> For the complete documentation index, see [llms.txt](https://docs.nested-knowledge.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.nested-knowledge.com/get-started/demo-nests/slr-glp-1-agonists-for-obesity.md). # SLR: GLP-1 Agonists for Obesity A walkthrough of the systematic literature review demo nest on GLP-1 agonists for obesity. ### About This Review Randomized controlled trial evidence on GLP-1 receptor agonists for weight-related outcomes has grown quickly since 2020. This nest is a complete systematic literature review (SLR) asking a precise question: in adults with obesity, do these agents reduce weight, BMI, or waist circumference compared with placebo or active comparators, in RCTs published since 2020? It's built to show what a full-scale, defensible SLR looks like in Nested Knowledge. This includes every stage from search through manuscript, with an AI tool supporting each step and a human reviewer or adjudicator finalizing every decision. If you're evaluating Nested Knowledge for a review that needs to hold up to peer review, HTA scrutiny, or internal audit, this is the nest to open first. Want to move straight to a Q\&A with the evidence? Jump to Synthesis Chat ### Literature Search The review draws from **OpenAlex**, **PubMed**, and **ClinicalTrials.gov**. The sources were searched at different points as the review was updated: November 2025, June 2026, and December 2025, respectively. The query was built outside of Nested Knowledge and inputted as an Automatic Search on scheduled updates. Select **More** to view individual search details, and click **Results: ###** to view the associated studies. One additional record was added through expert recommendation rather than database search. Duplicates are automatically removed upon upload but available for review, see [Duplicate Review.](/review-workflow/literature-search/duplicate-review.md) Combining a broad bibliographic database (OpenAlex), a clinical literature standard (PubMed), and a trial registry (ClinicalTrials.gov) is a common pattern for a comprehensive SLR: bibliographic databases catch published, peer-reviewed evidence, while a registry search surfaces trials that may be completed but not yet published, or reported only in a registry posting. Supplementing with an expert-identified record shows a nest can incorporate items outside a formal search — see [Other Sources.](/review-workflow/literature-search/other-sources.md)

Automatic Search results for the GLP-1 obesity review.

**Try it yourself:** Use [**Smart Search**](/review-workflow/literature-search/ai-tools-in-search/smart-search.md) to input a related research question, and generate search queries. Finalize the query on a new Automatic Search or add a study manually through Other Sources. **See also:** [Literature Search](/review-workflow/literature-search.md), [Automatic Search](/review-workflow/literature-search/add-search/automatic-search.md), [Other Sources](/review-workflow/literature-search/other-sources.md), [Duplicate Review](/review-workflow/literature-search/duplicate-review.md), and [Smart Search](/review-workflow/literature-search/ai-tools-in-search/smart-search.md) ### Screening Exclusion-based Screening criteria was defined. Records were screened separately at abstract and full-text level (**Two Pass Screening**), with two independent reviewers per record and a third reviewer adjudicating any disagreement (**Dual Screening**). After training, **Robot Screener** filled one of the two reviewer seats. Only available in Abstract Screening. This combination matters for a review of this scale and stakes: two independent decisions per record catch errors a single reviewer would miss, and adjudication resolves disagreement with a documented, third-party judgment. Robot Screener demonstrates how an automated reviewer participates the same way a human reviewer does — it answers the same criteria and reaches its own decision, but it never has the final word. A second reviewer and a human adjudicator are still required before any record's fate is settled. **Try it yourself:** Select Adjudicate Screening under Abstract Screening. View a study with two screening decisions under the Abstract Screenings tab to compare human vs Robot Screener decision side-by-side.

Abstract screening decisions from a reviewer and Robot Screener. — Compare reviewer and Robot Screener decisions during abstract screening.

**See also:** [Configure Screening](/review-workflow/screening/configure-screening.md), [Robot Screener](/review-workflow/screening/exclusion-based-screening/robot-screener.md), [Screen Records](/review-workflow/screening/screen-records.md) ### Tagging and Extraction Data extraction combined three approaches: manual tagging, **Core Smart Tags**, and **Adaptive Smart Tags**. Core Smart Tags was ran early, giving an immediate PICO-level read across the whole corpus — useful for prioritizing which records to screen first, well before screening finished. Adaptive Smart Tags handled the heavier lift once studies were included: extracting weight, BMI, and waist circumference values consistently across 125 studies would take a reviewer weeks by hand, and manual extraction at that scale also introduces more opportunity for transcription error. Manual tagging stayed in the loop for judgment calls Smart Tags shouldn't make alone — labeling a subgroup analysis, for instance, or resolving an ambiguous reported value. **Try it yourself:** Examine applied tags using the icon or open a study in Tagging to view extracted data. View all the question tags and their generated evidence alongside the study Abstract or Full Text. Selecting a question auto-scrolls you to where the evidence was collected.

Tagging view showing applied tags for an included study. — Applied tags in the GLP-1 review.

Tagging view showing generated evidence for question tags. — Generated tag evidence alongside the study record.

**See also:** [Configure Tagging](/review-workflow/tagging/configure-tagging.md), [Core Smart Tags](/review-workflow/tagging/ai-tools-in-tagging/core-smart-tags.md), [Adaptive Smart Tags](/review-workflow/tagging/ai-tools-in-tagging/adaptive-smart-tags.md), and [Apply Tags](/review-workflow/tagging/apply-tags.md) ### Meta-Analytical Extraction **Meta-Analytical Extraction** configures each intervention and its data elements, then captures the numeric, arm-level data — means, standard deviations, sample sizes, effect estimates — that Quantitative Synthesis later pools into a single result. **Smart MA Extraction** assists this stage the same way Adaptive Smart Tags assists Tagging: it proposes values pulled from the source text for a reviewer to confirm, rather than entering them unsupervised. Note: this is a separate stage to Tagging. Tagging captures broad descriptive and qualitative detail, including PICO, study design, and adverse event categories. Meta-Analytical Extraction captures the specific numeric fields a pooled analysis needs. This evidence base includes different agents, doses, comparators, and follow-up durations. That separation lets a reviewer configure poolable comparisons, such as semaglutide 2.4 mg versus placebo at 68 weeks, without forcing every included study into one all-or-nothing analysis. **Try it yourself:** Open a study in MA Extraction to see the types of data elements extracted. Head to Quantitative Synthesis to view the automated NMA outputs.

Meta-Analytical Extraction fields for an included trial. — Meta-Analytical Extraction for an included GLP-1 trial.

**See also:** [Meta-Analytical Extraction](/review-workflow/meta-analytical-extraction.md), [Configure Meta-Analytical Extraction](/review-workflow/meta-analytical-extraction/configure-meta-analytical-extraction.md), [Extract Meta-Analytical Data](/review-workflow/meta-analytical-extraction/extract-meta-analytical-data.md), [Smart MA Extraction](/review-workflow/meta-analytical-extraction/smart-ma-extraction.md), [Meta-Analytical Methods](/review-workflow/synthesis/navigate-synthesis-outputs/quantitative-synthesis/meta-analytical-methods.md) ### Critical Appraisal Risk of bias was assessed with the **Cochrane RoB 2** tool, assisted by **Smart Critical Appraisal**. RoB 2 is the validated, field-standard tool for RCT risk of bias — using it, rather than a custom or informal checklist, is itself part of what makes the review's conclusions credible to a reader who knows the methodology. Smart Critical Appraisal speeds the first pass through RoB 2's domains, but as with screening, a human reviewer finalizes every appraisal judgment. 114 studies were appraised as low or some concern for risk of bias; 8 as high concern. **Try it yourself:** Open a study in Critical Appraisal and compare your judgment against Smart Critical Appraisal's initial pass.

Critical Appraisal view with Cochrane RoB 2 assessments. — Cochrane RoB 2 assessments in Critical Appraisal.

**See also:** [Critical Appraisal](/review-workflow/critical-appraisal.md), [Smart Critical Appraisal](/review-workflow/critical-appraisal/smart-critical-appraisal.md) ### Study Inspector Before extracted data moves into Synthesis, use **Study Inspector** to check the included-study set for gaps and inconsistencies. It acts as a quality-control checkpoint. For a review at this scale, a systematic reviewer could use Study Inspector to filter studies missing a full-text PDF before running Bulk Import Full Texts. They could find studies with a specific RoB 2 domain flagged high risk, or confirm which of the 125 included studies still lack completed Meta-Analytical Extraction. They could also export a defined subset, such as the tirzepatide-arm studies, through Download from Inspector for a targeted internal review. None of this changes any study's data on its own — Study Inspector filters and surfaces what's already been entered elsewhere, rather than collecting anything new. That's exactly why it's useful as a final check: it shows a reviewer the state of the whole corpus at once, in a way no single editor does on its own. **Try it yourself:** Filter the 125 included studies down to a single agent or dose, then export that subset through Download from Inspector to see the resulting file.

Study Inspector showing included GLP-1 studies. — Study Inspector for the included-study set.

**See also:** [Study Inspector](/review-workflow/study-inspector.md), [Filters Available in Study Inspector](/review-workflow/study-inspector/filter-studies/filters-available-in-study-inspector.md), [Full Text Import](/review-workflow/screening/full-text-import.md), [Bulk Actions](/review-workflow/study-inspector/bulk-actions.md), [Download from Inspector](/review-workflow/study-inspector/download-from-inspector.md), and [Advanced Study Inspector](/review-workflow/study-inspector/filter-studies/advanced-study-inspector.md) ### Synthesis However a nest is built, opening **Synthesis** surfaces the same set of views automatically, drawing on whatever's already been completed elsewhere in the nest — none of it needs to be built from scratch. **Qualitative Synthesis** generates on its own from applied tags, giving a synthesis-ready view of how those tags translate into findings. The **PRISMA** chart auto-populates the same way from screening decisions, producing a live, auditable count of records at each stage without anyone building it by hand. **Quantitative Synthesis** and **Critical Appraisal** visuals appear alongside these two, but only where that stage was actually completed in the nest — a review that skipped Meta-Analytical Extraction, for instance, simply won't have a Quantitative Synthesis view to show. **Synthesis Chat** is available immediately regardless of how much of the nest is populated — it's a conversational way to query whatever data exists so far, not a view that waits on a specific stage finishing. **Dashboard** is the one exception: it has to be built rather than generated, so it only appears if someone has configured one in this nest. **Smart Insights** are built here– they are generated AI-written text summaries from the extracted data, drawn together into a competitive landscape narrative. In this review **Quantitative Synthesis** was generated which pools the numeric data captured in Meta-Analytical Extraction. Explore how weight loss ranged from roughly 5% to over 20% depending on agent and dose, with waist circumference differences from about −4 cm to −15 cm; gastrointestinal adverse events were the main tolerability signal, more common during dose escalation. Smart Insights' role is to generate a fast first draft of that narrative from the extracted data — a human reviewer still shapes and verifies it before it's finalized, the same as every other AI-assisted step in this nest. View these in Dashboard below. The review also documents specific evidence gaps directly — for example, that JNJ-64565111 5.0 mg and lotiglipron 140 mg were each represented by only one study. Naming a gap like this, rather than omitting it, is exactly the kind of finding a qualitative synthesis is for. **Try it yourself:** Select areas of interest in the sunburst, or compare interventions and reported outcomes in Quantitative Synthesis.

Quantitative Synthesis view comparing GLP-1 interventions. — Quantitative Synthesis of GLP-1 intervention outcomes.

**See also:** [Synthesis](/review-workflow/synthesis.md), [Qualitative Synthesis](/review-workflow/synthesis/navigate-synthesis-outputs/qualitative-synthesis.md), [PRISMA Chart](/review-workflow/synthesis/navigate-synthesis-outputs/prisma-chart.md), [Meta-Analytical Extraction](/review-workflow/meta-analytical-extraction.md), [Quantitative Synthesis](/review-workflow/synthesis/navigate-synthesis-outputs/quantitative-synthesis.md), [Critical Appraisal](/review-workflow/critical-appraisal.md), [Critical Appraisal Outputs](/review-workflow/synthesis/navigate-synthesis-outputs/critical-appraisal-outputs.md), [Synthesis Chat](/review-workflow/synthesis/navigate-synthesis-outputs/synthesis-chat.md), [Dashboard Editor](/review-workflow/synthesis/set-up-synthesis-outputs/dashboard-editor.md), and [Smart Insights](/review-workflow/synthesis/set-up-synthesis-outputs/smart-insights.md) #### Dashboard **Dashboard** is where extracted data in included studies becomes a synthesis-ready summary view — tables and visuals a reviewer configures once extraction is far enough along, rather than a new place to enter data. That distinction is worth stating directly, since it's easy to blur: Search, Screening, Tagging, Meta-Analytical Extraction, and Critical Appraisal are editors — each is where a reviewer records a decision or a value about one study at a time. Dashboard sits on the other side of that line. Configuring a Dashboard view — choosing which data elements to display, and how — is itself a kind of editing, but it never collects new source data; it only aggregates and displays what the editors upstream have already captured. If a number on Dashboard looks wrong, the fix is in the editor that extracted it, not in Dashboard itself. **Try it yourself:** Maximise the study insight cards to interact with the data, or download table cards to view data in Excel.

Dashboard cards summarizing GLP-1 review data. — Dashboard summary of extracted GLP-1 review data.

**See also:** [Dashboard Editor](/review-workflow/synthesis/set-up-synthesis-outputs/dashboard-editor.md) and [Dashboard in Synthesis](/review-workflow/synthesis/navigate-synthesis-outputs/dashboard-in-synthesis.md) #### Synthesis Chat Before working through Dashboard section by section, it's worth trying **Synthesis Chat** directly — a conversational way to query everything this nest has already synthesized. Ask it something concrete, like which studies reported a waist circumference decrease greater than 10 cm, or how tirzepatide's results compare to semaglutide's across the included trials, and it answers from the nest's own extracted data rather than from general knowledge of the topic. If you're exploring this demo rather than running the original review, Synthesis Chat is often the fastest way to get a feel for what 125 studies' worth of extracted data can actually answer. **Try it yourself:** Ask Synthesis Chat a question this walkthrough hasn't already covered, and see how it grounds its answer in specific studies from the nest. **See also:** [Synthesis Chat](/review-workflow/synthesis/navigate-synthesis-outputs/synthesis-chat.md)

Synthesis Chat answering a question about GLP-1 studies. — Synthesis Chat response grounded in the included studies.

### Manuscript **Manuscript** draws the structured content built up across this nest — PRISMA counts, the Critical Appraisal summary, extracted results — into a formatted draft a reviewer completes and edits toward submission. For a review headed toward publication, this is where the earlier stages pay off directly: rather than manually re-copying screening counts into a PRISMA flow diagram or re-typing appraisal judgments into a results table, Manuscript pulls those numbers from where they were already recorded. It's a drafting aid, not a finished paper — a reviewer still writes, edits, and takes responsibility for the manuscript's final language. **Try it yourself:** View and download this review's Manuscript as a word document. **See also:** [Manuscript in Synthesis](/review-workflow/synthesis/navigate-synthesis-outputs/manuscript-in-synthesis.md), [PRISMA Chart](/review-workflow/synthesis/navigate-synthesis-outputs/prisma-chart.md), [Critical Appraisal Outputs](/review-workflow/synthesis/navigate-synthesis-outputs/critical-appraisal-outputs.md)

Manuscript view for the GLP-1 review. — Manuscript draft generated from the review data.

### Next Steps


Start A Pilot Trial	Set up a pilot trial of Nested Knowledge.	Start a Pilot Trial
Core Concepts	Understand the objects and terminology used in Nested Knowledge.	Core Concepts
Nest Setup	Learn how to set up your review in Nested Knowledge.	Nest Setup
Quickstart	Start building your first nest.	Quickstart

--- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.nested-knowledge.com/get-started/demo-nests/slr-glp-1-agonists-for-obesity.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.