and one
consolidated rating agreed by the pair. The discussion is part of the
study's record.
Why two different designs
Extraction and risk-of-bias assessment are different jobs, so the tools are
different. Extraction is single-extractor plus verification: efficient, with
an audit trail. Risk of bias is double independent: two sets of eyes on the
judgment calls, then agreement. The design follows the method, not the
other way around.
What the apps are — and what they are not
Both apps are working tools for this review, built to the review's own
forms. The risk-of-bias app is now public as the NUQUEST ROB app, at
rob.nesda.net:
anyone can create a project and run their own review in it, with their own
studies and reviewers. Access is by invitation — I create each account. The
user manual
is on the NESDA website.
The data-extraction app, SR Extractions, stays private to this review. But if you like the
approach, NESDA builds custom data-extraction forms for systematic reviews
— if you have a review and want a tool shaped around your forms,
email me and we can talk about it.
Where the review stands
Screening is complete and extraction and risk-of-bias assessment are
underway. The review runs until August 2027, and the results will be
published. I will write again about what the tools taught us — what worked,
what the reviewers tripped over, and what I would build differently next
time.
— Mei
News · October 2026
Dr. Mei Chung to present at ISNAMH 2026 in Hong Kong
Nutrition Evidence Synthesis and Data Analysis, L.L.C. (NESDA) founder Dr. Mei Chung
will present at the International Symposium on Nutritional Aspects of Musculoskeletal
Health (ISNAMH 2026), taking place December 7–9, 2026 at the Kerry Hotel in Hong Kong.
On December 8, Dr. Chung will present “Lifestyle factors and peak bone mass –
a scoping review of current evidence from the Bone Health and Osteoporosis Foundation.”
She will be attending the entire conference and welcomes the chance to connect with
longtime collaborators and new colleagues.
Learn more about the symposium.
Coming soon
AI in the research workflow
A new series is in the works: how artificial intelligence can help conduct research —
not as a black box, but as a practical assistant inside a rigorous evidence-synthesis
workflow.
Each post will walk through a real example from NESDA's own projects: where AI sped
things up, where it needed checking, and what a careful researcher should verify
before trusting its output. The first articles are being drafted now.
Data extraction, double-checked
Using AI to pull information from articles, and the verification steps that keep it honest.
In preparation
Writing and debugging R code with AI
Lessons from building analysis tools with an AI pair-programmer.
In preparation
Get in touch
Reach out for expert guidance on evidence synthesis, study methods, or a research
collaboration.
Whether you are developing nutrition guidelines, planning a systematic review, or working
through a complex dataset, NESDA can help you turn the evidence into decisions. The
fastest way to reach Dr. Chung is by email or phone.