This page documents the clinical trial evidence submitted to the US Food and Drug Administration for the licensure of the two hepatitis B vaccines currently on the CDC childhood immunization schedule: Recombivax HB (Merck, licensed 1986) and Engerix-B (GSK, licensed 1989).
The data below is drawn from the FDA-approved package inserts for each product and from Freedom of Information Act (FOIA) records obtained by the Informed Consent Action Network (ICAN). The primary factual claims — trial size, control arm, and post-dose monitoring window — are directly quoted from the package inserts.
| Field | Value |
|---|---|
| Licensed | 1986 |
| Technology | First recombinant DNA vaccine licensed in the United States |
| Trial population | 147 healthy infants and children (up to age 10) |
| Control group | None |
| Solicited safety monitoring window | 5 days after each dose |
| Unsolicited safety monitoring window | 5 days after each dose |
Primary source. FDA-approved package insert, Section 6.1 (Clinical Trials Experience): "In three clinical studies, 434 doses of RECOMBIVAX HB, 5 mcg, were administered to 147 healthy infants and children (up to 10 years of age) who were monitored for 5 days after each dose." See the current Recombivax HB label on DailyMed.
FOIA record. ICAN submitted a FOIA request to the FDA for clinical trial reports supporting the Recombivax HB licensure. The FDA's production contained no documents reflecting safety review beyond the 5-day post-dose window reported in the package insert. See ICAN for the legal filings.
| Field | Value |
|---|---|
| Licensed | 1989 |
| Trial population (all ages, cumulative) | 5,071 subjects across 36 studies (13,495 doses) |
| Pediatric cohorts (from label) | Neonates: n=381 (0/1/2 month schedule), n=52 (0/1/6 month schedule); children 6 mo – 10 yr: N=242 |
| Control group | None |
| Solicited safety monitoring window | 4 days after each dose |
| Unsolicited safety monitoring window | 4 days after each dose |
Primary source. FDA-approved package insert, Section 6.1 (Clinical Trials Experience): "In 36 clinical studies, a total of 13,495 doses of ENGERIX-B were administered to 5,071 healthy adults and children who were initially seronegative for hepatitis B virus… All subjects were monitored for 4 days post-administration." Pediatric cohort details appear in Sections 14.3–14.4. See the current Engerix-B label on DailyMed.
Deposition record. During his January 11, 2018 deposition by attorney Aaron Siri (Plotkin Deposition 2018), Stanley Plotkin initially stated that safety data had "surely" been collected beyond the 4-day window. Presented with the package insert, Plotkin acknowledged his prior statement was "speculation based on experience." The full deposition transcript is publicly hosted at sirillp.com/plotkin-depo-full. The transcript contains no citation to safety data collected beyond 4 days.
FOIA outcome. Following ICAN litigation against the FDA, the agency's production did not include documents reflecting safety monitoring of the pre-licensure trial beyond the short window reported in the package insert.
| Brand | Licensed | Trial Size | Control | Solicited | Unsolicited |
|---|---|---|---|---|---|
| Recombivax HB (Merck) | 1986 | 147 infants/children (434 doses) | None | 5 days | 5 days |
| Engerix-B (GSK) | 1989 | 5,071 subjects across all ages (pediatric: ~675) | None | 4 days | 4 days |
The following ingredient data is drawn directly from the FDA-approved package inserts for each product (linked in the Sources section). Both vaccines are aluminum-adjuvanted; neither pediatric formulation contains thimerosal. The active ingredient in both is recombinant hepatitis B surface antigen (HBsAg) produced in genetically modified Saccharomyces cerevisiae (baker's yeast).
| Ingredient | Amount per 0.5 mL dose | Notes |
|---|---|---|
| Hepatitis B surface antigen (HBsAg) | 5 mcg | Active ingredient; recombinant, yeast-derived |
| Aluminum (as amorphous aluminum hydroxyphosphate sulfate) | ~0.25 mg | Adjuvant |
| Residual formaldehyde | < 7.5 mcg | Manufacturing residue (label: "< 15 mcg/mL") |
| Residual yeast protein | < 1% of dose | Manufacturing residue |
| Thimerosal (mercury) | None | Preservative-free |
Label quote: "approximately 0.5 mg of aluminum (provided as amorphous aluminum hydroxyphosphate sulfate, previously referred to as aluminum hydroxide) per mL of vaccine" — per 0.5 mL dose, ≈ 0.25 mg.
| Ingredient | Amount per 0.5 mL dose | Notes |
|---|---|---|
| Hepatitis B surface antigen (HBsAg) | 10 mcg | Active ingredient; recombinant, yeast-derived |
| Aluminum (as aluminum hydroxide) | 0.25 mg | Adjuvant |
| Sodium chloride | 4 mg | Tonicity / buffer |
| Sodium phosphate (dibasic and monobasic, dihydrate) | 0.45 mg + 0.35 mg | Buffer |
| Residual yeast protein | ≤ 5% of dose | Manufacturing residue |
| Thimerosal (mercury) | None | Preservative-free |
Label quote: "Each 0.5-mL pediatric/adolescent dose contains 10 mcg of HBsAg adsorbed on 0.25 mg aluminum as aluminum hydroxide."
The CDC childhood schedule administers three doses of hepatitis B vaccine (birth, 2 months, 6 months). At 0.25 mg aluminum per dose, the cumulative aluminum burden from the hepatitis B series alone is:
| Dose | Timing | Aluminum (cumulative) |
|---|---|---|
| 1 | Birth (day 0) | 0.25 mg |
| 2 | 2 months | 0.50 mg |
| 3 | 6 months | 0.75 mg |
This figure reflects the hepatitis B series in isolation. Infants on the full CDC schedule also receive aluminum-adjuvanted DTaP, Hib (some formulations), PCV13/PCV15, and Hepatitis A, producing a larger cumulative aluminum exposure across the first year of life. For the broader treatment of injected aluminum, absorption kinetics, and regulatory reference levels, see Aluminum Adjuvants.
A clinical trial with n=147 cannot, under standard assumptions, statistically distinguish a background rate of zero from an adverse-event rate of 1 in 1,000.
This is a general property of trial size, not a claim about Recombivax HB specifically. It means: whatever the trial did find, a harm at that frequency could have been missed.
Under FDA practice, a licensed vaccine may serve as an active comparator in subsequent vaccine trials. Once Recombivax HB and Engerix-B were licensed, they were used as comparators for newer products:
An active comparator is only epistemically useful if the comparator's own safety profile has been established against a true baseline — i.e., against subjects who received no vaccine or an inert placebo. The purpose of a control arm is to measure the background rate of any given adverse event in an otherwise-similar population, so that the trial can attribute the excess rate in the treatment arm to the product.
Recombivax HB and Engerix-B were licensed without an inert control arm. Their adverse-event rates against baseline are therefore not established. When these products were then used as comparators in later vaccine trials, the downstream trial could only answer a narrower question:
> "Does the new vaccine cause the event at a different rate than Recombivax / Engerix-B?"
It could not answer the broader question:
> "Does the new vaccine cause the event at a higher rate than baseline?"
Any adverse effect that Recombivax or Engerix-B does cause at some background rate becomes invisible in the downstream comparison. If both arms produce the same rate of a given event, the trial concludes "no difference between arms" — which is true, but is not the same as "no safety signal." The shared event rate could be entirely vaccine-caused; the design cannot distinguish the two.
This dynamic extends down the comparator chain. If Vaccine A is licensed without a placebo arm, and Vaccine B is licensed using Vaccine A as its comparator, then Vaccine C (using B as comparator) inherits the unexamined baseline from A through B. Each successive licensure relies on the assumption that the earlier product's safety profile is adequately known — but for the hepatitis B parent products, that assumption was never tested against an inert control.
The standard defense of this practice is ethical: it is argued that exposing trial subjects to a disease-relevant placebo would be unethical once an effective product exists. This defense is addressed in the steelman discussion below (see defense #3, "The 'no placebo' critique misunderstands the ethics of vaccine trials"). The short version: the ethics argument constrains the choice between placebo and active comparator in subsequent trials — but it does not retroactively supply the baseline data that the original licensure of Recombivax HB and Engerix-B never collected.
| Dose | Timing |
|---|---|
| Dose 1 | Birth |
| Dose 2 | 2 months |
| Dose 3 | 6 months |
Hepatitis B virus (HBV) is transmitted primarily through blood, sexual contact, and shared injection equipment. In the United States, perinatal (mother-to-infant) transmission at birth is the primary HBV exposure route for newborns.
Standard prenatal care in the US includes HBsAg (hepatitis B surface antigen) screening of pregnant women. Infants born to HBsAg-positive mothers are managed under a specific protocol: hepatitis B vaccine plus hepatitis B immune globulin (HBIG) within 12 hours of birth.
For infants born to HBsAg-negative mothers (the majority of US births), the CDC rationale for universal birth-dose vaccination cites two categories of residual risk: (1) maternal false-negative screening results and seroconversion during pregnancy, and (2) prevention of later childhood or adolescent acquisition.
FOIA records obtained by ICAN from the CDC reflect no documented cases of hepatitis B transmission in a US school setting.
Several arguments are routinely raised in defense of the Recombivax and Engerix pre-licensure designs. The strongest forms are presented below, with an honest assessment of where each holds and where it weakens.
The argument. Acute vaccine reactions (fever, injection-site swelling, systemic symptoms) peak within 48–72 hours post-dose. A 4–5 day solicited-event window is consistent with standard acute reactogenicity protocols. Delayed or chronic effects are designed to be studied through post-licensure epidemiology (VAERS, VSD (Vaccine Safety Datalink)), not pre-licensure RCTs.
Where it holds. The 48–72 hour peak for acute local and systemic reactions is well-supported in the reactogenicity literature.
Where it weakens. The argument requires post-licensure systems to actually detect delayed effects. VAERS is a passive voluntary reporting system; VAERS documents the Harvard Pilgrim / AHRQ finding that "fewer than 1% of vaccine adverse events are reported." VSD access has been limited for independent researchers (VSD (Vaccine Safety Datalink)). If the downstream system is incomplete, the trade-off — "short trial plus strong surveillance" — loses one of its terms.
The argument. Phase 3 vaccine trials are powered for their prespecified primary endpoints (typically efficacy and common reactogenicity), not for the detection of every possible rare harm. A 147-subject trial may be appropriate if the endpoint definitions permit it.
Where it holds. This is how trial design works in principle. Prespecified endpoints drive the power calculation.
Where it weakens. The logic becomes circular when the endpoint definitions themselves are narrow: small trial → no rare events observed → "no safety signal" → small trial accepted as adequate for the next product. By the 2000s, most vaccine pre-licensure trials had grown to several thousand subjects (e.g., HPV, rotavirus, pneumococcal conjugate). The 147-subject Recombivax trial is small even by pre-licensure vaccine standards of its era, not just by modern standards.
The argument. When a vaccine targets a known-serious disease, it can be considered unethical to randomize subjects to an inert control, because the control arm is exposed to known risk. Active comparators (another licensed vaccine) are the ethical and scientific standard.
Where it holds. The ethics framework is real. For later vaccines targeting actively circulating diseases with known morbidity, the argument is strong.
Where it weakens. The Recombivax trial used no control arm at all — not an inert placebo, and not an active comparator. The ethics argument applies to the choice between inert placebo and active comparator; it does not justify the absence of any control group. Additionally, in 1986 the US pediatric incidence of HBV was low, weakening the "known-risk exposure" component of the ethics argument as applied to this specific trial.
The argument. Both products have been administered worldwide for 30+ years. Catastrophic safety signals of the kind a larger pre-licensure trial might have caught would have surfaced in population-level data. Specific concerns (multiple sclerosis, Guillain-Barré syndrome) have been the subject of published epidemiological studies.
Where it holds. Long-term wide-population use does provide a form of observational safety evidence. For rare events at the 1-in-100,000+ level that are catastrophic and identifiable, a signal would likely emerge.
Where it weakens. "Signal emergence" depends on detection capacity. Subclinical effects, conditions with multifactorial etiologies, and conditions with delayed onset are genuinely difficult to detect at the population level without active surveillance and a control denominator — neither of which has been reliably in place for hepatitis B vaccines in the US. The published MS and GBS epidemiology is not unanimous; it is a mixed literature.
There are also two deeper problems with the "population-level data is the real safety database" claim:
The "what's left if VAERS is dismissed" problem. Defenders of short pre-licensure trials frequently point to VAERS and VSD (Vaccine Safety Datalink) as the systems that pick up what trials miss. But when VAERS reports cite safety signals, the same medical community typically dismisses VAERS as unreliable because it is passive, voluntary, and subject to reporter bias. That critique is valid. The problem is what it leaves: VSD access has been limited for independent researchers (VSD (Vaccine Safety Datalink)); ecological and observational studies are vulnerable to healthy-user bias and confounding; and international passive systems (Yellow Card in the UK, EudraVigilance in the EU) have the same limitations as VAERS. If VAERS is dismissed and VSD is gated and observational studies are methodologically weaker, the "post-licensure surveillance will catch it" defense loses the system it depends on.
The background-rate problem. If a vaccine program causes a small increase in a chronic condition that has many other causes, population-level data cannot isolate the vaccine's contribution without an unvaccinated control denominator. US childhood chronic-disease prevalence has risen substantially over the period that the recommended schedule has expanded: a 2025 peer-reviewed analysis of PEDSnet and the National Survey of Children's Health found that the share of 3-to-17-year-olds with at least one chronic condition rose from 25.8% to 31.0% between 2011 and 2023 in the general-survey population, and from 39.9% to 45.7% in pediatric health-system data over the same period ([Academic Pediatrics, 2025](https://www.academicpedsjnl.net/article/S1876-2859(25)00035-X/fulltext); covered by UCLA Health). Longer-horizon NHIS data show activity-limiting chronic conditions in children rising from roughly 2% in the early 1960s to 8% by 2010, though that comparison is complicated by changes in survey methodology and diagnostic definitions. This rise has many candidate causes — diet, obesity, environmental exposures, screen time, diagnostic expansion, and others — and is not evidence that vaccines caused it. But it does mean the "nothing bad has shown up at the population level" argument is harder to make cleanly: something has shown up, the attribution is contested, and the surveillance systems required to untangle which inputs contributed cannot do so without the unvaccinated comparison that US surveillance does not collect.
Hepatitis B Vaccines (Post-Licensure), Pre-Licensure Safety Testing, Childhood Vaccine Schedule, Plotkin Deposition 2018, ICAN, Stanley Plotkin, 1986 Act (National Childhood Vaccine Injury Act), Financial Immunity for Vaccine Makers, Merck, GSK
1. Recombivax HB (Merck) — FDA-approved package insert, Section 6.1 (Clinical Trials Experience). Current label: Recombivax HB on DailyMed.
2. Engerix-B (GSK) — FDA-approved package insert, Sections 6.1, 14.3, 14.4. Current label: Engerix-B on DailyMed.
3. Stanley Plotkin deposition, January 11, 2018 (Aaron Siri, examining counsel; Oakland County Circuit Court, Family Division, Michigan). Full transcript (sirillp.com). See Plotkin Deposition 2018.
4. ICAN v. FDA, FOIA production records. Case filings and production documents are published by the Informed Consent Action Network; see ICAN for the legal docket.
5. Hanley JA, Lippman-Hand A. "If nothing goes wrong, is everything all right? Interpreting zero numerators." JAMA. 1983;249(13):1743–1745. PMID: 6827763. PubMed record.
6. Prevalence and Trends in Pediatric-Onset Chronic Conditions in the United States, 2011–2023. Academic Pediatrics, 2025. [Full text](https://www.academicpedsjnl.net/article/S1876-2859(25)00035-X/fulltext). Summary via UCLA Health news release.
Frequently Asked Questions