The Arithmetic of Curing Everything

The shape of what we do not have

According to the World Health Organization's ICD-11, NORD (the National Organization for Rare Disorders), and the Monarch Initiative's computational harmonization of major disease knowledge bases, there are roughly 10,000 distinct, clinically recognized diseases.123 Fewer than 5% of them have an approved therapy.2

Most of what we can treat sits at the head of the disease distribution — hypertension, diabetes, hyperlipidemia, depression, arthritis. Most of what we cannot treat lives in the tail: roughly 7,000 rare diseases, each affecting fewer than 200,000 Americans by the Orphan Drug Act's definition,5 the vast majority with no approved therapy at all.

That sounds like a narrow problem, because any single rare disease is small. It isn't. Collectively, rare diseases affect about 1 in 10 Americans — roughly 30 million people,26 nearly as many as the 40.1 million Americans living with diabetes.7 The diseases are rare; the population living with one of them is not.

Will we close the gap? When? Not this century, and probably not the next. At the current rate of genuinely additive novel approvals — somewhere around 10 to 15 per year — covering the remaining ~9,500 conditions takes 600 to 900 years.◊ Christopher Austin, when he led NCATS, put the rare-disease version of the same math at "over 2,000 years."9

Why haven't we, already? Because the system was not built for the long tail. Pharmaceutical R&D runs on market-sized bets: a drug addressing 50 million patients can justify a billion-dollar development program; a drug addressing 5,000 cannot.4 The Orphan Drug Act of 1983 tried to correct for this with tax credits, user-fee waivers, and 7 years of market exclusivity, and it worked — for the subset of rare diseases commercially viable even with subsidies. But the economic logic of the blockbuster drug still pulls R&D toward conditions that already have treatments, where the fifth PD-1 inhibitor for lung cancer can earn more than a first-ever therapy for a disease few have heard of. Federal basic-science funding has not tracked changes in disease burden either; the research–disease mismatch has widened rather than narrowed.

The rest of this section is a deeper look at the disease landscape itself, because the shape of the distribution is what makes the problem hard.

The figure makes the concentration visible. The top 10 conditions — hypertension, obesity, hyperlipidemia, anxiety, arthritis, chronic pain, depression, substance use, migraine, diabetes — account for roughly 82% of all person-diagnoses in the US. The top 25 get you to ~93%. The remaining ~9,975 conditions, most of them rare, make up the long tail — where nearly all of the untreatable disease sits.

The head and the tail are different problems requiring different tools. Progress at the head is incremental: better drugs, better diagnostics, better adherence for conditions we already know how to treat imperfectly. Progress in the tail is categorical: moving a disease from the "zero approved therapies" column to the "one approved therapy" column. The two kinds of progress move different numbers. Only the second kind shrinks the 9,500-disease gap.

The current rate of progress

The natural next question is: how fast are we filling the gap? The cleanest measure is FDA novel drug approvals — new molecular entities and new biologics entering the US market for the first time. It is not a perfect proxy (a novel approval isn't the same as a first-ever treatment for a previously untreatable condition), but it is the number that shows up in almost every policy and investment conversation.

Here is the last forty years.

The dark ages, 2002–2010: the Vioxx story

The trough in the middle of that chart has a name. Vioxx (rofecoxib) was a Merck painkiller approved by the FDA in May 199910 and aggressively marketed for arthritis, acute pain, and migraines. It was a COX-2 inhibitor — a newer class of NSAID designed to relieve pain with fewer gastrointestinal side effects than traditional anti-inflammatories. By the time Merck voluntarily pulled it from the market on September 30, 2004, more than 80 million patients had taken it and annual sales had topped $2.5 billion11 — the largest prescription-drug withdrawal in history. A Lancet analysis estimated that Vioxx caused roughly 88,000 heart attacks and 38,000 deaths in the US alone;12 an FDA drug-safety analyst's estimate ran higher, with as many as 139,000 serious cardiovascular events and 55,000 premature deaths attributable to the drug.13 Internal documents later showed Merck had downplayed cardiovascular signals visible in the VIGOR trial as early as 2000; the confirmatory APPROVe trial in 2004 forced the withdrawal.14

The regulatory reaction was severe. The FDA tightened safety-trial requirements, pharma companies consolidated, and the industry retreated into an institutional crouch — the defensive posture an organization adopts after a high-profile failure, where avoiding the next scandal becomes more important than approving the next therapy. Risk-averse me-too drugs crowded out novel mechanisms. The gray bars in the middle of the chart are the ~22-per-year output of that crouch: fewer approvals, fewer new mechanisms, fewer first-in-class therapies. A decade of post-Vioxx caution.

The modern acceleration, 2014–present: policy that worked

The green bars on the right of the chart tell a different story — one worth dwelling on. Approvals roughly doubled to ~47 per year, and the change is traceable to a specific set of policy choices:

The FDA Safety and Innovation Act of 2012 (FDASIA) created the Breakthrough Therapy designation, giving drugs that show early signs of substantial improvement over existing therapy an expedited pathway with intensive FDA guidance.15 Through June 2024, the FDA had received 1,516 breakthrough-designation requests, granted 587, and approved 317 breakthrough-designated products16 — hundreds of therapies that moved from lab to patient faster because of a single piece of legislation. FDASIA also broadened the 1992 Accelerated Approval pathway17 and codified Priority Review. The Orphan Drug Act incentives, created in 1983, matured over the same period: roughly half of 2025's novel approvals received orphan designation.18

This is the part that deserves to be stated plainly. Legislative choices saved lives at scale. Policies designed in committee rooms translated, within a decade, into tens of thousands of patients receiving therapies they would otherwise have waited years for — or never received. The 2014–present acceleration is concrete evidence that we can decide to improve millions of lives. The system responds to the rules we write for it.

The sobering part

Even at the new, higher rate, the raw approval count understates the progress problem. In 2025, CDER approved 46 novel drugs; 23 received orphan designation and 20 were classified as first-in-class.18 In 2024 the total was 50,19 in 2023 it was 55.20 But "first-in-class" means novel mechanism — the nth PD-1 inhibitor for lung cancer can be first-in-class by target while adding nothing to the disease-coverage count. The net rate of conditions moving from "untreatable" to "treatable" is closer to 10–20 per year.

Keep the scale in view. There are roughly 30 million Americans living with a rare disease, and fewer than 5% of those rare diseases have any approved therapy.2 That is the untreated population we are trying to address. At the genuinely additive rate of ~10–15 new disease-coverings per year,◊ the 30M Americans in the rare-disease tail are, effectively, waiting centuries.

The pipeline we actually have

The pipeline is the leading indicator. Every drug approved ten years from now is in a lab somewhere today. Citeline's Pharmaprojects database tracked 22,825 active drugs in the global R&D pipeline at the start of 2024, up 7.2% year-over-year,21 with roughly 22,940 at the start of 2026.22 Somewhere between 10,000 and 12,000 of those are in active clinical development (Phases I–III);21 the rest are preclinical.

But the pipeline is heavily front-loaded. Most drugs are in early stages, and the overwhelming majority of them will never be approved. This is the attrition funnel — and its shape is the single most important thing to understand about drug development.

Three things jump out of the funnel. First, the attrition is ruthless and getting worse. A new drug entering Phase I today has a 6.7% chance of ever being approved — down from 10.4% in 2014.23 Phase II is the toughest hurdle at just 28% survival;23 biology reveals itself there, and it is usually unkind. Second, the pipeline is large enough to produce roughly 2,000 eventual approvals from current programs over the next decade◊ — which sounds substantial. Third, and most importantly, most of those approvals will not address new diseases.

The novelty mix inside the pipeline is where this gets uncomfortable. Only a minority of programs are first-in-class by mechanism, and the majority are best-in-class follow-ons, new indications for existing drugs, or biosimilars and reformulations. And first-in-class is itself generous — it means novel mechanism, not novel disease. The fifth PD-1 inhibitor for lung cancer can be first-in-class by target while adding nothing to the disease-coverage count.◊

The honest math

Put the numbers together. If we need therapies for the ~9,000 diseases we currently cannot treat, and the rate of genuinely additive approvals is 10–15 per year,◊ covering the remaining disease space takes 600 to 900 years.◊ This is the same order of magnitude as Christopher Austin's "over 2,000 years" estimate for rare diseases specifically.9 Even under optimistic projections — pipeline growth, AI-driven efficiency gains pushing the additive rate to 25–30/year by the late 2030s◊ — you still arrive at 300+ years without a paradigm shift.

The cost arithmetic is similarly grim. The most widely-cited estimate comes from Joseph DiMasi and colleagues at the Tufts Center for the Study of Drug Development: the capitalized cost of developing a new drug that reaches market is $2.558 billion (2013 dollars), rising to $2.87 billion when post-approval R&D is included.24 Only about 9.6% of drugs entering clinical trials ever receive FDA approval.24 A more recent ecosystem-wide analysis by Amitabh Chandra and colleagues, covering the full biopharma R&D landscape (not just the top 20 companies), found that total global biopharma R&D investment reached $276 billion in 2021 — which works out to roughly $5 billion in total ecosystem spending per approved drug.25 At those unit costs, brute-forcing therapies for the remaining 9,000 diseases would run into the tens of trillions of dollars◊ — decades of current global biomedical R&D spending.

Put that way, the answer to "how long would it take to cure everything?" is: longer than civilization has existed, at a cost larger than any war we have ever fought. This is not a minor inefficiency. It is a paradigm that does not work at the scale of the problem.

What would actually get us there

The roadmap is not primarily about spending more money linearly. It requires structural shifts in how therapeutics are developed, approved, and paid for. Six leverage points matter.

A realistic thirty-year scenario

Here is what the next three decades could look like if the leverage points above compound on each other. What follows is entirely speculative — an optimistic but not fantastical scenario where AI delivers, platforms scale, and regulators adapt. No part of this section is drawn from a published forecast.◊

Total cost across thirty years: $3–5 trillion, or roughly $100–170 billion per year globally. That is a 30–50% increase over current biomedical R&D spending — but directed dramatically more efficiently. It is not a fantasy number. It is well within the range of what coordinated effort can mobilize.

The critical variable is not money. It is whether AI actually delivers the 3–5× efficiency gains current trajectories suggest it might, and whether platform therapies scale the way early gene-therapy programs hint they can. If both hold, curing the overwhelming majority of human disease is a within-a-generation problem. If they don't, we are looking at 50–100+ years at much higher cost — still finite, but far enough out that most people reading this will not see the end of it.