Methodology
How type1.science ranks, scores and cites.
type1.science is a living index of the research and tools for type 1 diabetes, organized around four pillars — Cure, Living, Preventing and Trials. The ratings are editorial judgments informed by cited evidence. The criteria, default weights and reasoning are published so you can inspect those judgments. Trials are indexed by their study details, not scored. A dated changelog lists every material change, month by month.
01 / Scoring
A transparent 0–100 score
Within each category, every option is scored 0–100 on a fixed set of published criteria — for CGMs that includes accuracy, lag, interoperability and access; for cures it includes insulin independence, durability and immunosuppression burden. Editors interpret the available evidence to assign each rating. The overall score is a weighted average of those ratings; the formula is reproducible, but the inputs and default weights involve judgment.
A higher score means a more favorable assessment under the selected criteria. A score of 78 is not 78% effectiveness, a 78% chance of success, or a treatment recommendation. Compare scores within the same category and weighting only. Results from different study populations do not establish a head-to-head comparison. Public scores are rounded to whole numbers. Exact weighted arithmetic remains available in the breakdown; it does not imply that level of scientific certainty.
What the ranking can and cannot tell you
The formula is not a validated prediction of which treatment will work best for a person. A one- or two-point editorial gap has no established clinical meaning. Evidence labels are shown separately; the calculation does not apply a hidden evidence-quality multiplier. A future product can receive provisional ratings, but intended features and predecessor results must be identified as such.
Published glucose outcomes depend on the population, starting glucose control, age, target settings and follow-up. Both achieved levels and changes from baseline need that context. Comparing unrelated cohorts does not remove these differences. A product generation cannot inherit a predecessor’s clinical superiority simply because it shares an algorithm family.
The same calculation powers the list and item pages. Lists sort using the result before rounding, with names breaking exact ties. Missing ratings are omitted and the remaining weights renormalized; this is not evidence of good performance. Skipping every criterion restores the defaults. Item pages always show default weights; the Living list identifies your saved personal weights.
02 / Weighting
You decide what counts
There is no single “best.” The cheapest sensor, the most accurate one and the one you can actually get this month are rarely the same product, so the ranking is yours to shape.
Each category has published default weights. In the Living section, you can open Weight the criteria and set how much each one matters — Skip, Some, A lot, or Top. The list re-sorts live and remembers your weighting on your device. Cure and Preventing currently use their published default weights.
AID default: outcomes and everyday life
Revised 16 September 2026: 50% for glucose outcomes and low-glucose protection; 40% for daily experience; 10% for access and cost. Freedom and form factor receive 20%, including tubing, wearability and documented water-use limits. The total glucose weighting is explicit because time in range, average glucose and variability overlap. Exercise spans both daily use and glucose safety; these groups are priorities, not statistically independent measures.
This is an editorial starting point. It has not been validated through a representative user-preference study or shown to be optimal. Change the weights to reflect your priorities. Availability and eligibility still need checking separately, even if a product scores well.
03 / Evidence
Evidence levels
Each item has an evidence label summarizing its overall evidence. This is an editorial classification, not a formal GRADE assessment of every outcome. Individual claims can have different levels of support.
04 / Stage
Development stage
Therapy cards summarize development using a five-step bar. This is a shorthand, not a prediction of success or a fixed timetable:
Preclinical → Phase 1 → Phase 2 → Phase 3 → Approved. Device and observational studies may have no drug phase.
Alongside it, a plain-language status tells you what that means for access today:
05 / Preferences
Preferences stay on your device
The site stores display preferences, saved filters and Living score weights in your browser. There is no account. These preferences are not a clinical assessment, and filters do not establish treatment or trial eligibility. Clearing browser storage removes saved preferences.
Loading the website and opening external sources still sends ordinary web requests to the relevant hosting services. Local preference storage does not mean that browsing creates no network traffic or server logs.
06 / Sourcing
What the dates mean
We prioritize regulator documents, trial registries and original peer-reviewed research. Sponsor announcements and conference abstracts are labeled; their findings may change after full publication. A source resolving successfully is not proof that it supports a claim.
Last reviewed records a content review. For trials, registry checked is when we read the registry, and registry updated is when the registry last posted a change. A recent check can still find an old or unconfirmed record. Estimated completion dates are plans, not results. Overall recruitment can differ from recruitment at a local site.
Age and region filters summarize the records. Full eligibility can depend on diagnosis timing, biomarkers, prior treatment and local requirements. Regulatory approval, commercial supply and public subsidy are separate questions. Confirm current access or enrollment with the relevant provider.
07 / Accountability
AI-assisted, openly attributed
The changelog names the model and reasoning effort used for each refresh, its contributors, material corrections and unresolved verification limits. Model attribution describes how the work was produced; it does not establish clinical expertise or independent review.
Content and scores are AI-assisted and pending independent clinical review. Sources may be incomplete or outdated, and errors are possible. Design concepts are unranked proposals: desired features do not establish product performance. We retain negative and discontinued studies so the index reflects setbacks as well as progress.
08 / Monthly review
A record of checks, not just a new date
Each monthly review should examine existing claims and search for missing developments across therapies, devices, insulins, screening and trials. The review process records sources and exact claims, distinguishes full from partial checks, tracks new candidates and calls for an independent second pass.
Primary outcomes, denominators, adverse events, regulator decisions and local access require separate checks. Automated checks flag incomplete documentation and changed records; they cannot establish that every statement is true or that no research is missing. The changelog records published corrections and material changes. Review dates do not certify every claim; consult the sources and limitations in each record.
Named pipeline products remain in the index with their actual evidence and development status. Where sufficiently assessable evidence is available, they can receive provisional editorial ratings. Where performance is undisclosed or a rating would rely on unsupported assumptions, the record remains visible without a numerical score. Removing a score does not establish poor performance. No results means effectiveness is unknown. Editorial design concepts have no numerical scores and appear separately from product rankings. Report a correction or missing development →
Not medical advice
type1.science is science journalism, not medical advice. It is built to help you ask better questions and understand the landscape — treatment decisions belong with you and your care team.