Review Reports - A Possible Unmet Need: Pneumococcal Vaccination in the Workplaces—A Systematic Review of Invasive Pneumococcal Disease Among Shipyard Workers

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is a well-conducted, clearly reported systematic review and meta-analysis on an important occupational health topic – invasive pneumococcal diseases (IPD). IPD remains a major vaccine-preventable disease. Shipyard workers are exposed to welding fumes and other respiratory hazards, and recent outbreaks (e.g., Finland, France, Norway) have raised concern. The authors synthesize data from six IPD outbreaks in European shipyards (2015–2025), providing pooled incidence estimates, clinical outcomes, serotype distribution, and implications for prevention. The manuscript is relevant, methodologically sound, and offers practical value for occupational health services and public health policymakers. I only have several minor comments:

The authors used the Newcastle‑Ottawa Scale for cohort studies, but outbreaks are not traditional cohorts. This is acknowledged, but consider adding a sentence that no validated tool for outbreak studies exists – which is fine, but transparency helps readers.
Line 62: “these infections in are defined as invasive pneumococcal diseases” Remove “in”: “these infections are defined as…”
Lines 76-77: “did account 12% to 13%” replace with “accounted for 12% to 13%”
Line 349: “whole of 67 cases” “total of 67 cases”
Lines 420, 769: “Forrest plots” Should be “Forest plots”
Figure 5 (Forest plots): The quality is acceptable but resolution could be improved for publication.
Table 5 is labeled twice (one should be Table 6 for Egger’s test). Renumber.

Author Response

We would like to warmly thank you the present reviewer for their very positive appraisal.

In fact, we have edited the paper in order to cope with their requirements, as follows:

(1) The authors used the Newcastle‑Ottawa Scale for cohort studies, but outbreaks are not traditional cohorts. This is acknowledged, but consider adding a sentence that no validated tool for outbreak studies exists – which is fine, but transparency helps readers.

REPLY: in fact, no instrument are actually internationally acknowledged for the quality appraisal of outbreak studies. However, as the ORION framework has been designed for nosocomial outbreaks and its design appears quite appropriate for our aims, an adapted version has been designed and has been included in the present review.

See methods:

"In the absence of a dedicated framework for outbreak investigations, the cohort version of the Newcastle–Ottawa Scale (NOS) was applied to all included studies, complemented by an adapted version of the “Outbreak Reports and Intervention studies Of Nosocomial infection (ORION)” statement to address outbreak-specific reporting characteristics [48,49]".

(2) Comments 2 to 7: thank you, we've edited the main text and substantially solved the whole of the typos.

Thank you again for the time you spent in the appraisal of our study.

on the behalf of all Authors,

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript addresses an impactful systematic review in an under-characterized occupational risk setting. The topic is relevant for occupational health, infectious diseases, and public health policy. The manuscript is generally comprehensive and methodologically ambitious, but it currently overreaches in parts, has internal inconsistencies, and would benefit from sharper framing and tighter epidemiological rigor.

Major points.

There is a clear mismatch between the PECO-framed research question and the included evidence. While the question implies a comparative risk assessment, the analysis relies on outbreak reports without control groups, which cannot support causal inference or relative risk estimation. As a result, claims of increased occupational risk are not directly supported, and comparisons with general population incidence are ecological and indirect. The objective should be reframed as a descriptive synthesis of IPD outbreaks in shipyard settings, and causal language regarding occupational exposure should be removed or substantially softened unless supported by appropriate comparative data..
The analysis is based on a limited dataset (6 outbreaks, 8 studies) characterized by substantial heterogeneity (I² ~90%) and likely publication and reporting biases. Despite this, pooled estimates are presented with a level of precision that exceeds the strength of the underlying evidence, and subgroup analyses are interpreted more confidently than warranted given their low statistical power. These estimates should be framed as exploratory and hypothesis-generating, with cautious interpretation of subgroup differences. The authors should also consider whether a meta-analysis meaningfully adds value over a structured descriptive synthesis.
Although variability in case definitions is acknowledged, its impact is insufficiently addressed. Substantial differences, such as restriction to serotype 4 in Norway and broader inclusion criteria in later Finnish studies (e.g., NAAT and radiological diagnoses), as well as inconsistent handling of non-sterile site isolates, undermine comparability across outbreaks. This likely introduces systematic under- or over-ascertainment, compromising incidence estimates. The manuscript would benefit from a dedicated sensitivity analysis to assess this bias and, where feasible, stratification by stricter versus expanded case definitions.
The incidence is calculated using the total shipyard workforce, implicitly assuming homogeneous exposure. This overlooks substantial workforce heterogeneity (e.g., welders vs. administrative staff) and likely results in exposure misclassification, biasing estimates downward. The authors should clearly state that these incidence figures are crude and likely diluted, and avoid presenting them as accurate measures of risk among truly exposed workers.
The characterization of outcomes as “relatively favorable” is plausible but overstated. Comparisons with general IPD populations are implicit rather than formally analyzed, and the CFR (0.8%) is based on a single death, rendering the estimate unstable. Moreover, hospitalization and ICU admission rates remain substantial. The conclusion should be tempered to state that outcomes appear lower than is typically reported in older populations, likely reflecting workforce demographics, and to avoid making strong claims based on limited data.
The reported predominance of serotype 4 (67.2%) is noteworthy but likely influenced by methodological and epidemiological constraints. In particular, the Norwegian outbreak was restricted to serotype 4 by design, and the observed dominance of ST801 suggests a clonal (founder) effect rather than a broadly generalizable pattern. The manuscript should clearly distinguish outbreak-specific clonal expansion from wider serotype distribution in occupational settings, and temper downstream implications for vaccine policy accordingly.
The manuscript makes policy-level vaccination recommendations that are not fully supported by the data. There is no direct evidence of vaccine effectiveness, strategies varied across outbreaks, and uptake was not systematically analyzed. Vaccination should be framed as a plausible intervention, avoiding strong or unqualified recommendations.
The manuscript applies funnel plots, Egger’s test, and the LFK index despite a very small number of included studies (n ≈ 6–8), where these methods lack reliability and statistical power. This limitation should be explicitly acknowledged by stating that assessments of publication bias are underpowered and must be interpreted with caution. Simplifying this section would improve methodological rigor and clarity.

Minor points

The figures are largely descriptive and add limited analytical value. Consider enhancing them with a timeline of outbreaks and, where feasible, stratification by exposure category.
The methods need clarification: provide a brief justification for using the Freeman–Tukey transformation rather than the logit approach. Specify whether zero-event corrections were applied. Clarify how overlapping reports from the same outbreaks were identified and handled (data merging strategy).

Comments on the Quality of English Language

The text is occasionally repetitive and would benefit from careful editing for concision. Several sentences are overly long and difficult to follow, and minor grammatical issues are present throughout.

Author Response

Estimated Reviewer,

we would like to thank you for your comments; we've edited our paper following your recommendations, and we're confident that the interplay between your suggestions and our efforts did eventually improve the overall quality of the paper. More precisely:

(1) There is a clear mismatch between the PECO-framed research question and the included evidence. While the question implies a comparative risk assessment, the analysis relies on outbreak reports without control groups, which cannot support causal inference or relative risk estimation. As a result, claims of increased occupational risk are not directly supported, and comparisons with general population incidence are ecological and indirect. The objective should be reframed as a descriptive synthesis of IPD outbreaks in shipyard settings, and causal language regarding occupational exposure should be removed or substantially softened unless supported by appropriate comparative data..

REPLY: the PECO research question has been reframed as follows: A Patient/Population/Problem; Exposure/Context; Outcome strategy [43,44] was designed given the descriptive nature of outbreak reports and the absence of a consistent comparator. In fact, the review was designed in order to systematically characterize reported outbreaks of pneumococcal disease in shipyard settings, describing their epidemiological features, clinical burden, and associated occupational and environmental conditions, with particular attention to exposure to metal and welding fumes.

(2) The analysis is based on a limited dataset (6 outbreaks, 8 studies) characterized by substantial heterogeneity (I² ~90%) and likely publication and reporting biases. Despite this, pooled estimates are presented with a level of precision that exceeds the strength of the underlying evidence, and subgroup analyses are interpreted more confidently than warranted given their low statistical power. These estimates should be framed as exploratory and hypothesis-generating, with cautious interpretation of subgroup differences. The authors should also consider whether a meta-analysis meaningfully adds value over a structured descriptive synthesis.

REPLY: we have edited the main test in order to stress that the present study may be considered as an exploratory and hypothesis-generating study, with cautious interpretation of all data.

"A second limitation is the small number of included outbreaks and events. Accordingly, the meta-analysis was undertaken as an exploratory synthesis of patterns across outbreaks. The substantial heterogeneity observed limits the interpretability of pooled estimates, which should be viewed as descriptive summaries rather than precise effect measures, reflecting greater between-outbreak variability than within-study precision. Subgroup analyses were similarly underpowered. Therefore, the findings should be regarded as hypothesis-generating".

For the same reason, the extensive analysis of the risk of bias and publication bias has been moved to Appendix B, and the section in the results simplified as follows: "Formal assessments of publication bias were performed but are inherently underpowered and unreliable with the small number of included studies. Visual and statistical diagnostics showed inconsistent patterns across outcomes and did not allow robust conclusions. These analyses are therefore reported as Appendix B (Appendix Figures B1, B2, and B3, and Appendix Table B1)."

(3) Although variability in case definitions is acknowledged, its impact is insufficiently addressed. Substantial differences, such as restriction to serotype 4 in Norway and broader inclusion criteria in later Finnish studies (e.g., NAAT and radiological diagnoses), as well as inconsistent handling of non-sterile site isolates, undermine comparability across outbreaks. This likely introduces systematic under- or over-ascertainment, compromising incidence estimates. The manuscript would benefit from a dedicated sensitivity analysis to assess this bias and, where feasible, stratification by stricter versus expanded case definitions.

REPLY: we substantially agree with your comments. Therefore we've included a subanalysis of laboratory confirmed IPD (see Figure 3) and reframed limits section as follows:

A second limitation is the small number of included outbreaks and events. Accordingly, the meta-analysis was undertaken as an exploratory synthesis of patterns across outbreaks. The substantial heterogeneity observed limits the interpretability of pooled estimates, which should be viewed as descriptive summaries rather than precise effect measures, reflecting greater between-outbreak variability than within-study precision. Subgroup analyses were similarly underpowered. Therefore, the findings should be regarded as hypothesis-generating.

Third, although the characteristics of invasive pneumococcal disease (IPD) are well established by public health authorities [10,17,18], the analysis and comparison of case definitions across the included studies indicate that reporting heterogeneity may have affected our pooled estimates. Case definitions differed in several key aspects. First, the Norwegian outbreak definition focused exclusively on serotype 4 cases [73], and therefore non–serotype 4 infections may have been missed, potentially leading to under-ascertainment. Second, while most studies restricted case identification to shipyard workers, the report by Cassir et al. 2020 [40,69] also included crew members. Although this broader inclusion supports a role for environmental factors in outbreak propagation, it may have diluted the contribution of occupational exposures, such as welding and metal fumes. Third, even within the same setting (i.e., the Turku shipyard), consistent case definitions were not applied across investigations [70–72]. More recent reports, such as Manca et al. 2025 [71] incorporated nucleic-acid amplification tests and radiological criteria for lobar pneumonia, which were not used in earlier investigations. This evolution in diagnostic criteria may have resulted in differential ascertainment, with potential over-ascertainment in more recent studies or, conversely, under-ascertainment in earlier ones. Overall, these discrepancies likely resulted in both under- and over-ascertainment of cases, thereby limiting the comparability of incidence and attack-rate estimates across outbreaks. To partially address this issue, we conducted a sensitivity analysis restricted to laboratory-confirmed IPD cases, which yielded lower incidence estimates. However, this approach increases specificity at the expense of sensitivity and does not fully account for differences in diagnostic intensity across studies. Therefore, heterogeneity in case definitions represents a major source of bias, and our findings should be interpreted with caution.

Please note that a sensitivity analysis has been implemented also for attack rates and incidence rates of laboratory confirmed cases, and not only for previous estimates.

(4) The incidence is calculated using the total shipyard workforce, implicitly assuming homogeneous exposure. This overlooks substantial workforce heterogeneity (e.g., welders vs. administrative staff) and likely results in exposure misclassification, biasing estimates downward. The authors should clearly state that these incidence figures are crude and likely diluted, and avoid presenting them as accurate measures of risk among truly exposed workers.

REPLY: we agreed with your comments and edited the paper as follows (limits section):

...

A further limitation relates to the use of the total shipyard workforce as the denominator for incidence calculations, which implicitly assumes homogeneous exposure across workers. On the contrary, shipyard workforces are highly heterogeneous and dynamic, with substantial short-term variability in workforce size and composition [37,39,41,66,69,70,72–74]. Moreover, different professional groups do not share the same exposure profiles [37,41,69,113]: for example, workers directly involved in welding or operating within shipyard facilities or inside vessels may have substantially different exposure compared with administrative staff or support personnel, who may share only limited common environments (e.g., canteens) or none. Therefore, the use of aggregate workforce figures likely introduces exposure misclassification and may dilute incidence estimates, leading to underestimation of risk among truly exposed workers. Notably, several included studies reported only approximate or coarse estimates of workforce size, further limiting the precision of the denominators. Therefore, the reported incidence figures should be interpreted as crude and indicative measures rather than accurate estimates of risk in specific occupational subgroups.

(5) The characterization of outcomes as “relatively favorable” is plausible but overstated. Comparisons with general IPD populations are implicit rather than formally analyzed, and the CFR (0.8%) is based on a single death, rendering the estimate unstable. Moreover, hospitalization and ICU admission rates remain substantial. The conclusion should be tempered to state that outcomes appear lower than is typically reported in older populations, likely reflecting workforce demographics, and to avoid making strong claims based on limited data.

REPLY: results have been modified ("Pneumococcal infections and IPD are major causes of communicable disease morbidity [10,14,17,18], with substantial associated social and economic costs. According to the 2025 ECDC report on IPD [10], among 7,000 cases with known outcome (39.5%) documented in the EU/EAA countries during 2022, 12.8% died, with higher case fatality ratios observed in older age groups (17.1% among those aged ≥65 years and 10.9% among those aged 45–64 years). In the present analysis, the proportion of cases requiring ICU admission was 13.0%, and the CFR was 0.8%, corresponding to a single death. While these figures appear lower than those reported in general adult populations, particularly older age groups, such comparisons should be interpreted cautiously, as they were not formally assessed and likely reflect the younger age distribution of the workforce (median age 39–48 years across studies). Moreover, the CFR estimate is based on a very small number of events and is therefore inherently unstable. Importantly, hospitalization rates remained substantial, with incidence estimates 30 to 100 times higher than those reported in the corresponding working-age groups (25–44 and 45–64 years) in the general population of the same countries during the assessed timeframes [10,17,18].") and also conclusions have been edited, as follows: Despite this, clinical outcomes were associated with low rates of intensive care unit admission and a case fatality ratio below 1%, findings that could be tentatively explained by the relatively young age of the affected workforce than by specific pathogen-related factors.

(6) The reported predominance of serotype 4 (67.2%) is noteworthy but likely influenced by methodological and epidemiological constraints. In particular, the Norwegian outbreak was restricted to serotype 4 by design, and the observed dominance of ST801 suggests a clonal (founder) effect rather than a broadly generalizable pattern. The manuscript should clearly distinguish outbreak-specific clonal expansion from wider serotype distribution in occupational settings, and temper downstream implications for vaccine policy accordingly.

REPLY: thank you, we stressed the methodological issues in the limits section ("First, the Norwegian outbreak definition focused exclusively on serotype 4 cases [73], and therefore non–serotype 4 infections may have been missed, potentially leading to under-ascertainment")

(7) The manuscript makes policy-level vaccination recommendations that are not fully supported by the data. There is no direct evidence of vaccine effectiveness, strategies varied across outbreaks, and uptake was not systematically analyzed. Vaccination should be framed as a plausible intervention, avoiding strong or unqualified recommendations.

REPLY: again we agree and the corresponding section has been reframed as follows:

"The prevention of pneumococcal infection and invasive pneumococcal disease (IPD) among shipyard workers has been identified as a relevant concern for occupational health practice [20–23,25,32,33,35,38,63]. Several national bodies, including the British Department of Health, the German Ständige Impfkommission (STIKO), and the Austrian National Vaccination Plan, recommend pneumococcal vaccination for workers exposed to welding or metal fumes [77–79]. However, recommendations remain heterogeneous across countries [80,81], and the present study does not directly assess vaccine effectiveness, optimal strategies, or uptake in outbreak settings. Therefore, any considerations regarding vaccination should be interpreted as policy-relevant implications rather than direct evidence-based conclusions from this analysis. For example, pneumococcal vaccines are not currently included by French official recommendations for workplace immunizations. Similarly, Italian legal framework does not formally recommend the workplace prophylaxis of IPD by means of pneumococcal vaccines, but the Italian Consolidated Act on Health and Safety at Work (Legislative Decree No. 81/2008) stresses the role of Occupational Physicians in the occupational risk assessment [82,83] , as the medical professional responsible for the ultimate policy decision to recommend or not the implementation of a given vaccination in a specific occupational setting following a context-specific risk–benefit assessment.

Results from the present systematic review also share some hints on the more appropriate formulations to be recommended in daily practice. Several pneumococcal vaccines are commercially available, including PPSV and conjugate (PCV) formulations [3,4,14,84–87] Although both PPSV and PCV vaccines have been used during outbreak management, the available data are heterogeneous and do not allow formal comparison of their effectiveness in reducing outbreak size or severity. Distinctive properties and effectiveness profiles of PPSV and PCV have been discussed elsewhere [84,88,89]. Briefly, the cheaper PPSV have an adult effectiveness estimated in around 24% (95%CI 5 to 40%) [86,90], having been considered as an effective option for healthy adults who have not completed a previous vaccination schedule with PCV [78,79,91]. However, as PPSV are unable to elicit an effective mucosal immunity [86,90,92], their implementation does not affect the circulation of the pathogen, nor does it guarantee the actual protection of unvaccinated workers. Conjugate vaccines may therefore offer theoretical advantages, including the induction of mucosal immunity and potential reduction of transmission [93–95], particularly when dealing with crowded workplaces, shared households and accommodations, possibly reducing the risk of outbreaks in the specific settings of construction yards and naval shipyards [29,40,73,96]. In other words, while a vaccination strategy based on PPSV can be considered as a rational approach for fulfilling the primary aim of the Occupational Health practice for safeguarding and promoting the health of a certain worker [97], prioritizing the use of PCV could contribute to the containment of IPD burden of disease, also benefiting individuals who are more difficult to reach through catch-up campaigns. Still, these potential features were not directly assessed in the included studies [29,39,40,69,70] and also vaccination uptake was not systematically reported across studies, impairing the formal assessment of its impact on outbreak dynamics. For instance, although vaccination campaigns were rapidly and effectively organized in all Finnish outbreaks, vaccination uptake declined markedly, from over 4,000 workers in 2019 [70] and over 3,000 in 2023 [72] to 1500 in 2025 [71], corresponding to approximately 57.1%, 33.3%, and 16.7% of eligible workers, respectively; however, the available data did not allow the authors to assess whether these differences had any measurable impact on outbreak duration or severity.

Another issue to be considered in terms of policy is represented by vaccine formulations. PPSV23 has historically covered approximately 60–75% of adult IPD cases (see Appendix Table A5) [14,90], while earlier conjugate vaccines (e.g., PCV7) had more limited coverage [14,95], subsequently expanded in newer formulations up to PCV20 [95,98,99]. The recent introduction of PCV21 further broadens vaccine coverage by including the 21 most prevalent pneumococcal serotypes that cause IPD in adults and therefore allowing the potential replacement of PPSV23 [98], but excludes several serotypes included in PCV20

Selection of vaccine strategies should therefore consider the serotype distribution observed in outbreaks. In the present series, the predominance of serotype 4 (67.2%) highlights the relevance of vaccines covering this serotype (PPSV23, PCV13, PCV15, PCV20), whereas PCV21 would have provided more limited coverage despite broader overall effectiveness in older populations [95,98] (see Appendix Table A5). While PPSV23 covered all detected serotypes and PCV20 omitted only one case (serotype 9N), PCV13 and PCV15 provided comparable coverage. Notably, the emergence of serotypes such as 8, 9N, and 9V in more recent outbreaks (Figure 5) further underscores the need for context-specific vaccine selection. In the end, it should be stressed that these considerations are based on descriptive serotype distributions and should not be interpreted as evidence of differential vaccine effectiveness in outbreak prevention.

Overall, vaccination represents a plausible and policy-relevant preventive strategy in high-risk occupational settings; however, the present findings do not allow direct assessment of the comparative effectiveness of specific vaccine formulations or strategies. Further studies are needed to evaluate optimal vaccination approaches in these contexts".

(8) The manuscript applies funnel plots, Egger’s test, and the LFK index despite a very small number of included studies (n ≈ 6–8), where these methods lack reliability and statistical power. This limitation should be explicitly acknowledged by stating that assessments of publication bias are underpowered and must be interpreted with caution. Simplifying this section would improve methodological rigor and clarity.

REPLY: see above point 2: the extensive analysis of the risk of bias and publication bias has been moved to Appendix B, and the section in the results simplified as follows: "Formal assessments of publication bias were performed but are inherently underpowered and unreliable with the small number of included studies. Visual and statistical diagnostics showed inconsistent patterns across outcomes and did not allow robust conclusions. These analyses are therefore reported as Appendix B (Appendix Figures B1, B2, and B3, and Appendix Table B1)."

MINOR COMMENTS

The figures are largely descriptive and add limited analytical value. Consider enhancing them with a timeline of outbreaks and, where feasible, stratification by exposure category.
REPLY: the timeline is included in figure 2; we have also included Figure 3 and Figure 4 for providing a more appropriate visual assistance to the appraisal of incidence and attack rates.
The methods need clarification: provide a brief justification for using the Freeman–Tukey transformation rather than the logit approach. Specify whether zero-event corrections were applied. Clarify how overlapping reports from the same outbreaks were identified and handled (data merging strategy).
we've edited the main text as follows:
The meta-analysis was performed as an exploratory analysis to summarize patterns across outbreaks. Between-study variance (τ²) was estimated by maximum likelihood methods. Prior to pooling, proportions were transformed using the Freeman–Tukey double arcsine transformation, which is generally preferred over logit transformation when dealing with small and heterogeneous samples because it stabilizes variances and allows inclusion of studies with proportions close to 0 or 1 without requiring continuity corrections. This approach was particularly relevant given the small number of events and the presence of extreme proportions in some outbreaks. In contrast, the logit transformation may be less stable in the presence of sparse data and typically requires continuity corrections when zero events are observed [50]. No zero-event corrections were required, as only outbreak-based datasets were included, inherently precluding zero-event studies for the outcomes considered.

Thank you for your comment: as you can see, the paper has been extensively edited.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have substantially improved the manuscript in response to my previous comments. Key issues have been addressed, including reframing the study as descriptive, tempering interpretations, and clarifying limitations. However, a few points still need clarification bebore be accepted

1-The discussion of serotype distribution, particularly the predominance of serotype 4, should more clearly distinguish between outbreak-specific dynamics (e.g., clonal expansion, potential founder effects) and broader epidemiological patterns. This would help avoid overgeneralization and better contextualize implications for vaccine strategies.

2-While the exploratory nature of the meta-analysis is now acknowledged, a brief justification for retaining a quantitative synthesis rather than relying solely on a structured descriptive approach would strengthen the methodological rationale.