Real-world outcomes from AgelessRx patients

Read this first

Not a clinical trial. Self-reported at routine check-ins. No placebo, no blinding, no control arm. Descriptive, not proof.

Survivorship bias. Customers who quit rarely check in again, so benefit figures skew toward those who stayed.

Measured against your own start where the check-ins rate outcomes. Improvement means an order rated an outcome better than at its own first check-in, by 30-day bin. Legacy check-ins asked instead whether a member had seen an improvement, and the two are never compared.

Questions differ by product. Similar outcomes are grouped by goal; the exact outcome sits in each product's page.

Side-effect lists differ too. “Never reported” means none of the options on that product's own checklist, at any check-in.

Same numbers as our dashboards. Computed by the Applied Science team's clinical-outcomes notebooks; this page reads them, it does not recompute them.

Refreshed monthly. Regenerated from the medical record, each member counted once. An outcome percentage needs at least 100 answers; side-effect and population figures need 20 members.

Not medical advice. Whether a treatment suits you is for you and a licensed clinician.

Improved vs own baseline. Share of treated orders whose rating of the outcome at that check-in was better than the same order's rating at its first check-in. Ratings that were already at the ceiling on day 0 cannot improve and are excluded. Orders rating the same or worse are in the denominator.

Reported improvement by. Cumulative share of treated orders that had reported an improvement on any outcome question by that check-in, adjusted for orders that were due a check-in and did not respond (the Applied Science team's normalized first-improvement curve). One positive answer on any question is enough and an order stays counted afterwards, so this is a cumulative, best-case estimate. No interval is published.

Reported a first improvement on this outcome by. The same normalized first-improvement estimate restricted to one outcome question.

Rated Good or better. Share of check-in answers in that 30-day bin that rated the outcome Good, Very Good or Excellent, out of answers that gave a rating from Very Poor to Excellent. Not applicable and not-tracking answers are excluded. This is a level, not a change, so it is what a product reports where its check-ins begin at 30 days and there is no day-0 rating to compare against.

Reported improvement on any measured outcome. Every answer to the have-you-seen-an-improvement check-in questions for the product in that 30-day bin, pooled across outcomes: the share that reported a mild, moderate or significant improvement. Not-tracking answers are excluded. Counts answers, so a member answering four questions contributes four times.

Answered positively on any measured outcome. For products whose check-ins rate outcomes or ask about satisfaction rather than asking about improvement: every answer to those questions in that 30-day bin, pooled across outcomes, and the share that was Good or better, satisfied or better, or none-to-mild. Counts answers, so a member answering several questions contributes several times.

Reported an improvement. Share of check-in answers in that 30-day bin reporting a mild, moderate or significant improvement in the outcome since starting, out of answers that gave a verdict (gotten worse, no noticeable improvement, or an improvement). Not-tracking answers are excluded. Asked on check-ins that do not rate the outcome on a Poor-to-Excellent scale; a level, not a change from a day-0 rating.

Reported losing weight. Share of check-in answers in that 30-day bin reporting any weight lost since starting, out of answers that gave a figure (gained, no change, or pounds lost). Not-tracking answers are excluded. A level, not a change from a day-0 rating.

Satisfied or very satisfied. Share of check-in answers in that 30-day bin answering satisfied, mostly satisfied, very satisfied or extremely satisfied, out of all satisfaction answers. Neutral and somewhat satisfied count against this figure.

No or mild symptoms. Share of check-in answers in that 30-day bin rating the symptom none, very mild or mild (or never, rarely, not at all, a little), out of all symptom answers. Moderate or worse counts against this figure.

Two or fewer symptom days a week. Share of check-in answers in that 30-day bin reporting aches and pains on two or fewer of the past seven days, out of all answers to that question.

Says it helps. Share of check-in answers in that 30-day bin saying the medication helps a lot or somewhat and continues to help, out of all answers to that question (including 'it was helpful but is no longer' and 'it never helped'). Trazodone asks this instead of rating the outcome.

Never reported a side effect. Share of members asked the side-effect checklist who did not select any side effect on any check-in for the product. Members who reported one at any check-in count against this figure even if a later check-in reported none. Not published where the checklist has no 'none' option or reached under half of surveyed members.

Reported any side effect. Share of surveyed members who selected at least one side effect on at least one check-in. The complement of never reported a side effect.

Reported this side effect. Share of surveyed members who selected the side effect on at least one check-in. Members can select several, so rates do not sum to the any-side-effect figure.

Worst severity reported. Among members who reported the side effect and rated its severity, the share whose worst rating across all their check-ins fell in each level.

Where the data comes from. Follow-up check-ins in the medical record, processed by the Applied Science team's clinical-outcomes notebooks, one per product family. This page reads the frames those notebooks publish; nothing here is typed in or recomputed.

Side effects. Counted per member across every check-in for the product: a member who selected a side effect once is a reporter even if later check-ins selected none. Severity is the worst level a member ever rated.

Who is counted. Approvals from 1 January 2023. Surveyed members answered at least one check-in. Demographics show shares only under 100 surveyed members.

Every improvement figure has a worsened share beside it: the answers that were lower than the day-0 rating, or said gotten worse, gained weight, dissatisfied or severe. The rest were unchanged.

Check-in response falls with time. Each cell reports the share of all treated orders that answered a check-in in that 30-day bin; orders too recent to have reached the bin are in the denominator, so it understates the true response rate. Headline figures use the latest check-in up to 90 days that at least a quarter of orders answered, or a later one where it clears the same bar.

Intervals are 95% Wilson score intervals for a share. A triangle marks a figure whose interval is wider than 20 points; read it as a rough indication only. First-improvement curves are cumulative estimates and carry no interval.

An outcome percentage needs at least 100 answers. Where a check-in has fewer, no percentage is published for it and the cell shows a dash, because a share of a few dozen answers reads as far more certain than it is. Side-effect and population figures are published from 20 members or 20 treated orders. Demographic shares are shown without counts where a product has fewer than 100 surveyed members.

2026.09 — first version on the internal dashboards' definitions

• Population: approvals from 1 January 2023; surveyed = an answered check-in within 180 days of approval.
• Benefit measured against each order's own first check-in, by 30-day bin, ceiling excluded. Products whose check-ins never ask a rating are reported separately as self-reported change.
• Side effects counted per member: never reported across the whole 180-day history. This reads lower than the "chose none at latest check-in" figure published before September 2026.
• Every figure carries its sample size and a 95% interval. Demographics and primary reason for purchase added per product.

2026.09.16 — figures now come straight from the Applied Science "Clinical Outcomes" notebooks

• The page reads the same per-family Hex notebooks the internal clinical dashboards use (one per product family, 23 in all). The pipeline's own warehouse SQL, taxonomy and answer scorer are retired.
• Product families replace SKUs: variants of one product (injection, nasal spray, patch) are one row, with each variant broken out on the product page.
• Improved vs own baseline is the notebooks' improvement-from-baseline measure by 30-day check-in bin. The self-reported change table is replaced by the notebooks' normalized first-improvement curve, which adjusts for orders that never answered a check-in.
• Side effects use the notebooks' pooled member-level incidence and worst-severity distributions.
• States represented is no longer published (not in the notebook frames).

2026.09.30 — two generations of check-ins, told apart and labelled

• Check-ins exist in two generations. Earlier ones began at day 30 and asked whether members had seen an improvement since starting. Current ones begin with a baseline check-in around day 3 and ask for a rating, so a change can be measured against the member's own starting point. Every figure now says which one produced it, and the page says not to compare a figure measured one way against one measured the other.
• Products carrying both generations publish the current method once it has enough answers, and the legacy method until then, which is what a large older product had been missing: its historical answers were never published, so its figures rested on the small number of members who had reached the newer form.
• An outcome percentage now needs at least 100 answers. Where a check-in has fewer, no percentage is published for it and the cell shows a dash, whose hover says whether the question was never asked or was answered by too few members. Side-effect and population figures keep the lower floor of 20.
• A later check-in past 90 days is published only where enough of that product's orders answered it.
• "Approved members" is no longer shown as a figure. Every published number counts surveyed members, so that is the only population figure quoted; the term is still defined, because the definition of a surveyed member depends on it.
• Check-in columns stop at 90 days unless enough of that product's orders answered a later one, and each column now says how many members answered it.
• Product and goal page introductions rewritten to match the tone of the main page: what we publish and why, including when members do not improve, rather than opening with a count. A member's own reason for taking a product now reads as a sentence ("most often taking it to encourage weight loss").
• The generated reading above each table no longer says an outcome "got worse" where the question only asked members to rate it. A rating of Poor is not a report of getting worse, and only the two change-style questions can be read that way.