Publication

• Title: A randomized, placebo-controlled trial of 13-valent pneumococcal conjugate vaccination to accelerate immune recovery after sepsis
• Acronym: VACIRiSS
• Year: 2026
• Journal published in: Science Translational Medicine
• Citation: Shankar-Hari M, Smith P, Szakmany T, Summers C, McGuigan PJ, Brealey D, et al. A randomized, placebo-controlled trial of 13-valent pneumococcal conjugate vaccination to accelerate immune recovery after sepsis. Sci Transl Med. 2026;18(862):eaeb4113.

Context & Rationale

• Background

  • Survival from the acute septic episode does not mark the end of risk. By one year, approximately 45% to 50% of sepsis survivors are rehospitalised and about 15% have died; infection is among the commonest reasons for readmission.¹²
  • Post-sepsis immune dysfunction is biologically complex rather than a simple sequential transition from hyperinflammation to immunoparalysis. Persistent inflammation, lymphocyte apoptosis and depletion, impaired antigen presentation, altered innate and adaptive cell function, and immune ageing can coexist within the same patient population.³
  • PCV13 was an attractive experimental probe because it was licensed, familiar, generally safe, and capable of inducing T-cell-dependent pneumococcal antibody responses. Its established clinical efficacy, however, was against vaccine-serotype pneumococcal disease rather than infections in general; in CAPiTA it reduced vaccine-type disease but did not significantly reduce all-cause community-acquired pneumonia.⁴
  • A conjugate vaccine was chosen instead of a live-attenuated vaccine because the depth and duration of post-sepsis immunosuppression were uncertain; PCV13 offered a clinically familiar antigen challenge without the replication risk of a live organism.
  • VACIRiSS therefore tested a more ambitious proposition than conventional pneumococcal vaccination: that PCV13 could act both as a controlled antigen challenge and as a non-specific, or heterologous, immune stimulus capable of improving clinically important post-sepsis outcomes.
• Research Question/Hypothesis

  • Clinical hypothesis: a single dose of PCV13 administered during recovery from sepsis would prolong the time to first infection-related rehospitalisation or death over 365 days compared with saline placebo.
  • Biological hypothesis: PCV13 would produce serotype-specific IgG responses and measurable non-specific effects across lymphocyte subsets, cytokines and the whole-blood transcriptome, thereby accelerating immune recovery.
• Why This Matters

  • This was the first randomised trial to test vaccination as a therapeutic immune intervention in adult sepsis survivors, rather than merely describing post-sepsis immune abnormalities.
  • A one-off, inexpensive and widely deployable intervention at the transition from critical care to ward care would be highly attractive if it reduced recurrent infection, rehospitalisation or death.
  • The study also addressed a central translational problem in sepsis: whether a uniform immune stimulus can benefit a clinically and immunologically heterogeneous syndrome, or whether treatment must be selected according to mechanism, timing and host phenotype.

Design & Methods

• Research Question: In adult ICU or high-dependency-unit survivors recovering from sepsis, does one intramuscular dose of PCV13, compared with saline placebo, reduce the rate of first infection-related rehospitalisation or death within 365 days, and what antigen-specific and broader immune responses does it generate?
• Study Type:
  • Investigator-initiated, phase IV, multicentre, parallel-group, superiority, 1:1 randomised, quadruple-blinded, saline placebo-controlled trial with an internal pilot phase.
  • Conducted in 13 adult general critical care units in UK National Health Service hospitals.
  • Recruitment ran from 27 July 2018 to 22 April 2022; follow-up was completed on 22 April 2023.
  • Registered as NCT03565159 and ISRCTN11401066; funded by the National Institute for Health and Care Research. Pfizer supplied PCV13.⁵
• Population:
  • Screening and enrolment: 4,033 adult sepsis survivors were assessed and 214 were randomised: 104 to PCV13 and 110 to placebo.
  • Inclusion: age ≥18 years; registered with a general practitioner; ICU/HDU admission for suspected or proven infection with SOFA ≥2; clinically improving and expected to step down to HDU or ward care within 24 to 48 hours; consent from the patient, legal representative or professional consultee.
  • Key exclusions: temperature ≥38.0°C in the preceding 24 hours; vaccine or diphtheria-toxoid hypersensitivity; another vaccination within 7 days; pregnancy or lactation; treatment limitations; nursing-home or long-term-care residence; platelets <50 × 10⁹/L or INR >1.3; splenectomy; pneumococcal sepsis; and APACHE II-defined immunodeficiency or immunosuppression, including HIV, recent malignancy treatment, dialysis or transplantation, nephrotic syndrome, prolonged systemic corticosteroids or other immunosuppressive treatment.
  • Illness severity: mean APACHE II score was 16.5 ± 6.4 with PCV13 and 16.8 ± 7.9 with placebo; median SOFA score was 7 (IQR 4 to 9) and 7 (IQR 4.5 to 10), respectively.
  • Timing in critical illness: median critical care stay before randomisation was 5.3 days (IQR 2.5 to 8.7) and 5.4 days (IQR 3.1 to 8.7), respectively.
• Intervention:
  • One 0.5-mL intramuscular dose of 13-valent pneumococcal conjugate vaccine, PCV13 (Prevenar 13).
  • The dose was intended within 72 hours of randomisation; administration could be delayed if the participant deteriorated and resumed when clinically improved.
  • Blood was obtained before vaccination and at approximately day 10 and day 30 for serotype-specific IgG, immunophenotyping, cytokine and transcriptomic analyses.
• Comparison:
  • One 0.5-mL intramuscular injection of 0.9% sodium chloride prepared in an identically masked syringe.
  • All other treatment was determined by the clinical teams. Other vaccinations during follow-up were recorded but were not prohibited.
• Blinding: Participants, site research staff, the clinicians administering the injection, statisticians, follow-up personnel, data-linkage teams and biological laboratories were blinded. Separate unblinded pharmacy or preparation staff dispensed the allocated injection. This was unusually comprehensive masking for a mechanistic critical care trial.
• Statistics:
  • Power calculation: assuming a 32.1% one-year incidence of infection-related rehospitalisation or death, 214 participants provided 90% power at a two-sided 5% significance level to detect a hazard ratio of 0.36—a 64% relative reduction.
  • Primary analysis: intention-to-treat in all participants whose data could be used, with age-stratified incidence-rate ratios and incidence-rate differences as the principal population summaries; a Cox model provided the hazard ratio, with adjustment for age, site frailty and the Sepsis Survivor Prognosis score.
  • Secondary analyses: binary outcomes were analysed with risk ratios; time-to-event outcomes used cause-specific Cox models. No adjustment for multiplicity was made, and missing clinical outcomes were not imputed.
  • One participant in each group withdrew permission to use data on the day of randomisation, leaving 103 and 109 participants in the primary analysis.⁵
• Follow-Up Period: 365 days after randomisation, with direct assessments around days 10, 30 and 90, telephone follow-up around days 180 and 365, and linkage to hospital, mortality and general-practice data.

Key Results

This trial was not stopped early. It reached the planned sample of 214 participants and completed one-year follow-up.

• The primary estimate lay on the side of possible harm rather than benefit, but its wide interval was compatible with effects ranging from a modest reduction to a substantial increase in hazard; the study excluded the very large benefit it was designed to detect, not every clinically relevant benefit.
• The recurrent-infection and serious-adverse-event findings should be taken seriously but remain exploratory: the reinfection definition was broad, several comparisons were made without multiplicity adjustment, risk-ratio and risk-difference intervals were discordant, and no serious event was attributed to vaccination.
• PCV13 produced measurable antigen-specific and cellular responses without an excessive cytokine response, but this biological activity did not translate into better clinical outcomes.

Internal Validity

• Randomisation and Allocation: A central password-protected web system used variable blocks and stratification by site and age (≤65 versus >65 years). Allocation was concealed until randomisation, and there is no evidence of foreknowledge or selection manipulation.
• Dropout or Post-randomisation Exclusion: One participant in each group withdrew permission to use data on the day of randomisation. Three of 104 PCV13 participants and 5 of 110 placebo participants did not receive the allocated injection; 3 and 6, respectively, were lost to follow-up. Primary analyses therefore included 103 and 109 participants, giving low and reasonably balanced attrition.
• Performance and Detection Bias: Blinding extended well beyond participants and bedside investigators to injection administrators, follow-up personnel, statisticians, data-linkage staff and laboratories. Separate unblinded teams prepared the injections. Similar local and systemic reaction rates make major functional unblinding unlikely.
• Protocol Adherence: The allocated intervention was given to 101/104 (97%) in the PCV13 group and 105/110 (95%) in the placebo group. No crossover was reported. The actual distribution of time from randomisation to injection, and the number requiring delayed administration after deterioration, were not reported.
• Baseline Characteristics: Age, comorbidity burden, APACHE II, SOFA score and pre-randomisation critical care stay were similar. Chance imbalances remained: moderate preadmission dependency was 9.6% versus 22.7%, medical admission 60.6% versus 70.0%, gastrointestinal infection 30.1% versus 21.8%, and bloodstream infection 3.9% versus 10.9% in PCV13 versus placebo.
• Heterogeneity: Participants differed in infection source, prior health, inflammatory state and immune phenotype. Neither age (interaction P=0.26) nor Sepsis Survivor Prognosis category (interaction P=0.55) identified a different clinical effect, but the trial was not powered to detect plausible subgroup interactions and did not select patients by immune biomarker.
• Timing: Randomisation occurred after a median of approximately 5.3 to 5.4 days in critical care, when step-down was expected within 24 to 48 hours. This was clinically practical and targeted the period of greatest subsequent risk, but conventional antibody protection develops over days to weeks while events were counted immediately after randomisation.
• Dose: The standard licensed 0.5-mL PCV13 dose was appropriate for testing ordinary vaccine immunogenicity and demonstrated biological activity. It was not established as the optimal dose, platform or schedule for non-specific immune modulation after sepsis.
• Separation of the Variable of Interest: Treatment delivery was well separated and day-30 serotype-specific IgG rose for 6 of 13 serotypes with PCV13. The response was nevertheless variable, and the study did not measure opsonophagocytic function or protection against pneumococcal infection.
• Prior Vaccination and Co-interventions: Previous pneumococcal vaccination was reported in 24/104 (23.3%) versus 19/110 (17.3%), and status was unknown in 15.1% versus 11.8%. Other vaccines during follow-up were allowed. Thus, PCV13 represented a first exposure for some participants and revaccination for others, adding immunological heterogeneity that was not a randomisation stratum.
• Outcome Assessment: Mortality ascertainment through civil registration was objective. Infection-related rehospitalisation and “reinfection” relied on case-report data, ICD-10 coding and antibiotic prescribing rather than independent clinical adjudication or microbiological confirmation, increasing the risk of non-differential misclassification.
• Statistical Rigor: The principal effect measures were broadly concordant, and sensitivity analysis restricted to the pre-COVID period gave HR 1.11; 95% CI 0.58 to 2.11. However, the study was powered only for HR 0.36, secondary analyses were not multiplicity-adjusted, and the final SAP was completed after recruitment, although before database lock.

Conclusion on Internal Validity: Internal validity is strong for the conclusion that PCV13 did not produce the very large clinical benefit hypothesised. It is moderate overall for quantifying realistic benefit or harm because the sample was small, the infection outcomes lacked microbiological specificity, immune exposure was heterogeneous, and secondary safety and reinfection findings were vulnerable to multiplicity and misclassification.

External Validity

• Recruitment Fraction: Only 214 of 4,033 screened patients (5.3%) were randomised. Major reasons for exclusion included APACHE II-defined immune comorbidity (n=1,502), limitations of care (n=626), pneumococcal sepsis (n=89), refusal (n=317) and multiple logistical or social reasons (n=942).
• Selected Survivor Phenotype: Nursing-home residents, patients with treatment limitations, major immunosuppression, dialysis or transplant dependence, recent malignancy treatment, HIV, splenectomy and current pneumococcal sepsis were excluded. These groups are common among real-world sepsis survivors and often carry both the highest recurrent-infection risk and established indications for pneumococcal vaccination.
• Participant Characteristics: Mean age was approximately 60 years, 88% to 94% self-reported as white, and most participants had been independent before admission. Only 13 deaths occurred among 212 analysed participants over one year, substantially fewer than the approximately 15% post-discharge mortality described in broad sepsis-survivor cohorts, consistent with selection of a fitter population.
• Representativeness Within the UK: Compared with contemporaneous UK critical care data, participants were broadly representative but marginally older, more severely ill and more often white. The trial therefore informs recovering, relatively independent, non-immunosuppressed UK ICU survivors better than the whole post-sepsis population.
• Healthcare-System Applicability: Central data linkage, NHS general-practice records and hospital coding supported follow-up but may not transfer directly to systems without national records or comparable rehospitalisation pathways. Applicability to resource-limited settings, ethnically diverse populations and different vaccination programmes is uncertain.
• Intervention Relevance: The trial studied PCV13. The current UK adult and clinical-risk programme has moved to higher-valency PCV20, so direct extrapolation to the product now used in routine adult practice is limited.

Conclusion on External Validity: External validity is limited to moderate. The findings apply most directly to relatively fit, recovering, non-immunosuppressed adults in UK critical care and should not be used to withhold indicated pneumococcal vaccination from older adults or recognised clinical-risk groups, nor assumed to predict the effects of PCV20.

Strengths & Limitations

• Strengths:
  • First randomised, placebo-controlled clinical test of vaccination as an immune-recovery intervention after sepsis.
  • Thirteen-centre UK design, central concealed randomisation and unusually comprehensive quadruple blinding.
  • High treatment fidelity, low attrition and routine-data linkage across a clinically important one-year follow-up period.
  • A clinically meaningful primary outcome that incorporated death into the composite rather than treating it simply as censoring.
  • Embedded, allocation-blinded immunology spanning serotype-specific antibodies, lymphocyte phenotyping, cytokines and whole-blood transcriptomics, with serology performed in a WHO reference laboratory.
  • Investigator-initiated public funding, prespecified clinical analyses and availability of biological source data and analysis code.
• Limitations:
  • The sample size was designed around an implausibly large HR of 0.36 and was consequently underpowered for modest but clinically meaningful effects.
  • The composite was dominated by infection-related rehospitalisation and included all-cause death, which is clinically important but biologically non-specific.
  • Infection outcomes were derived from diagnostic coding and antibiotic use without microbiological confirmation or blinded event adjudication.
  • Only 5.3% of screened patients were enrolled, with exclusion of many frail, institutionalised, immunosuppressed and pneumococcal-vaccine-eligible survivors.
  • Prior pneumococcal vaccination and other vaccines were not standardised; no biomarker was used to select a potentially responsive immune phenotype.
  • Binding IgG rather than functional opsonophagocytic activity was measured, pneumococcal disease itself was not an outcome, and trained immunity was not assessed.
  • Published biological results concentrated on days 10 and 30 despite a 365-day clinical question, so durability and later immune recovery remain uncertain.
  • Extensive exploratory biological and clinical secondary analyses create false-positive risk, while several numerical inconsistencies in the published results require correction.
  • PCV13 is no longer the principal adult pneumococcal product in the current UK programme.

Interpretation & Why It Matters

• Clinical Practice

  PCV13 should not be administered solely to “accelerate immune recovery” or prevent general post-sepsis infection in otherwise non-indicated survivors. The trial found no reduction in the primary outcome, and there was no consistent favourable signal across clinical secondary outcomes.
• What Immunogenicity Means

  The rise in antibody against selected serotypes confirms that many sepsis survivors can mount a vaccine response during early recovery. It does not show that the antibodies were functionally protective, durable, or capable of preventing the broad mixture of infections represented in the clinical endpoint.
• Safety Interpretation

  The higher proportion with serious adverse events and the recurrent-infection signal cannot be dismissed, but neither establishes vaccine-induced harm. Their correct interpretation is a reason for caution and replication, not a reason to conclude that all pneumococcal vaccination after sepsis is dangerous.
• Precision Rather Than Uniform Stimulation

  The coexistence of high- and low-inflammation states, immunosenescence-like lymphocyte patterns and widely variable IgG responses supports a mechanism-based strategy: define the dominant immune defect, select patients likely to respond, and match the intervention and timing to that biology. Precision vaccinology and contemporary sepsis-immunobiology frameworks point in this direction.⁶⁷
• Routine Vaccination Still Applies

  VACIRiSS deliberately excluded many people with standard pneumococcal-vaccine indications. Age- and comorbidity-based vaccination should therefore continue unchanged. In the UK, PCV20 is now used for adults from 65 years and for listed clinical-risk groups; previous sepsis alone is not a listed indication.⁸

Controversies & Other Evidence

• The clinical target was much broader than the established vaccine effect. PCV13 is designed to prevent disease caused by 13 pneumococcal serotypes, yet VACIRiSS excluded current pneumococcal sepsis and did not measure pneumococcal or vaccine-type infection. Its success therefore depended principally on a large heterologous effect against all infections and all-cause death. In CAPiTA, PCV13 efficacy was 45.6% against vaccine-type community-acquired pneumonia and 75.0% against vaccine-type invasive disease, but only 5.1% against all-cause community-acquired pneumonia (95% CI −5.1% to 14.2%).⁴
• The trial tested a very large effect and should not be read as an equivalence study. The power calculation assumed HR 0.36. The observed HR 1.23; 95% CI 0.80 to 1.91 convincingly argues against that transformative benefit, but it does not exclude a smaller reduction or increase in risk. “No benefit demonstrated” is more accurate than “no biological effect” or “proof of no effect”.
• Timing may have disadvantaged conventional vaccine protection. Events accrued from randomisation, whereas peak serotype-specific IgG was assessed at day 30. Early readmissions or antibiotic-treated episodes could occur before a mature antibody response. This is less problematic for the proposed rapid heterologous effect, but it weakens the test of conventional antigen-specific protection.
• The apparent infection signal is difficult to interpret. Reinfection included any infection-related rehospitalisation or primary-care antibiotic course. The adjusted RR intervals at days 30, 180 and 365 excluded 1, while the corresponding risk-difference intervals included 0, and no multiplicity adjustment was applied. The authors reasonably note that post-vaccination systemic symptoms could have been labelled as infection and treated, but coding and prescribing data cannot distinguish this from true infection.
• The antibiotic result is weaker than the abstract wording implies. Time to first primary-care antibiotic treatment gave HR 1.34; 95% CI 0.95 to 1.88. This is compatible with earlier treatment but does not establish it statistically.
• The serious-adverse-event imbalance deserves independent replication. Serious events affected 51.9% versus 35.4% of participants (P=0.02), despite no excess of prespecified local or systemic reactions and no event being judged vaccine-related. Chance, broad event capture and baseline heterogeneity remain plausible, but the finding is not reassuring enough to ignore in future post-sepsis vaccine trials.
• Prespecification was transparent but later than ideal. The protocol framed the primary objective as a hazard-ratio comparison, while the final SAP designated age-stratified IRR and IRD as the principal population summaries and the Cox HR as a secondary analysis. SAP version 1.0 was dated 7 November 2022, after recruitment ended on 22 April 2022 but before database lock; version 1.2 added the Sepsis Survivor Prognosis subgroup on 21 December 2023. The consistency of IRR and HR estimates is reassuring, but a fully finalised pre-recruitment SAP would have offered stronger protection against analytic flexibility.⁵
• Several published numerical inconsistencies require correction. The 365-day infection-related rehospitalisation row prints 19/102 (18.6%) for PCV13, although the primary-component analysis reports 40 PCV13 readmissions and the row’s RR 1.17 and RD +5.0 percentage points are incompatible with 19/102. The day-30 reinfection row prints a placebo denominator of 102, whereas 34 events and 32.7% correspond to 104 and the narrative uses 104. The text also prints placebo CRP as 954 (IQR 21 to 123) mg/L, which is internally impossible, and gives a slightly different all-cause-rehospitalisation HR from Table 2. These appear to be reporting errors rather than evidence that the analyses were wrong, but the primary publication should be corrected.
• The biological findings are hypothesis-generating, not validation of a mechanism. PCV13 increased binding IgG for only 6 of 13 serotypes compared with placebo; functional antibody activity, pneumococcal outcomes and epigenetic trained immunity were not measured. Many cellular and transcriptomic findings were within-group longitudinal comparisons or exploratory correlations, and the control cohort consisted of elective cardiac-surgery participants rather than population healthy volunteers. The protocol planned broader immune-recovery assessment through day 90, whereas the principal published mechanistic comparisons ended at day 30.
• Baseline predictors remain plausible but unproven treatment-selection tools. In older adults, age, sex, T-helper phenotypes and a cytotoxicity-associated signature have been associated with PCV13 responsiveness, while contemporary sepsis frameworks argue for biologically informative subtyping rather than uniform immunomodulation.⁶⁷ VACIRiSS reproduced some associations with age, body mass index and cytotoxicity modules, but its high- versus low-inflammation endotypes did not clearly enrich antibody responders and were not validated against clinical benefit.
• Current vaccine policy answers a different question. The UK programme now uses PCV20 for adults aged 65 years and over and for defined clinical-risk groups.⁸ The guidance predates the final VACIRiSS publication and therefore does not incorporate its results. VACIRiSS does not test PCV20, does not challenge established serotype-specific vaccine efficacy, and does not support withholding guideline-indicated vaccination after sepsis.

Summary

• VACIRiSS randomised 214 recovering UK ICU sepsis survivors to a single 0.5-mL dose of PCV13 or saline placebo, with excellent masking and high treatment adherence.
• PCV13 did not reduce the primary outcome of infection-related rehospitalisation or death: HR 1.23; 95% CI 0.80 to 1.91; P=0.35.
• Broadly defined reinfections were more frequent by adjusted relative-risk analyses, and serious adverse events affected 51.9% versus 35.4%; these signals warrant caution but do not establish vaccine-related harm.
• PCV13 induced variable serotype-specific IgG and measurable cellular and transcriptomic changes, but immune activation did not translate into clinical benefit.
• The trial argues against routine untargeted PCV13 as a post-sepsis immune-recovery treatment and redirects research towards biomarker-selected patients, mechanism-specific interventions and better-defined timing and outcomes.

Overall Takeaway

VACIRiSS is an important first-in-population, mechanistically rich randomised trial, but it is not practice-changing in favour of vaccination: PCV13 given during early recovery from critical illness generated variable immune responses without reducing one-year infection-related rehospitalisation or death. Its principal contribution is to show that post-sepsis immune dysfunction cannot be assumed to respond beneficially to uniform stimulation and that future immunomodulation will require greater biological precision.

Bibliography

• 1.Shankar-Hari M, Saha R, Wilson J, Prescott HC, Harrison DA, Rowan K, Rubenfeld GD, Adhikari NKJ. Rate and risk factors for rehospitalisation in sepsis survivors: systematic review and meta-analysis. Intensive Care Med. 2020;46(4):619-636.
• 2.Prescott HC, Iwashyna TJ, Blackwood B, Calandra T, Chlan LL, Choong K, et al. Understanding and enhancing sepsis survivorship: priorities for research and practice. Am J Respir Crit Care Med. 2019;200(8):972-981.
• 3.van der Poll T, Shankar-Hari M, Wiersinga WJ. The immunology of sepsis. Immunity. 2021;54(11):2450-2464.
• 4.Bonten MJM, Huijts SM, Bolkenbaas M, Webber C, Patterson S, Gault S, et al. Polysaccharide conjugate vaccine against pneumococcal pneumonia in adults. N Engl J Med. 2015;372(12):1114-1125.
• 5.Shankar-Hari M, Harrison DA. Pneumococcal vaccination to accelerate immune recovery in sepsis survivors: randomised placebo-controlled trial. Protocol version 6.2 (1 November 2021) and statistical analysis plan with version history through 1.2 (21 December 2023). Supplementary material to Shankar-Hari et al. Sci Transl Med. 2026;18(862):eaeb4113.
• 6.Ravichandran S, Erra-Diaz F, Karakaslar OE, Marches R, Kenyon-Pesce L, Rossi R, et al. Distinct baseline immune characteristics associated with responses to conjugated and unconjugated pneumococcal polysaccharide vaccines in older adults. Nat Immunol. 2024;25(2):316-329.
• 7.Shankar-Hari M, Calandra T, Soares MP, Bauer M, Wiersinga WJ, Prescott HC, et al. Reframing sepsis immunobiology for translation: towards informative subtyping and targeted immunomodulatory therapies. Lancet Respir Med. 2024;12(4):323-336.
• 8.UK Health Security Agency. Pneumococcal: the Green Book, chapter 25. London: UKHSA; 2026. Updated 5 August 2026.