Evidensity

For every systematic review you commission, ten evidence questions go unanswered.

Every claim, cited

The Problem

The reality of evidence generation in pharma

A single systematic literature review takes an average of 1.72 years and costs approximately $141,000 (Borah et al., 2017). The ten largest pharma companies each publish around 119 per year. But for every SLR you commission, there are dozens of internal evidence questions — portfolio decisions, business development & licensing (BD&L) due diligence, medical affairs queries — that never justify the spend. The evidence base decays the day after completion, and every month spent waiting for evidence is a month patients wait for access decisions.

What if you could run a full systematic review overnight, at a fraction of the cost — and get evidence to decision-makers while it still matters?

The Deliverable

What you receive: an interactive evidence bundle

Every Evidensity engagement is delivered as a single, self-contained interactive bundle — the same format as the two worked examples below. It opens in any browser with nothing to install: browse the full corpus, trace every claim to its verbatim source quote, inspect the per-paper credibility scorecard, and explore the forest plots and pooled estimates. Not a static summary — the complete, auditable analysis, in your hands.

GLP-1 & cardiovascular outcomes → DOACs vs warfarin in atrial fibrillation →

Two live examples — each generated from a single plain-text question, then checked against the published landmark meta-analysis of the same trials.

Live Example · Validated

GLP-1 Receptor Agonists & Cardiovascular Outcomes

A worked example — and a test of the method against a known answer. Generated from a single plain-text question, then checked against the published landmark meta-analysis of the same trials.

Research Question
In adults with type 2 diabetes, what is the effect of GLP-1 receptor agonists versus placebo on the risk of 3-point major adverse cardiovascular events (cardiovascular death, non-fatal myocardial infarction, or non-fatal stroke)?

Explore the interactive evidence bundle →

Browse every paper, trace each claim to its source quote, inspect the credibility scorecard and forest plots — the full analysis, not a static summary.

1,567 Papers Discovered
10+ Databases Searched
184 Screened Through
183 Analysed
19 Meta-Analyses
Key Result — Reproducing a Landmark

Pooled 3-point MACE: 0.86 across eight cardiovascular outcome trials

3-Point MACE — GLP-1 Receptor Agonists vs Placebo Study Estimate [95% CI] Weight Husain et al. (2019) +0.79 [+0.57, +1.10] 2.0% Gerstein et al. (2021) +0.73 [+0.58, +0.92] 4.1% McGuire et al. (2025) +0.86 [+0.77, +0.96] 18.1% Hernandez et al. (2018) +0.78 [+0.68, +0.90] 11.2% Gerstein et al. (2019) +0.88 [+0.79, +0.99] 17.3% Marso et al. (2016) +0.87 [+0.78, +0.97] 18.5% Marso et al. (2016) +0.74 [+0.58, +0.95] 3.6% Holman et al. (2017) +0.91 [+0.83, +1.00] 25.3% Pooled (RE) +0.86 [+0.82, +0.90] I² = 0.9% τ² = 0.0000 Q = 7.06 (df=7, p=0.423) Low heterogeneity PI: +0.86 [+0.81, +0.91] ← Favours treatment Favours control →
Forest plot: 3-point MACE, GLP-1 receptor agonists versus placebo across the eight pivotal cardiovascular outcome trials. REML random-effects pooling — pooled risk ratio 0.86 [0.82, 0.90], I² = 0.9%. Generated automatically from the extracted trial data; independently reproduces Sattar et al. (2021).

Across the eight pivotal GLP-1 cardiovascular outcome trials, our pooled estimates — produced from the plain-text question above, every figure traceable to its source — reproduce the published landmark meta-analysis (Sattar et al., Lancet Diabetes & Endocrinology, 2021) closely across every outcome:

Outcome (pooled risk ratio) Evidensity Sattar 2021
3-point MACE0.86 (0.82–0.90)0.86 (0.80–0.93)
Cardiovascular death0.87 (0.81–0.94)0.87 (0.80–0.94)
All-cause mortality0.89 (0.84–0.94)0.88 (0.82–0.94)
Fatal / non-fatal stroke0.86 (0.76–0.99)0.83 (0.76–0.92)
Fatal / non-fatal myocardial infarction0.86 (0.74–1.01)0.90 (0.83–0.98)
Heart-failure hospitalisation0.92 (0.82–1.03)0.89 (0.82–0.98)

The two columns are independent: ours generated by the pipeline from the question above; Sattar's transcribed from the open-access publication. The headline MACE estimate matches to two significant figures with the same low heterogeneity (I² ≈ 1%); every outcome agrees in direction and magnitude. The wider intervals on MI and stroke reflect fewer poolable arms, not a divergent result.

Live Example · Validated

Direct Oral Anticoagulants vs Warfarin in Atrial Fibrillation

A second worked example — generated from a single plain-text question, then benchmarked against the landmark meta-analysis of the pivotal trials (Ruff et al., Lancet, 2014).

Research Question
In adults with atrial fibrillation, what is the effect of direct oral anticoagulants (DOACs) versus warfarin on stroke or systemic embolism, as reported in the pivotal phase-3 randomised controlled trials?
Key Result — Reproducing a Landmark

Stroke / systemic embolism: a 14% reduction, with zero heterogeneity

Stroke or Systemic Embolism — DOACs vs Warfarin Study Estimate [95% CI] Weight Connolly et al. (2009) +0.91 [+0.74, +1.11] 18.2% Patel et al. (2011) +0.88 [+0.75, +1.04] 27.4% Granger et al. (2011) +0.79 [+0.66, +0.95] 22.6% Giugliano et al. (2013) +0.87 [+0.73, +1.04] 23.9% Fox et al. (2011) +0.86 [+0.63, +1.17] 7.8% Pooled (RE) +0.86 [+0.79, +0.94] I² = 0.0% τ² = 0.0000 Q = 1.22 (df=4, p=0.874) Low heterogeneity PI: +0.86 [+0.75, +0.99] ← Favours treatment Favours control →
Forest plot: stroke or systemic embolism, DOACs vs warfarin across the pivotal randomised trials. REML random-effects pooling — pooled risk ratio 0.86 [0.79, 0.94], I² = 0.0%. Generated automatically from the extracted trial data.

On the outcome that defines the class, the pooled estimate across the pivotal randomised trials is 0.86 [0.79–0.94] with zero between-trial heterogeneity (I² = 0%) — a 14% relative reduction in stroke or systemic embolism, produced from the plain-text question above with every estimate traceable to its source. This reproduces the primary efficacy finding of the landmark meta-analysis (Ruff et al., 2014: risk ratio 0.81 [0.73–0.91]).

Explore the interactive evidence bundle →

How It Works

Not a chatbot. A systematic process.

Research
Question
Systematic
Search
Relevance
Screening
Full-Text
Extraction
Quantitative
Pooling
SLR
Report

20 specialised agents · 10+ academic databases · PRISMA-aligned methodology

What separates this from deep research tools

One corpus, unlimited angles

Once the evidence base is built, follow-up questions come back almost instantly. Need the same corpus reframed for a payer committee? A different therapeutic comparison? A sub-population breakdown? Same evidence, new analysis — overnight.

Output is tailored to audience: Cochrane-style with GRADE certainty ratings for HTA dossiers, executive briefs for leadership, or data-forward insight reports for internal teams.

Use Cases

Three use cases, one principle

Pre-screening before CRO spend or payer submission

Scope the evidence landscape overnight before committing a six-figure engagement or entering a reimbursement negotiation. Know whether the evidence base supports your position before you walk into the room.

Internal evidence scans

Portfolio decisions, BD&L due diligence, medical affairs queries, competitive intelligence. The questions that matter but never justify a formal SLR.

Living updates

Re-run the same search criteria quarterly. Flag new papers that would change conclusions. Keep your evidence base current between formal reviews.

This doesn't replace regulatory-submission SLRs. It handles everything else.

Comparison

The landscape

General AI Research Academic AI Tools Traditional SLR Evidensity
Search Unstructured web Academic databases Systematic, manual 10+ academic databases, systematic criteria
Extraction Prose summary Abstract-level Full-text, manual Full-text structured schemas
Grounding Often hallucinated Citation-level Gold standard Verbatim quote + section
Quantitative Pooling None None Separate biostatistician REML random-effects, forest plots
Credibility None None Manual (RoB-2, GRADE) Multi-dimensional per paper
Timeline Minutes Minutes 1.72 years avg Hours
Cost Low Low ~$141,000 per review Enquire
Output Chat response Paper list Publication-ready report Interactive evidence bundle + report

Strong    Partial    Limited or none

Jack Flynn
Founder, Evidensity

If faster evidence would have changed any access timeline on your watch, I'd welcome the chance to show this to the people still fighting that fight.