FDA Boxed Warnings by the Numbers (2026): 1,084 Ingredients, NSAIDs Lead

A regulatory-affairs or drug-safety lead planning a label strategy must understand the FDA's most severe safety-related labeling instrument: the boxed warning. Commonly referred to as a "black box warning," this notification is placed at the very top of a drug’s prescribing information to alert clinicians and patients to serious risks, such as life-threatening adverse events, potential misuse, or clinical limitations. A registry-quantified analysis answers the question directly: a reproducible scan of the openFDA drug-label database (snapshot exported 25 July 2026) identifies 32,974 label records carrying a boxed warning out of 260,986 registered labels — roughly one in eight — spanning at least 1,084 distinct normalized active-ingredient names.

That 1,084 figure is a floor, not a census, and it is significantly broader than the commonly cited "over 400 drugs" figure repeated in consumer health portals and medical references. Two things drive the gap. First, the denominators differ: the legacy "400+" figure counts unique prescription active moieties, while the label registry counts every ingredient name attached to any registered label, including hundreds of generic and repackager versions of the same molecule. Second — and this is the part no SERP result discloses — only 12,683 of the 32,974 boxed-warning records (38.5%) carry a harmonized openfda ingredient name at all; the remaining 20,291 records have an empty openfda block and cannot be mapped to an ingredient. Every ingredient and class number below therefore describes that named 38.5% subset, which means the true ingredient count is higher than 1,084, not lower.

Within the named subset, a small group of high-risk drug classes dominates. Antidepressants lead with 1,491 boxed-warning label records, followed by nonsteroidal anti-inflammatory drugs (NSAIDs) with 1,386, opioids with 689, antipsychotics with 681, hormones and estrogens with 551, and diabetes agents with 408.

Understanding these concentration patterns, the historical approval dynamics, and the regulatory pathways for adding or revising warnings is essential for pharmaceutical strategists managing postmarket safety and brand positioning.

---

How many drugs carry an FDA boxed warning?

To understand the prevalence of boxed warnings, one must distinguish between individual label records and unique active ingredients. The FDA registers every package size, strength, and manufacturer-labeler combination as a separate Structured Product Labeling (SPL) document.

A scan of the openFDA drug-label registry reveals:

Three method points matter before anyone quotes these numbers.

The record count is not a product count. The FDA registers every package size, strength, and manufacturer-labeler combination as a separate Structured Product Labeling (SPL) document, so one molecule can generate hundreds of records. Levothyroxine alone accounts for 285 boxed-warning label records.

The ingredient count is a lower bound. The 1,084 figure is derived only from the 38.5% of boxed-warning records that carry a harmonized openfda.generic_name. The other 20,291 records still contain SPL product data elements and package display panels, but they were never mapped to an NDC product entry, so their ingredients are invisible to this cut. Any ingredient that appears exclusively in unharmonized labels is missing from the 1,084.

The 1,084 entries are ingredient names, not pure moieties. 298 of them are combination strings that the normalization step keeps intact — abacavir and lamivudine, hydrocodone and acetaminophen, and long prenatal-vitamin ingredient lists all count once. Splitting those strings into their parts and re-merging yields 1,065 distinct moieties, so the two numbers happen to land close together, but they are not the same measurement.

The gap between the registry count and the common "400+ drugs" figure is therefore driven by generic multiplication and naming scope — not, as is often assumed, by over-the-counter products:

Note the OTC row, because it corrects a common assumption. Only 19 of the 32,974 boxed-warning records are OTC products (0.06%). Over-the-counter NSAIDs carry their cardiovascular and gastrointestinal risk language in the Drug Facts "Warnings" section, not in a boxed warning — so the 1,386 NSAID records counted here are prescription products, not the ibuprofen on a pharmacy shelf.

For regulatory strategists, both numbers are valid in their own context. If you are analyzing innovator prescription drug development, the 400+ moiety baseline is the relevant benchmark. If you are analyzing clinical and supply-chain risk exposure — where generic manufacturers and repackagers dominate the label population — the registry cut is the operational baseline, provided you carry the 38.5% coverage caveat with it.

---

Which therapeutic classes have the most boxed warnings?

Boxed warnings are not evenly distributed across the pharmacopeia. Instead, they are heavily concentrated in a few high-exposure therapeutic classes, driven by class-wide regulatory determinations.

The top class buckets by label-record count show this concentration. The final column is the one worth studying: it divides records by distinct ingredients and separates class breadth from generic multiplication.

Each record is assigned to at most one bucket, matched on the label's generic-name field, and shares are computed against the 12,683-record named subset — not against all 32,974 boxed-warning records. Nine buckets covering 49.1% of the named subset means the other half is a long tail: oncology cytotoxics, immunosuppressants, anticoagulants, anesthetics, and hundreds of single-product warnings.

The safety rationale behind each bucket differs, and the rationale is what determines whether a new asset inherits the warning:

Read the two tables together and the operational picture sharpens. The "Hormones / estrogens" bucket is the broadest class here — 58 distinct ingredients — but has the lowest label multiplication (10 labels per ingredient), because it aggregates three unrelated warning rationales across estrogens, contraceptives, and testosterone. Thyroid is the mirror image: just three ingredients generating 328 records, because levothyroxine is one of the most heavily genericized molecules in the United States. Neither class is "more dangerous" than the other; the record counts are measuring supply-chain structure as much as safety.

This class concentration is a crucial consideration for clinical and safety teams. If a company is developing a new asset in one of the genuinely class-wide categories, a boxed warning should be treated as a high-probability regulatory outcome rather than an exception.

For example, the antidepressant class warning (1,491 records) applies to all selective serotonin reuptake inhibitors (SSRIs), serotonin-norepinephrine reuptake inhibitors (SNRIs), tricyclic antidepressants, and atypical agents. Even if a novel CNS drug demonstrates a cleaner clinical profile in trials, the FDA's historical precedent is to mandate the class-wide suicidality boxed warning for all drugs approved under the antidepressant class.

Similarly, within the antipsychotic category (681 records), older typical antipsychotics (like haloperidol) and newer atypical agents (like aripiprazole and olanzapine) share a severe class-wide warning regarding increased mortality when used off-label to treat dementia-related behavioral disturbances in elderly patients. Furthermore, clozapine represents a unique intra-class peak, carrying five separate boxed warnings (severe neutropenia; orthostatic hypotension, bradycardia and syncope; seizures; myocarditis and cardiomyopathy; and increased mortality in elderly patients with dementia-related psychosis).

Which individual ingredients carry the most boxed-warning labels?

Aggregating by class hides which specific molecules dominate the registry. Ranking the named subset by individual normalized ingredient produces a leaderboard that looks nothing like a "most dangerous drugs" list — and that is precisely the point:

The top of this list is dominated by cheap, high-volume, heavily genericized primary-care medicines — levothyroxine, metformin, lisinopril, losartan — not by oncology or specialty products. A boxed warning on levothyroxine is not a signal that thyroid replacement is unusually hazardous; it is a 1970s-era statement about misuse for weight loss that now propagates across 285 separate generic labels.

The operational lesson for anyone building a label-risk screen: rank by ingredient, never by label-record count. Counting records ranks manufacturers' filing behavior. Counting ingredients ranks regulatory decisions.

---

Why do antidepressants and NSAIDs dominate the boxed-warning count?

The sheer volume of antidepressant (1,491) and NSAID (1,386) records in the registry is driven by two factors: high prescribing volume and class-wide regulatory mandates.

The Antidepressant Suicidality Warning

In 2004, following a series of pediatric clinical trial reviews, the FDA issued a class-wide boxed warning for all antidepressants, highlighting an increased risk of suicidal ideation and behavior in children and adolescents. In 2007, this warning was extended to young adults aged 18 to 24.

Because antidepressants are among the most widely prescribed drug classes in the United States, and because there are dozens of distinct molecules (and hundreds of generic versions across multiple strengths and repackagers), this single class-wide mandate generated thousands of individual label changes, inflating the registry's warning count.

The NSAID Cardiovascular and GI Warning

In 2005, in the wake of the withdrawal of rofecoxib (Vioxx) due to cardiovascular safety concerns, the FDA mandated a class-wide boxed warning for all prescription nonsteroidal anti-inflammatory drugs (NSAIDs). This warning alerts prescribers to:

• Cardiovascular Risk: Increased risk of serious cardiovascular thrombotic events, myocardial infarction, and stroke, which can be fatal.
• Gastrointestinal Risk: Increased risk of serious gastrointestinal adverse events, including bleeding, ulceration, and perforation of the stomach or intestines.

Like antidepressants, NSAIDs (including ibuprofen, naproxen, celecoxib, and meloxicam) are marketed at massive generic volume by hundreds of labelers. The class warning applies to all prescription NSAIDs, ensuring that any new generic entry or repackaged product must carry the warning.

For safety leads, the takeaway is that when the FDA determines a risk is class-wide, it applies a blanket requirement across all member drugs regardless of individual molecular differences. This creates a regulatory "floor" that later entrants cannot negotiate away, even with superior clinical data, unless they can prove their drug belongs to a distinct clinical category.

---

How often are boxed warnings added versus removed?

A boxed warning is not necessarily permanent, but it is highly stable. The addition or revision of a warning is a dynamic process driven by postmarket safety surveillance and coordinated regulatory reviews.

Two landmark cohort studies bracket this question, and they are frequently misquoted — including by pages currently ranking for this topic. Read carefully, because they count different things.

Lasser et al. (JAMA 2002) examined all 548 new chemical entities approved by the FDA between 1975 and 1999. By 2000, 45 of them (8.2%) had acquired one or more new boxed warnings that were not present at approval, and 16 (2.9%) had been withdrawn — 56 drugs (10.2%) in total experiencing either outcome. In Kaplan-Meier analysis, the estimated probability of acquiring a new boxed warning or being withdrawn over 25 years was 20%, half of the labeling changes occurred within 7 years of introduction, and half of the withdrawals occurred within 2 years.

Note the distinction the 8.2% figure requires: it is the observed proportion by the year 2000, not a 25-year lifetime risk. The 25-year projection is the 20% Kaplan-Meier estimate, and it covers warnings and withdrawals combined. Quoting 8.2% as a lifetime rate understates the risk by more than half.

Downing et al. (JAMA 2017) evaluated 222 novel therapeutics approved between 2001 and 2010, followed for a median of 11.7 years (IQR 8.7–13.8):

The events column and the drugs column are not interchangeable, and conflating them is the single most common error in secondary coverage of this study. There were 61 boxed-warning events, but they landed on 43 drugs — some products picked up more than one. So the correct statement is that 19.4% of the cohort received a postmarket boxed warning, not 27.5%. Across all event types, 32.0% of novel therapeutics experienced at least one postmarket safety event, and the cumulative incidence at 10 years was 30.8%.

The median time from approval to a first postmarket safety event was 4.2 years (IQR 2.5–6.0) — consistent with Lasser's finding that half of boxed-warning additions arrive within 7 years. For launch planning, that means the boxed-warning risk window sits squarely inside the period when a brand is still building its formulary position, not at the end of the lifecycle.

The three cohort withdrawals were valdecoxib (Bextra, 2005), tegaserod (Zelnorm, 2007), and efalizumab (Raptiva, 2009) — valdecoxib for cardiovascular risk and severe skin reactions, tegaserod for cardiovascular ischemic events, and efalizumab for progressive multifocal leukoencephalopathy. Alosetron (Lotronex), often listed alongside these, is not in this cohort: it was approved in 2000 and withdrawn the same year, outside the 2001–2010 approval window.

---

What changed in the 2026 menopausal hormone-therapy boxed warnings?

While the addition of a boxed warning is common, the removal or restriction of a boxed warning is exceedingly rare, requiring overwhelming clinical evidence. A major regulatory hook in 2026 is the FDA's coordinated revision of the menopausal hormone therapy (MHT) boxed warnings.

The sequence matters, because it is the clearest recent template for how a class-wide boxed warning actually comes off:

The FDA's action removed three specific risk statements from the boxed warning: cardiovascular disease, breast cancer, and probable dementia. The agency also asked manufacturers to drop the older instruction to use the lowest effective dose for the shortest possible time.

Two details are widely misreported and are worth stating precisely:

• The removals were not limited to low-dose local formulations. The first six products span all four HRT categories — topical vaginal estrogen, systemic estrogen-alone, systemic combination estrogen-plus-progestogen, and systemic progestogen-alone — namely Prometrium, Divigel, Cenestin, Enjuvia, Estring, and Bijuva. This was not a carve-out for vaginal estrogen; it was a reconsideration of the underlying WHI interpretation across the class.
• The probable-dementia statement was removed, not narrowed. The FDA's announcement lists it among the three statements deleted from the boxed warning, not rewritten with an age qualifier.

A third detail is the one launch teams should actually notice: transdermal estradiol was not in the first batch. Most international clinical guidelines favour non-oral estradiol, so its absence from the initial six is a sequencing artifact of which manufacturers filed first — not a signal about the relative safety of the route. Removing a class-wide boxed warning proceeds product-by-product, on each sponsor's own supplement timeline, even when the agency has already announced the class-level decision.

For brand teams, the precedent cuts a specific way. The lever here was not a formulation-specific argument about local versus systemic exposure; it was a regulator-initiated re-reading of the pivotal evidence base, followed by 29 sponsors filing supplements into an open door. That is a much rarer opening than a routine labeling supplement, and it is worth being honest about how rarely it appears: adding a boxed warning is a routine postmarket event, while removing one at class scale has happened a handful of times in the FDA's history.

---

Are fast-tracked drugs more likely to receive a boxed warning?

A key strategic question for pipeline planners is whether the FDA's expedited pathways (Accelerated Approval, Priority Review, Fast Track, or Breakthrough Therapy designation) increase the risk of a postmarket boxed warning.

Downing et al. (JAMA 2017) addressed this directly, and it is worth quoting the actual effect sizes rather than the widely circulated "3.5 times more likely" claim — a figure that appears on several currently-ranking pages, is rendered inconsistently as both "3.5 times" and "3.5 percent" on the same page, and does not correspond to any result the study reports. Here is what the study actually found, expressed as incidence rate ratios for postmarket safety events:

Three things follow that the SERP consensus gets wrong.

The expedited-pathway effect is real but smaller than advertised. Accelerated approval carried an IRR of 2.20 — roughly double the postmarket safety-event rate, not 3.5 times. The confidence interval (1.15–4.21) is wide, so the point estimate should be treated as directional.

The strongest single predictor was therapeutic area, not pathway. Psychiatric therapeutics carried an IRR of 3.78 — the highest in the study. That is the number the "3.5x" claim appears to be a garbled version of, and it belongs to a completely different variable. It also converges with the registry data above: antidepressants and antipsychotics together account for 2,172 boxed-warning label records, the two largest buckets in the entire named subset. The class concentration and the cohort risk model are telling the same story from opposite directions.

Speed of review was protective, not harmful. Drugs reviewed in under 200 days had an IRR of 0.46. What raised risk was approval near a deadline (IRR 1.90) — a proxy for rushed end-of-cycle decisions, not for short review clocks generally. "Fast review equals unsafe drug" is not what the data show.

This differential is not a failure of the FDA's review process. Expedited pathways are reserved for serious or life-threatening conditions with unmet medical need, where the agency deliberately accepts higher clinical uncertainty at approval and relies on postmarket confirmatory trials and active surveillance.

For pipeline strategists, the decision rule that falls out of this is:

The window to plan for is the first 4 to 7 years post-approval, where both studies place the bulk of the events. A postmarket boxed warning can severely restrict commercial potential, as seen in the JAK inhibitor FDA boxed warnings case, where postmarket cardiovascular and malignancy signals led to restrictions that eroded first-line positioning.

---

The Label Negotiation Process: Prior Approval vs. CBE Supplements

When a safety signal is identified, the process of updating the prescribing information to add or modify a boxed warning involves intense negotiation between the sponsor and the FDA. This labeling workflow is managed through the FDA's supplemental NDA (sNDA) or supplemental BLA (sBLA) pathways:

1. Prior Approval Supplement (PAS): This pathway is used for major changes that require complete FDA review and authorization before the updated label can be distributed. Adding a new boxed warning almost always requires a PAS. The sponsor submits clinical trial analyses, literature reviews, and proposed text, and the FDA has up to six months (or more, if major revisions are negotiated) to approve the final wording.
2. Changes Being Effected (CBE-30): The sponsor notifies the FDA of a labeling change and waits 30 days before distributing the new label. The FDA can object or modify the label during this period. CBE-30 is typically used for moderate safety updates that do not involve adding a boxed warning.
3. Changes Being Effected Immediate (CBE-0): The sponsor distributes the new label immediately upon filing the supplement with the FDA. This is reserved for urgent safety updates to protect public health, but it is rarely used to add a new boxed warning unless directed by the FDA via a formal Safety Labeling Change (SLC) notification.

When the FDA initiates a boxed warning addition via an SLC, the sponsor has 30 days to respond. The sponsor can either accept the proposed text or submit a counter-proposal with alternative wording or scientific arguments against the warning. This negotiation is critical: the specific phrasing within the box can determine whether the drug remains commercially viable or is restricted to salvage therapy.

---

International Context: FDA Boxed Warnings vs. EMA and PMDA Mechanisms

Global launch teams must recognize that a safety warning in the United States does not automatically translate into identical regulatory actions abroad. Each region’s regulatory body uses distinct instruments to communicate safety risks:

European Medicines Agency (EMA)

In the European Union, the EMA does not use a direct visual equivalent of the FDA's boxed warning. Instead, it relies on the following mechanisms:

• Direct Healthcare Professional Communications (DHPC): Formal letters sent directly to doctors to communicate urgent safety updates, drug recalls, or restricted indications.
• The Black Triangle (▼): An inverted black triangle printed on the package leaflet and Summary of Product Characteristics (SmPC) for newly authorized medicines or those subject to additional postmarket monitoring (e.g., all biologics and products approved under conditional marketing authorization).
• Annex II Conditions: Specific restrictions on prescribing or mandatory registries built into the marketing authorization conditions.

Pharmaceuticals and Medical Devices Agency (PMDA - Japan)

In Japan, the PMDA utilizes the "Blue Letter" (Rapid Safety Communication) and "Yellow Letter" (Emergency Safety Communication) systems. These letters are sent directly to healthcare institutions to mandate immediate changes in clinical use, often following severe postmarket adverse events.

This regulatory divergence means that a drug carrying an FDA boxed warning for a specific risk might only carry standard warnings in its European SmPC, or conversely, face strict prescribing restrictions in Europe that the FDA does not replicate. Launch teams must plan separate safety negotiation strategies for each major regulatory jurisdiction.

---

Safety Surveillance: Adverse Event Reporting vs. Labeling Determinations

It is vital for pharmacovigilance professionals to separate the FDA's labeling determinations (boxed warnings) from raw post-market surveillance reports, such as those found in the FDA Adverse Event Reporting System (FAERS).

The three largest boxed-warning classes in the table above each have a companion post on this site quantifying their reported adverse events rather than their labeling: opioid analgesic FAERS adverse events by the numbers, SSRI and SNRI antidepressant FAERS adverse events by the numbers, and NSAID adverse events in FAERS by the numbers. Reading a class's FAERS report volume next to its boxed-warning label count is instructive precisely because the two rankings do not match — and neither one is an incidence rate.

FAERS is a passive surveillance database containing millions of voluntary safety reports. Because of its passive nature:

• Causality is not proven: A report in FAERS does not prove that the drug caused the adverse event.
• Reporting bias dominates: Highly publicized drugs or those involved in litigation generate disproportionately more reports.
• Denominators are missing: Raw report counts do not provide the total number of patients exposed to the drug, making it impossible to calculate true clinical incidence.

In contrast, a boxed warning represents the FDA's formal, evidence-backed regulatory judgment. Before mandating a boxed warning, the FDA reviews clinical trials, epidemiology studies, meta-analyses, and validated safety signals. A boxed warning is a legal mandate that requires manufacturers to update their prescribing information, whereas FAERS reports are raw signal inputs that may or may not lead to labeling changes.

When designing safety messaging, access teams must ensure that clinicians understand this distinction: while FAERS reports represent passive, unverified signals, a boxed warning represents a clinical risk boundary that must actively guide prescribing decisions.

---

FAQ

Is a boxed warning the same as a black box warning?

Yes. "Black box warning" is the historical and common industry term for a boxed warning, named after the prominent black border that surrounds the safety statement at the top of the prescribing information. The FDA officially refers to this format as a "boxed warning."

Does a boxed warning mean a drug is unsafe or should not be used?

No. A boxed warning does not mean a drug is unsafe or should be avoided. It is a tool to help clinicians weigh the benefits and risks for individual patients. Many life-saving medications (such as chemotherapy agents, anticoagulants, and transplant medicines) carry boxed warnings because their high therapeutic benefit outweighs their substantial safety risks when managed correctly.

Why is the boxed-warning ingredient count higher than the '400 drugs' figure I have seen?

The "400+ drugs" figure typically refers to unique, innovator prescription active moieties. The 1,084 count represents distinct normalized ingredient names in the openFDA label registry, which includes generic formulations and multi-ingredient combination names counted as single entries. It is not driven by over-the-counter products — only 19 of 32,974 boxed-warning records are OTC. And it is a floor rather than a total, because 61.5% of boxed-warning label records carry no harmonized ingredient name at all.

How many drugs receive a boxed warning after approval?

In the Downing et al. (JAMA 2017) cohort of 222 novel therapeutics approved 2001–2010, 43 drugs (19.4%) received a postmarket boxed warning over a median 11.7 years of follow-up, from 61 total boxed-warning events. Counting 32.0% is also correct but answers a broader question — that is the share of drugs experiencing any postmarket safety event, including FDA safety communications and withdrawals.

Can a boxed warning be removed once it is added?

Yes, but it is uncommon. Removal normally requires a labeling supplement containing clinical or real-world evidence that the risk is lower than originally believed, or that risk-management measures have mitigated it. The February 2026 menopausal hormone-therapy revisions are the largest recent example, and they were unusual in being FDA-initiated: the agency announced the removal in November 2025 after its own literature review, then approved sponsor supplements in batches.

Does a boxed warning affect payer coverage?

Not automatically, and the published evidence suggests less than practitioners assume. It is a labeling instrument, not a coverage determination — but it commonly becomes an input to utilization-management criteria, step-therapy sequencing, and specialty-pharmacy monitoring requirements, especially where a boxed-warning-free alternative exists in the same class.

---

Sources

• U.S. Food and Drug Administration (FDA). "openFDA Drug Labeling Endpoint." Registry snapshot analysed 25 July 2026 (260,986 label records). open.fda.gov
• U.S. Food and Drug Administration (FDA). "Prescribing Information for Medicines." Accessed July 2026. fda.gov
• U.S. Food and Drug Administration (FDA). "FDA Approves Labeling Changes to Menopausal Hormone Therapy Products." Press announcement, 12 February 2026. fda.gov
• U.S. Department of Health and Human Services / FDA. "HHS Advances Women's Health, Removes Misleading FDA Warnings on Hormone Replacement Therapy." Press announcement, 10 November 2025. fda.gov
• Downing, N. S., Shah, N. D., Aminawung, J. A., et al. "Postmarket Safety Events Among Novel Therapeutics Approved by the US Food and Drug Administration, 2001-2010." JAMA, 2017; 317(18): 1854-1863. jamanetwork.com
• Lasser, K. E., Allen, P. D., Woolhandler, S. J., Himmelstein, D. U., Wolfe, S. M., Bor, D. H. "Timing of New Black Box Warnings and Withdrawals for Prescription Medications." JAMA, 2002; 287(17): 2215-2220. pubmed.ncbi.nlm.nih.gov
• Makary, M. A., Nguyen, C. P., Høeg, T. B., Tidmarsh, G. F. "Updated Labeling for Menopausal Hormone Therapy." JAMA, 2026; 335(2): 117-118. pubmed.ncbi.nlm.nih.gov
• National Center for Biotechnology Information (NCBI). "Box Warning (StatPearls)." Accessed July 2026. ncbi.nlm.nih.gov
• DailyMed. "Structured Product Labeling (SPL) Resources." Accessed July 2026. dailymed.nlm.nih.gov
• European Medicines Agency (EMA). "Direct Healthcare Professional Communications (DHPC)." Referenced for international comparison. ema.europa.eu