Epistemic Asymmetry Drives CIDRAP Study That Claims (After Cherry-Picking) “Strong Evidence” of Efficacy of Birth Dose of Hepatitis B Vaccine

All by itself, this paper actually destroys the credibility of the argument that ACIP got it wrong. This should be a textbook case used as an example of confirmation bias.

The hepatitis B vaccine birth dose has been part of U.S. immunization policy since 1991. The evidence of effectiveness of later doses has not been questioned. It is claimed to have reduced pediatric hepatitis B virus (HBV) infections and to have contributed to long-term declines in chronic disease burden.

However, despite widespread vaccination, evidence indicates it has not substantially impacted long-term chronic HBV prevalence in the United States. Data from multiple National Health and Nutrition Examination Survey (NHANES) studies spanning from 1988 to 2016 consistently show that chronic HBV prevalence has remained stable at approximately 0.3% to 0.35%. This stable prevalence primarily reflects the high rates among foreign-born populations, particularly non-Hispanic Asians, who continue to have prevalence rates substantially higher (3.85%) than U.S.-born individuals (0.15%).

Thus, while the birth dose policy has successfully lowered acute pediatric infections, one of its primary rationales—reducing chronic HBV burden and subsequent hepatocellular carcinoma in the general population—has not been realized as anticipated.

What is in dispute is something narrower and more contemporary: whether, in 2026, for infants born to mothers who test negative for HBsAg during pregnancy, the universal birth dose must remain a standard recommendation, or whether shared clinical decision-making with a delayed first dose (≥2 months) is scientifically defensible in a low-prevalence environment.

Into that debate stepped a high-profile “Special Article” in Pediatrics, produced by CIDRAP’s newly launched “Vaccine Integrity Project” and co-authored by former CDC Director Rochelle Walensky. The paper declares that it conducted an “independent review” and found “strong evidence” for the safety and effectiveness of the birth dose, and “no evidence to support a change in vaccine recommendations.”

The conclusion is categorical. The methods are not.

This article is not fraudulent. It does not fabricate data. The hepatitis B vaccine is safe and effective. The birth dose works. The problem is not the vaccine. The problem is epistemic asymmetry — unequal standards of evidence applied to competing policy positions.

The “Comprehensive but Not Systematic” Escape Hatch

In its Methods section, the paper states that it is not a complete systematic review, but rather a “comprehensive review, analysis, and synthesis.” Later, it explicitly acknowledges that it did not include many elements of a systematic review: no systematic search methodology, no meta-analysis, and no formal quality-of-evidence assessment.

Yet the Abstract asserts that the authors found “strong evidence” and “no evidence to support a change”.

Those contradictions matter: They are the old behavior of the ousted old guard seen many times before where conclusions drawn in the abstract or title are not supported by or not supportable by the study results.

A non-systematic review cannot legitimately support a definitive “no evidence” conclusion about a policy change. “No evidence found” only carries weight if the search process is demonstrably exhaustive, reproducible, and structured to detect contrary findings.

This paper does not provide:

• A PRISMA flow diagram.

• A pre-registered protocol (e.g., PROSPERO).

• Dual independent screening procedures.

• Risk-of-bias assessment (RoB 2, ROBINS-I, etc.).

• GRADE tables grading the overall certainty of evidence.

• Transparent documentation of inclusion and exclusion decisions.

The authors describe the GRADE and Evidence-to-Recommendations (EtR) frameworks used by ACIP in detail. They do not apply those frameworks themselves.

These would help establish the claimed and implied rigor. But, as you will see, the study has no rigor. They invoke the architecture of systematic review without assuming its obligations, write permission slips for inclusion bias, and proceed from assumptions to conclusions. As if no one would notice. This is only is the first asymmetry.

Scope Drift: “US Populations Only” — Except When Not

The Methods section states that safety analyses were limited to US-licensed products used in US populations, with rare exceptions if highlighted in ACIP reviews.

But the tables include multiple trials and cohorts from Egypt, Israel, Colombia, Vietnam, India, China, Taiwan, and the Netherlands. Large non-US datasets are incorporated when they strengthen immunogenicity or long-term protection claims.

Meanwhile, European policy differences are acknowledged and then dismissed on the grounds that the United States is “unique.” Special pleading does not help their case.

Non-US data are admissible when supportive. Non-US systems are incomparable when inconvenient.

That is selective inclusion framed as comprehensiveness. It’s also known as cherry-picking.

The 99% Decline: Association Framed as Causation

The paper states that infant vaccination has resulted in a 99% reduction in pediatric HBV infections.

The decline is real. Surveillance data show dramatic reductions in acute HBV among children since the early 1990s.

But the paper does not:

• Decompose the independent contribution of the birth dose versus completion of the 3-dose series.

• Quantify the role of adult vaccination programs.

• Adjust for improved prenatal screening.

• Model immigration-driven prevalence changes.

• Account for secular declines in community transmission.

The 99% figure is deployed rhetorically as causal proof of the birth dose’s indispensability. It is, at best, a strong association embedded within a multifactorial intervention era.

Correlation does not equal causation. Policy recalibration requires careful and exact causal parsing. The paper offers trend narration.

The Denominator They Never Calculate

The December 2025 ACIP vote applied only to infants born to mothers who tested HBsAg-negative during pregnancy.

The relevant question is therefore precise:

What is the probability that an infant born to a confirmed HBsAg-negative mother acquires HBV before 2 months of age?

The paper never performs this calculation.

Consider the structure of the missing analysis:

• Approximately 3.6 million births annually in the United States.

• Roughly 0.5% of pregnant women are HBsAg-positive, or about 18,000 per year.

• Commercial assays report sensitivities of 97–100%.

• Maternal seroconversion between screening and delivery is rare but nonzero.

• Reported perinatal cases have been fewer than 20 annually in recent years, likely undercounted.

Even allowing for underreporting, the expected number of infections among infants of mothers who screened negative is extraordinarily small.

In fact, the authors themselves concede that “the risk of HBV transmission to infants born to HBsAg-negative women is very low.”

That sentence alone acknowledges the core premise of the ACIP recalibration.

If risk in this subgroup is “very low,” then recalibrating from a universal standard recommendation to shared clinical decision-making is scientifically arguable. The paper never quantifies the residual risk in this subgroup, yet concludes that no policy change is supported.

This is not a trivial omission. It is the central analytic gap.

Straw-Man Framing of the ACIP Decision

The ACIP vote did not alter recommendations for infants born to HBsAg-positive or unknown-status mothers. Those infants continue to receive birth dose plus HBIG.

Yet the paper repeatedly invokes:

• Screening failures in HBsAg-positive mothers.

• Postnatal household transmission risks from known carriers.

• Errors in documentation for positive-status mothers.

These are real concerns. But they are not the policy target of Vote 1.

Defending universal birth dose by invoking risks in populations unaffected by the change is a scope mismatch. It conflates universal protection arguments with subgroup-specific risk analysis.

The Silence on Improving Screening is Deafening

Another tell is the study is utterly silent on improving screening.

The paper identifies screening failures extensively — but exclusively as ammunition for the birth dose argument, never as a problem warranting its own solution:

Screening failure 1 — Coverage gaps:

“From 2009–2017, approximately 50% of infants born to women with HBV infection were not identified through routine prenatal screening” (Reference 87)

Screening failure 2 — Technical limitations:

“Commercially available clinical enzyme immunoassays range in sensitivity from 97% to 100% but may have reduced sensitivity for certain HBV genotypes or during early acute infection”

Screening failure 3 — Post-screening seroconversion:

“HBV infection in pregnancy may occur after prenatal screening, a negative prenatal test result does not guarantee maternal infection status at the time of delivery”

Screening failure 4 — Communication breakdowns:

“Communication of maternal hepatitis B test results among laboratories, prenatal clinics, and obstetric and pediatric hospital care teams is not always timely and accurate”

Screening failure 5 — Missing hospital policies:

“A lack of written hospital policies and standing orders for the birth dose can result in missed opportunities”

What They Never Say

At no point in the paper do the authors:

• Recommend improving prenatal screening coverage toward universal completion

• Recommend improving screening timing — e.g., repeat testing in the third trimester for high-risk women

• Recommend improving sensitivity for variant HBV genotypes

• Recommend improving laboratory-to-clinical communication systems

• Recommend standardized electronic health record integration of maternal HBsAg results

• Recommend hospital policy reform to mandate standing orders

• Recommend improving postnatal surveillance to capture the true perinatal infection burden they admit is undercounted

Why This Is a Profound Logical Problem

The authors construct their argument as a binary:

Screening is unreliable → therefore universal birth dose is necessary

But the logically complete set of responses to “screening is unreliable” includes:

1. Universal birth dose regardless of screening (their preferred solution)

2. Fix the screening (never mentioned)

3. Both simultaneously (never mentioned)

4. Stratify risk and target interventions (dismissed without engagement)

By never calling for improved screening, the authors are implicitly treating screening failure as a permanent, immutable feature of the US healthcare system rather than a remediable problem. This is a form of learned helplessness embedded in the policy argument.

All the authors want is a “must remain universal” conclusion. That is a textbook example of “analysis-to-result” (confirmation bias) and their silence on improved screening is deafening.

Europe, Systems, and the Unasked Question

Most EU/EEA countries do not implement universal birth dose policies. They rely on high prenatal screening coverage and infant vaccination beginning later in infancy. Many have achieved low prevalence and elimination benchmarks.

The paper acknowledges this and attributes the difference to U.S.-specific healthcare fragmentation.

But it does not ask a deeper systems question:

Is universal birth dose compensating for healthcare system failures that could instead be addressed through improved screening fidelity, data integration, and postpartum follow-up?

If birth dose functions as a workaround for structural fragmentation, that is a systems design issue — not purely an immunologic one.

That possibility is never explored.

Unquantified Warnings of “Resurgence”

The paper warns that declining coverage could lead to future resurgence, with susceptibility evolving over years.

That is plausible.

But no data are provided. Not evening modeling is provided.

No scenario analyses. No probabilistic estimates. No cost-effectiveness comparisons.

Unquantified warnings are rhetorical devices, not evidence. A paper that demands rigorous, systematic review standards from ACIP should meet those standards when forecasting harm.

Reputational Borrowing of GRADE

The authors describe GRADE and EtR frameworks in detail and emphasize their importance in guiding vaccine policy. They conclude by stating that structured, systematic, and transparent reviews remain paramount.

Yet they did not conduct such a review.

They borrow the authority of systematic architecture while declining to perform it. That is epistemic asymmetry in its purest form.

Institutional Timing and Context

The Vaccine Integrity Project was launched in April 2025. A CIDRAP report closely related to this article was dated December 2, 2025 — three days before the December 5 ACIP vote it purports to address.

The Pediatrics article appears in February 2026 as a “Special Article,” co-authored by a former CDC director.

The pattern is not organic academic curiosity. It is organized policy defense.

That does not invalidate the science. But it contextualizes the motivation.

Nothing forecloses legitimate debate about optimal timing in a contemporary low-prevalence environment for infants born to confirmed HBsAg-negative mothers. But this paper actually destroys the credibility of the argument that ACIP got it wrong.

Policy recalibration in response to objective assessment of the evidence is not anti-vaccine. It is what evidence-based medicine and logical, reasonable and ethical public health policy requires.

The Core Asymmetry

When defending the birth dose, the paper relies on historical success, trend associations, and absence of new safety signals to declare “strong evidence.”

When evaluating the ACIP shift, it demands modeling, warns of resurgence, and emphasizes uncertainty.

The evidentiary burden is unequal.

If critics of vaccines must meet strict methodological standards, defenders of existing policy must meet those same standards when declaring that no alternative is supported.

Otherwise, public trust erodes — not because the vaccine is unsafe, but because institutional confidence outpaces methodological rigor.

The science supports the vaccine.

Whether universal birth dose remains the optimal policy for every infant in 2026 is a narrower, quantitative question.

That question was not systematically answered here.

And disclaimers do not convert narrative comprehensiveness into adjudicative rigor.

NB: This critique could have continue on the lack of counterfactual analysis. The paper never defines:

• What epidemiologic threshold would justify reconsideration?

• What number of cases prevented per 100,000 births constitutes policy indispensability?

• What absolute risk difference is being defended?

• What is the number needed to treat (NNT) in the HBsAg-negative subgroup?

Without counterfactual modeling, the authors cannot logically claim that “no evidence supports change,” because they never define what magnitude of benefit is required to maintain universality.

There is also no absolute risk reduction assessment. The paper relies heavily on relative reductions (99% decline).

• Absolute per-birth risk

• Absolute reduction attributable to birth dose timing

• Confidence intervals around residual risk

• Temporal trend stratified by maternal HBsAg status

Relative reduction rhetoric is epidemiologically persuasive but policy decisions hinge on absolute risk in defined subgroups.

There is no sensitivity analysis

While screening sensitivity is high (97–100%), the paper does not:

• Perform best-case / worst-case modeling

• Provide upper-bound estimates for false-negative impact

• Estimate infections under varying screening failure rates

• Model impact if screening improved modestly

Without sensitivity analysis, the claim that universal birth dose is indispensable rests on untested assumptions about screening failure permanence.

This is a major methodological absence.

The paper does not examine timing-specific immunological context.

The authors show some evidence of birth-dose immunogenicity. But They do not show:

• Inferiority of first-dose-at-2-months in HBsAg-negative infants

• Breakthrough infection rates before 2 months in low-risk infants

• Timing-specific comparative seroconversion curves

• Non-inferiority margins

They do not demonstrate timing indispensability. That is a different evidentiary burden.

There is no heterogeneity analysis.

The paper treats U.S. infants as homogeneous.

There is no:

• Stratification by maternal country of origin

• Stratification by geographic prevalence

• Stratification by risk environment

• Urban vs rural variation

• Medicaid vs private care differences

If screening failure is heterogeneous, then universal policy justification requires demonstrating that risk is diffuse rather than concentrated, which they never show.

There is no cost-effectiveness or resource allocation analysis.

• They argue universality must remain. They do not:

• Estimate cost per infection prevented in the HBsAg-negative subgroup

• Compare marginal benefit vs marginal cost

• Evaluate alternative resource investment (screening improvement vs universal dosing).

Even if safety is high, public health decisions require resource justification. They omit that entirely.

• There is no evidence of surveillance sensitivity.

• They cite <20 annual perinatal cases.

• They admit undercounting.

They do not:

• Estimate surveillance sensitivity

• Provide capture-recapture estimates

• Quantify underreporting factor

• Provide confidence intervals for “true” burden

If resurgence is feared, surveillance performance must be analyzed.

They do not do that.

There is no explicit null hypothesis testing.

This is subtle but important. The paper never defines:

• What is the null hypothesis?

• Is the null “birth dose is unnecessary in HBsAg-negative infants”?

• Or is the null “birth dose remains indispensable”?

They implicitly treat universality as the default state requiring no quantitative defense.

That is a status-quo bias embedded as methodology.

There is no conflict-of-interest transparency.

Structural proximity to policy defense and the past careers of many of the authors tell us this is an argument, not science. Their claim of independence cannot be accepted at face value.

They abuse logic, evidence and misuse languge.

The article uses phrases like:

• “Strong evidence”

• “Clear benefit”

• “No evidence supports change”

Without:

• Quantitative grading

• Effect size tables

• Certainty stratification

A short language audit section in publications like these would allow reader and editors to see disconnects. The absence of structured grading is devastating.

There is no harm-benefit framing specific to subgroups

They do not show:

• Benefit magnitude in HBsAg-negative subgroup

• Net clinical benefit comparison (birth dose vs delayed dose)

• Timing-specific adverse event comparison

• Injection burden consideration in low-risk infants

Safety in general is not the same as subgroup net-benefit necessity.

There is no explicit burden of proof clarification.

• Who bears the burden of proof?

• Those proposing change?

• Or those defending universality?

The paper assumes burden lies entirely with reformers.

That is a philosophical assumption disguised as methodology.

They dismiss, and do not compare, other countries.

They dismiss Europe.

They do not:

• Compare perinatal infection rates longitudinally

• Compare chronic carrier rates

• Compare surveillance quality

• Provide data on elimination benchmarks

• Adjust for immigration prevalence differences