Garbage‑In, Garbage‑Out. Ad‑Hoc Vaccination “Recommendation” Policies by AAP and ACOG Reflect Pre‑Kennedy‑Era “Science” and Are Thereby Disqualified

AAP and ACOG's vaccine recommendations may be hazardous to your child's health. Due to data chaos, no one knows.

Sep 23, 2025

Recent JAMA coverage attempts to normalizes AAP and ACOG’s stand‑alone COVID‑19 vaccination policies for infants and for pregnancy after CDC reversed or narrowed its federal recommendations. However, we see the cracks all over their veneer.

JAMA Medical News framed ACOG’s decision to continue recommending COVID‑19 vaccination in pregnancy as a benign deviation from CDC policy, and ran a Viewpoint arguing that changes for children and pregnant women were a “failure of process, policy, and science.” (Rubin, 2025; Gostin et al., 2025). ACOG simultaneously issued press guidance reaffirming maternal vaccination; AAP issued a new pediatric schedule that diverges from CDC, recommending universal vaccination for ages 6–23 months and risk‑based dosing for older children (Committee on Infectious Diseases, 2025; ACOG, 2025; HealthyChildren.org, 2025; Schnirring, 2025; Reuters, 2025).

Those policies rest on evidence streams that do not satisfy current causal‑inference standards. The studies that AAP/ACOG lean on—registry cohorts, test‑negative designs, and modeling exercises—import unresolved biases at the design stage. On just one point alone, co-administration of different vaccines in the same office visit, no robust data exist, only online argumentation. The assumption of safety has been combined with ideological screens based on denialism.

Recommendations built on that weak scaffolding transmit those biases unchanged. That is textbook garbage‑in, garbage‑out.

What JAMA Has Platformed

JAMA ran multiple pieces spotlighting ACOG’s decision to keep recommending COVID‑19 vaccination during pregnancy after CDC stepped back; JAMA also characterized CDC’s change for children and pregnancy as a “failure of process, policy, and science.” (Rubin, 2025; Gostin et al., 2025).That editorial framing matters: it confers scientific legitimacy on ACOG/AAP positions that now drive insurance coverage and clinical defaults. Yet legitimacy follows method, not press statements.

Shimabukuro was responsible for analyses leading to the “all-clear” for influenza vaccination during pregnancy as well.

Similarly, the very foundation for recommending Tdap during pregnancy — the 2020 Beccera et al. review — was systematically dismantled by Lyons-Weiler and colleagues (2022). The review ignored fetal immunotoxicity studies, animal developmental toxicology data, and biodistribution findings, while accepting surveillance data with known underreporting. Any continued recommendation for Tdap in pregnancy based on this review is unsupportable.

Why AAP/ACOG’s evidence does not clear a causal bar

ACOG’s position depends almost entirely on a now-debunked safety claim published by Shimabukuro et al. (June 2021). That paper reported no increased risk of miscarriage among vaccinated pregnant women but used a denominator that included women vaccinated after the first trimester — mathematically impossible to miscarry in the first trimester if you weren't vaccinated then. This flawed analysis was never corrected or retracted. Dr. Tom Shimabukuro, head of CDC’s Immunization Safety Office, presented the same data to ACIP as fact. The entire pregnancy safety claim should be viewed as irreparably compromised.

The pediatric and maternal recommendations lean on observational datasets and surveillance‑linked analyses that cannot, as configured, resolve the direction and size of effect. Specific, well‑characterized biases dominate these designs:

a) Immortal‑time and time‑zero errors. In registry cohorts, vaccinated person‑time frequently begins after a delay while unvaccinated time starts at calendar day 0. Any events before the shot cannot occur “in” the vaccinated stratum by definition, giving vaccinees free, event‑free time. This inflates apparent protection. The pharmacoepidemiology literature has treated this for two decades (Suissa, 2008).

b) Healthy‑vaccinee selection. People who obtain vaccination earlier are healthier, wealthier, and more health‑seeking. Classic pre‑season checks show large apparent mortality benefits before influenza season even begins—pure bias. COVID‑era cohorts inherit the same structure unless they measure and adjust functional status and care‑seeking directly (Jackson et al., 2006).

c) Depletion‑of‑susceptibles. As the most exposure‑prone or susceptible individuals get infected first, later comparisons pit “hardier survivors” in one arm against a mixed group in the other, distorting waning and sometimes producing spuriously negative or positive effectiveness (Lipsitch et al., 2019).

d) Test‑negative collider bias and unequal testing. Maternal and pediatric effectiveness studies often use test‑negative designs. If vaccinated patients test for screening, travel, or precaution, while unvaccinated test mainly when sicker, conditioning on “tested” induces bias. Negative‑control designs using the “recently vaccinated” group exist precisely to detect this bias, but are rarely mandatory in policy‑facing studies (Hitchings et al., 2022).

e) Outcome misclassification (“with” vs “for”). Hospital labels often count any SARS‑CoV‑2‑positive admission as a “COVID‑19 hospitalization.” EHR‑phenotyping studies show that a nontrivial share are incidental positives; direction of bias then depends on screening intensity by vaccination status. Maternal and infant outcomes are especially sensitive to this (Klann et al., 2022).

f) Prior infection and hybrid immunity. Unmeasured infection histories bias estimates in either direction. Simulations and empirical checks show that failing to account for baseline infection‑derived protection can even flip the sign of vaccine effectiveness in test‑negative designs (Wiegand et al., 2024).

g) Target‑trial misalignment. Most maternal/pediatric observational studies do not explicitly emulate a target trial (eligibility, time zero, dynamic treatment strategies, censoring). Without that, design drift accumulates the biases above (Hernán & Robins, 2016).

Reality check. JAMA’s own Medical News has acknowledged contemporaneous FDA label actions (e.g., myocarditis/pericarditis class warnings) while platforming continued blanket pregnancy recommendations; this juxtaposition illustrates the evidence conflict that ad‑hoc guidance papers paper over rather than resolve (Prasad & Makary, 2025).

3) Why this disqualifies ad‑hoc recommendations.
Recommendations are claims of causal benefit over harm. When the underlying evidence is dominated by the biases above, causal direction is uncertain and magnitude is unstable. Publishing and promoting recommendations that omit these limitations is not “science communication”; it is policy laundering. JAMA’s platform should not normalize it.

The minimum standard for any pregnancy or infant COVID‑19 recommendation.

Target‑trial emulation up front. Define eligibility, time‑zero alignment, treatment strategies, censoring, and follow‑up before analysis; publish the DAG and the code (Hernán & Robins, 2016).

Symmetric risk windows. Count outcomes from day 0 and from day X in parallel; report how “grace windows” alter effect sizes.
Active control for prior infection. Measure infection history (serology or documented positives) and analyze strata accordingly; otherwise, call the estimate descriptive, not causal (Wiegand et al., 2024).
Negative‑control analyses. Use recently vaccinated as an exposure negative control and non‑COVID endpoints as outcome negative controls, reporting both alongside the headline VE (Hitchings et al., 2022).
Outcome specificity. Prefer oxygen‑requirement/ICU definitions over “any admission with a positive swab”; validate the phenotype locally (Klann et al., 2022).
Transparent time‑varying hazards. Report time‑since‑dose hazards with depletion‑of‑susceptibles diagnostics. Publish full datasets for re-analysis (Lipsitch et al., 2019).
Fixed and pre-published data analysis plan and not adjusting for copredictors as confounders.

Until AAP and ACOG meet this bar, their policies are not evidence‑based; they are preference‑based. JAMA should stop treating them as interchangeable with science.

AAP and ACOG Are Not Institutions with Scientific Regulatory Oversight. They are Trade Associations.

AAP and ACOG are not scientific oversight bodies. They are medical trade associations, structurally aligned with manufacturer influence, and incentivized to preserve professional consensus. Their policy statements, while often treated as scientific pronouncements, are not based on gold-standard evidence and do not meet legal or clinical standards for establishing causality.

What JAMA should publish instead

The public now requires they adopt The Kennedy Bar. At a minimum, this entails (1) methods‑first audits of the key maternal and pediatric datasets against the target‑trial checklist; (2) side‑by‑side re‑analyses showing how immortal‑time removal, prior‑infection control, and negative‑control checks change the estimates; (3) a clear separation of model‑based “lives saved” narratives from design‑based effect estimates; and (4) explicit discussion of how federal policy changes alter baseline risk and acceptable uncertainty for new recommendations.

The policy stakes are not theoretical. Litigation and petitions have already pressed these evidentiary issues into courts, underscoring why methods—not messaging—must rule.

Bottom line

As trade associations, AAP and ACOG can maintain any clinical preference they like. They cannot call it science until their underlying studies satisfy basic design invariants: aligned time zero, symmetric risk windows, credible control for prior infection, negative‑control diagnostics, validated outcomes, and pre‑registered analysis plans that anyone can rerun. JAMA’s role is to enforce those invariants—not to launder ad‑hoc policies into de facto standards.

Documentation of positions reported (policy and news)

Gostin, L. O., Reiss, D., & Offit, P. A. (2025). Changed recommendations for COVID-19 vaccines for children and pregnant women. JAMA, 334(8), 663. https://doi.org/10.1001/jama.2025.10658
Rubin, R. (2025). The CDC no longer recommends COVID-19 shots during pregnancy—now what? JAMA, 334(6), 469. https://doi.org/10.1001/jama.2025.11889
Prasad, V., & Makary, M. A. (2025). US FDA Safety Labeling Change for mRNA COVID-19 Vaccines. JAMA. https://doi.org/10.1001/jama.2025.12675
ACOG: Updated maternal immunization guidance (COVID‑19, influenza, RSV) reaffirming vaccination in pregnancy (ACOG, 2025).
AAP: 2025–26 policy recommending COVID‑19 vaccination for all children 6–23 months and risk‑based vaccination for older ages. Policy statement and press materials (Committee on Infectious Diseases, 2025; HealthyChildren.org, 2025; Schnirring, 2025; Reuters, 2025).

Methods references (minimum causal‑inference toolkit)

Suissa, S. (2008). Immortal time bias in pharmacoepidemiology. American Journal of Epidemiology, 167(4), 492–499. https://doi.org/10.1093/aje/kwm324
Jackson, L. A., Jackson, M. L., Nelson, J. C., Neuzil, K. M., & Weiss, N. S. (2005). Evidence of bias in estimates of influenza vaccine effectiveness in seniors. International Journal of Epidemiology, 35(2), 337–344. https://doi.org/10.1093/ije/dyi274
Lipsitch, M., Goldstein, E., Ray, G. T., & Fireman, B. (2019). Depletion-of-susceptibles bias in influenza vaccine waning studies: How to ensure robust results. Epidemiology and Infection, 147. https://doi.org/10.1017/s0950268819001961
Hitchings, M. D. T., Lewnard, J. A., Dean, N. E., Ko, A. I., Ranzani, O. T., Andrews, J. R., & Cummings, D. A. T. (2022). Use of recently vaccinated individuals to detect bias in test-negative case–control studies of COVID-19 vaccine effectiveness. Epidemiology, 33(4), 450–456. https://doi.org/10.1097/ede.0000000000001484
Wiegand, R. E., Fireman, B., Najdowski, M., Tenforde, M. W., Link-Gelles, R., & Ferdinands, J. M. (2024). Bias and negative values of COVID-19 vaccine effectiveness estimates from a test-negative design without controlling for prior SARS-CoV-2 infection. Nature Communications, 15(1). https://doi.org/10.1038/s41467-024-54404-w
Klann, J. G., Strasser, Z. H., Hutch, M. R., Kennedy, C. J., Marwaha, J. S., Morris, M., Samayamuthu, M. J., Pfaff, A. C., Estiri, H., South, A. M., Weber, G. M., Yuan, W., Avillach, P., Wagholikar, K. B., Luo, Y., Omenn, G. S., Visweswaran, S., Holmes, J. H., Xia, Z., … Murphy, S. N. (2022). Distinguishing admissions specifically for COVID-19 from incidental sars-cov-2 admissions: National retrospective electronic health record study. Journal of Medical Internet Research, 24(5), e37931. https://doi.org/10.2196/37931
Hernán, M. A., & Robins, J. M. (2016). Using big data to emulate a target trial when a randomized trial is not available: Table 1. American Journal of Epidemiology, 183(8), 758–764. https://doi.org/10.1093/aje/kwv254

The Problematic Study of on COVID-19 Vaccination During Pregnancy

Shimabukuro, T. T., Kim, S. Y., Myers, T. R., Moro, P. L., Oduyebo, T., Panagiotakopoulos, L., Marquez, P. L., Olson, C. K., Liu, R., Chang, K. T., Ellington, S. R., Burkel, V. K., Smoots, A. N., Green, C. J., Licata, C., Zhang, B. C., Alimchandani, M., Mba-Jonas, A., Martin, S. W., … Meaney-Delman, D. M. (2021). Preliminary Findings of mRNA Covid-19 Vaccine Safety in Pregnant Persons. New England Journal of Medicine, 384(24), 2273–2282. https://doi.org/10.1056/nejmoa2104983

Full Reference List

ACOG. (2025). ACOG releases updated maternal immunization guidance COVID influenza RSV. ACOG. https://www.acog.org/news/news-releases/2025/08/acog-releases-updated-maternal-immunization-guidance-covid-influenza-rsv?

Becerra-Culqui, T. A., Getahun, D., Chiu, V., Sy, L. S., & Tseng, H. F. (2020). The association of prenatal tetanus, diphtheria, and acellular pertussis (tdap) vaccination with attention-deficit/hyperactivity disorder. American Journal of Epidemiology, 189(10), 1163–1172. https://doi.org/10.1093/aje/kwaa074

Committee on Infectious Diseases. (2025). Recommendations for COVID-19 vaccines in infants, children, and adolescents: Policy statement. Pediatrics. https://doi.org/10.1542/peds.2025-073924

Gostin, L. O., Reiss, D., & Offit, P. A. (2025). Changed recommendations for COVID-19 vaccines for children and pregnant women. JAMA, 334(8), 663. https://doi.org/10.1001/jama.2025.10658

HealthyChildren.org. (2025). AAP releases its own evidence-based immunization schedule. HealthyChildren.Org. https://www.healthychildren.org/English/news/Pages/AAP-releases-its-own-evidence-based-immunization-schedule.aspx?

Hernán, M. A., & Robins, J. M. (2016). Using big data to emulate a target trial when a randomized trial is not available: Table 1. American Journal of Epidemiology, 183(8), 758–764. https://doi.org/10.1093/aje/kwv254

Hitchings, M. D. T., Lewnard, J. A., Dean, N. E., Ko, A. I., Ranzani, O. T., Andrews, J. R., & Cummings, D. A. T. (2022). Use of recently vaccinated individuals to detect bias in test-negative case–control studies of COVID-19 vaccine effectiveness. Epidemiology, 33(4), 450–456. https://doi.org/10.1097/ede.0000000000001484

Jackson, L. A., Jackson, M. L., Nelson, J. C., Neuzil, K. M., & Weiss, N. S. (2005). Evidence of bias in estimates of influenza vaccine effectiveness in seniors. International Journal of Epidemiology, 35(2), 337–344. https://doi.org/10.1093/ije/dyi274

Klann, J. G., Strasser, Z. H., Hutch, M. R., Kennedy, C. J., Marwaha, J. S., Morris, M., Samayamuthu, M. J., Pfaff, A. C., Estiri, H., South, A. M., Weber, G. M., Yuan, W., Avillach, P., Wagholikar, K. B., Luo, Y., Omenn, G. S., Visweswaran, S., Holmes, J. H., Xia, Z., … Murphy, S. N. (2022). Distinguishing admissions specifically for COVID-19 from incidental sars-cov-2 admissions: National retrospective electronic health record study. Journal of Medical Internet Research, 24(5), e37931. https://doi.org/10.2196/37931

Lipsitch, M., Goldstein, E., Ray, G. T., & Fireman, B. (2019). Depletion-of-susceptibles bias in influenza vaccine waning studies: How to ensure robust results. Epidemiology and Infection, 147. https://doi.org/10.1017/s0950268819001961

Lyons-Weiler, J., Fujito, A., & Pajer, B. (2022). Maternal Gestational Tdap Vaccination and Autism: A Critique of Becerra-Culqui et al. (2018). International Journal of Vaccine Theory, Practice, and Research, 2(1), 287–300. https://doi.org/10.56098/ijvtpr.v2i1.39

Prasad, V., & Makary, M. A. (2025). US FDA Safety Labeling Change for mRNA COVID-19 Vaccines. JAMA. https://doi.org/10.1001/jama.2025.12675

Reuters. (2025, August 20). US pediatric group breaks with federal policy, recommends COVID vaccines for young children. Reuters. https://www.reuters.com/business/healthcare-pharmaceuticals/us-pediatric-group-breaks-with-federal-policy-recommends-covid-vaccines-young-2025-08-19/?

Rubin, R. (2025). The CDC no longer recommends COVID-19 shots during pregnancy—now what? JAMA, 334(6), 469. https://doi.org/10.1001/jama.2025.11889

Schnirring, L. (2025, August 19). AAP evidence-backed immunization schedule reflects break from CDC advisers. CIDRAP. https://www.cidrap.umn.edu/covid-19/aap-evidence-backed-immunization-schedule-reflects-break-cdc-advisers?

Suissa, S. (2008). Immortal time bias in pharmacoepidemiology. American Journal of Epidemiology, 167(4), 492–499. https://doi.org/10.1093/aje/kwm324

Wiegand, R. E., Fireman, B., Najdowski, M., Tenforde, M. W., Link-Gelles, R., & Ferdinands, J. M. (2024). Bias and negative values of COVID-19 vaccine effectiveness estimates from a test-negative design without controlling for prior SARS-CoV-2 infection. Nature Communications, 15(1). https://doi.org/10.1038/s41467-024-54404-w