Economic and productivity outcomes of workplace health programmes incorporating blood-based laboratory screening: a systematic review

Abstract

Background

Workplace health programmes increasingly include blood-based laboratory screening, promoted to lower costs and improve productivity, yet the economic case is contested and previous reviews have not isolated programmes with a blood-based screening component. We systematically reviewed the economic and productivity outcomes of such programmes.

Methods

Following PRISMA 2020 and a prospectively registered protocol, we searched PubMed, Scopus, Web of Science, EconLit, the Cochrane Library, and grey literature (WHO, ILO) from January 2000 to May 2026. Eligible studies evaluated workplace programmes incorporating blood-based laboratory or biomarker screening in working adults and reported at least one economic or productivity outcome. Selection, extraction, and quality appraisal, using design-appropriate tools (RoB 2, Newcastle-Ottawa Scale, NIH before–after tool, CHEC list), were performed in duplicate with adjudication. Because outcome metrics were too heterogeneous to pool, findings were synthesised narratively by direction of effect across six domains, with sensitivity and subgroup analyses.

Results

Fifty-six studies were included, mostly US observational or single-arm designs; six were randomised trials, none at low risk of bias. Healthcare-cost outcomes (29 studies) were predominantly favourable (21 studies) and remained so among higher-quality studies. Return on investment (20 studies) was uniformly favourable, with no attenuation in the direction of effect across study quality. Absenteeism (23 studies) was inconsistent, and presenteeism (7) and staff turnover (1) were rarely measured, reflecting reliance on administrative rather than validated measures. Patterns were robust in sensitivity analyses.

Conclusions

Workplace programmes incorporating blood-based screening were associated with favourable healthcare-cost and return-on-investment results in most studies, but the return-on-investment signal showed no attenuation with study quality and is hard to separate from reporting and selection bias; absenteeism was inconsistent and productivity rarely assessed. Favourable economics appeared almost exclusively where screening was coupled to a structured, coordinator- or physician-led pathway, not offered alone—a contrast to read cautiously, since screening-only programmes were seldom evaluated economically. The evidence does not support blood-based screening as a stand-alone cost-saving measure and suggests, as a hypothesis, that its value lies less in testing than in support that turns detection into sustained behaviour or treatment change. Rigorous, controlled, non-US studies with validated productivity measures are needed.

Systematic review registration

Open Science Framework, https://doi.org/10.17605/OSF.IO/M4W8C

Clinical trial number

Not applicable.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1186/s12913-026-15428-8.

Background

Poor employee health imposes substantial and recurring costs on employers, accruing through medical expenditure, sickness absence (absenteeism), and reduced productivity while at work (presenteeism) [1, 2]. To contain these costs and improve workforce health, workplace health-promotion and wellness programmes have become widespread, particularly among large employers in high-income economies [3, 4]. Such programmes are commonly framed as an investment expected to yield a financial return, and their economic evaluation has accordingly centred on metrics such as healthcare-cost savings and return on investment (ROI) [3, 5].

A large proportion of contemporary workplace programmes incorporate biometric and blood-based laboratory screening—for example, measurement of blood lipids, fasting glucose or glycated haemoglobin, and related biomarkers—typically embedded within a health-risk assessment [4]. The underlying logic is that measuring modifiable biological risk factors identifies employees who may benefit from intervention, enabling targeted risk reduction and, ultimately, downstream savings. Blood-based measurement is distinctive in what it can detect: prevalent but clinically silent and modifiable conditions—dysglycaemia and undiagnosed type 2 diabetes, dyslipidaemia, chronic kidney disease, anaemia and iron deficiency, and thyroid dysfunction—that are frequently asymptomatic and cannot be captured by self-report questionnaires, anthropometry, or blood-pressure measurement alone. This is the principal clinical rationale for laboratory testing in the workplace, and it distinguishes such programmes from those built on health-risk-appraisal questionnaires.

Blood-based programmes are also economically distinct in three respects. First, their costs are dominated by recurrent laboratory and phlebotomy inputs rather than the largely one-off costs of educational or organisational change [4, 6]. Second, they yield clinically actionable biomarker data rather than self-reported risk information alone [4]. Third, any benefit depends on the downstream intervention pathway—physician-led, coordinator-led, or self-directed—which previous investigators have identified as decisive for both programme effectiveness and the return that is measured [7, 8].

The economic case for workplace programmes is contested. Early and largely observational evidence was optimistic: an influential meta-analysis concluded that medical and absenteeism costs fell substantially relative to programme spending, reporting savings of approximately US$3.27 in medical costs and US$2.73 in absenteeism costs per dollar invested [9], and narrative reviews likewise reported favourable clinical and cost outcomes [3, 5]. Subsequent randomised controlled trials have tempered these conclusions. In the BJ’s Wholesale Club trial, a multi-site workplace wellness programme changed some self-reported health behaviours but produced no significant effect on clinical measures, healthcare spending, or employment outcomes at 18 months—a pattern that persisted at three years [10, 11]. The Illinois Workplace Wellness Study reached similar conclusions and demonstrated that much of the apparent saving reported in observational studies reflects the selection of healthier [12], lower-cost employees into participation rather than a programme effect [13]. A large independent evaluation drew comparably cautious conclusions [4]. Methodological quality appears central to this discrepancy: reported ROI has been shown to vary inversely with the rigour of the underlying study design, with the most favourable returns concentrated among methodologically weaker studies [7], and the measurement of returns from integrated programmes is inconsistent [8].

Despite this substantial literature, existing syntheses have evaluated workplace prevention in aggregate, without isolating programmes that include a blood-based laboratory screening component. A recent systematic review of the return on investment of workplace-based prevention interventions—among the broadest in scope, encompassing 138 interventions across 141 reports and spanning the full range of primary, secondary, and tertiary prevention through human, organisational, and technical approaches (from coaching, training, and physical-activity programmes to ergonomic and organisational measures)—searched the literature to 2021 and pooled these interventions together [6]. Although it included screening among its intervention categories, it did not classify interventions by the presence of a blood-based or laboratory component, nor analyse economic outcomes for that subgroup specifically. This is a consequential gap, because blood-based screening differs from other workplace prevention in precisely the dimensions that determine economic value: its cost structure is dominated by recurrent laboratory testing rather than one-off educational or organisational inputs; the biomarker data it produces are clinically actionable in a way that generic risk messaging is not; and its capacity to generate savings is contingent on the intervention pathway that follows detection. Whether workplace programmes incorporating blood-based laboratory or biomarker screening deliver economic or productivity benefits—and the programmatic conditions under which any such benefits arise—has therefore not been established.

Clarifying this question matters for the employers and occupational-health services that continue to invest in workplace biomarker screening, and for the policymakers who shape incentives for workplace prevention. We therefore conducted a systematic review of the economic and productivity outcomes of workplace health programmes incorporating a blood-based laboratory or biomarker screening component. In working adults, we synthesised the reported effects on healthcare costs, ROI, absenteeism, presenteeism, cost-effectiveness, and staff turnover, relative to no screening, usual care, or a pre-intervention baseline; and, because the existing literature suggests that both programme design and study conduct strongly shape apparent returns, we examined how outcomes varied by the type of screening, the nature of the post-screening intervention, and the method used to measure productivity.

Methods

Protocol and registration

This systematic review was conducted according to a protocol registered prospectively on the Open Science Framework (https://doi.org/10.17605/OSF.IO/M4W8C). It is reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement [14] and was planned following PRISMA-P [15]; the narrative synthesis is reported using the Synthesis Without Meta-analysis (SWiM) guideline [16]; the completed PRISMA 2020 and SWiM checklists are provided as a related file. Deviations from the registered protocol are reported explicitly below.

Eligibility criteria

Studies were eligible if they evaluated, in working adults aged 18–65 in any sector or country (P), a workplace health programme incorporating blood-based laboratory testing or biomarker measurement—venous or capillary blood, point-of-care testing, or dried blood spots (I)—compared with no screening, usual occupational care, or a pre-intervention baseline (C), and reported at least one economic or productivity outcome (O): return on investment (ROI), cost savings, healthcare expenditure, absenteeism, presenteeism, staff turnover, or cost-effectiveness. Eligible designs were randomised controlled trials (RCTs), quasi-experimental and controlled before–after studies, single-arm before–after studies, prospective and retrospective cohort studies, and cross-sectional studies or economic models incorporating an economic evaluation. Studies were eligible only if reported in English and published between 2000 and 2026.

Records were excluded for the following pre-specified reasons: questionnaire- or health-risk-appraisal-only programmes without blood-based measurement (E1); workplace drug testing (E2); infectious-disease screening only (E3); genetic or genomic testing (E4); anthropometric measures only, without blood (E5); no economic or productivity outcome (E6); reviews, protocols, or editorials (E7); non-working populations (E8); urinalysis only (E9); and programmes not based in the workplace (E10).

Information sources and search strategy

Five bibliographic databases were searched from 1 January 2000 to 25 May 2026 (PubMed on 24 May 2026): PubMed, Scopus, Web of Science, EconLit, and the Cochrane Library. Grey literature was sought from the World Health Organization and the International Labour Organization. The search combined controlled vocabulary and free-text terms for three concepts—the workplace setting, health screening and workplace wellness programmes, and economic or productivity outcomes—adapted to each database. To maximise sensitivity, the search was not restricted to blood-based terms; the requirement for a blood-based laboratory or biomarker component was instead applied at the eligibility-screening stage (exclusion code E1). Full search strategies for every source, with dates and result counts, are provided in Additional file 1. Reference lists of included studies and relevant reviews were screened to identify additional records.

Study selection

Records were collated and duplicate records removed, after which titles and abstracts were screened against the eligibility criteria, followed by full-text assessment of all potentially eligible reports. Every record was screened in duplicate by two human reviewers (DS, SM), who worked separately and recorded their own include or exclude decision for each record against the eligibility criteria. Large language models were used only as a preparatory aid: for each record they returned a provisional suggestion together with the supporting text, which each reviewer checked against the original record or full text before recording a decision (see Use of large language models). No record was included or excluded on the basis of a model output; every selection decision reported in this review was made by a human reviewer. Of the 190 reports sought for retrieval, 56 were not assessed at full text because they were published in a language other than English or fell outside the 2000–2026 publication window; the remaining 134 reports were assessed against the eligibility criteria. Disagreements at the title/abstract and full-text stages were resolved by a third reviewer (SA), who consulted the original record or full text, and reasons for exclusion at full text were recorded against the pre-specified codes (E1–E10). Reviewer concordance before adjudication is reported in Results, and every adjudicated item for which a written rationale was recorded is listed in the adjudication log (Additional file 1, S6).

Data extraction

A piloted, structured extraction form (Additional file 1) was used to capture study identifiers, country and sector, organisation type and size, study design, sample size, follow-up duration, funding source and conflicts of interest, comparator, the screening programme (panel composition, biomarkers, testing method), the post-screening intervention, all reported economic and productivity outcomes with their direction and magnitude, and the data required for the four synthesis dimensions (below). Extraction was performed in duplicate by the two reviewers with large-language-model support and reviewer verification; discrepancies were adjudicated by the third reviewer against the full text. Where the same study was reported in more than one record, data were consolidated to the study level; reports of the same studies were linked to the study level, so that 58 included reports corresponded to 56 unique studies.

Methodological quality assessment

Risk of bias and methodological quality were appraised using a two-layer, design-appropriate approach. Layer 1 (internal validity) applied: the Cochrane Risk-of-Bias tool version 2 (RoB 2) to RCTs [17]; the Newcastle–Ottawa Scale (NOS), cohort version, to cohort and quasi-experimental (including controlled before–after) studies [18]; a modified NOS for cross-sectional studies (maximum 10 stars) [19]; and the National Institutes of Health (NHLBI) Quality Assessment Tool for Before–After (Pre–Post) Studies With No Control Group for single-arm before–after studies [20]. Layer 2 applied the Consensus Health Economic Criteria (CHEC) list to studies presenting a full economic evaluation—defined as an analysis comparing costs and consequences with a defined perspective and comparator (cost-effectiveness, cost–benefit, cost–utility, or modelling studies) [21]; CHEC therefore overlapped Layer 1 and was applied to 13 studies.

Two protocol refinements are disclosed. First, single-arm before–after studies were appraised with the NIH Pre–Post tool rather than the NOS, because the NOS comparability and selection-of-the-non-exposed-cohort domains are unattainable for studies without a concurrent comparison group and would have artificially deflated their scores; the NOS-cohort tool was reserved for studies with a concurrent comparator. Accordingly, three studies initially labelled before–after that were found at full text to include a concurrent non-participant comparison group were re-classified as quasi-experimental (controlled before–after) and appraised with the NOS-cohort tool. Second, the NOS was used in preference to ROBINS-I given the anticipated predominance of single-arm and descriptive observational designs [22]. For all instruments, summary scores were computed programmatically from the item-level ratings rather than taken from any holistic judgement. For the narrative synthesis, NOS scores were banded as good (≥7), fair (5–6), or poor (<5); RoB 2 as low, some concerns, or high; the NIH tool as good, fair, or poor; and CHEC reported as a total out of 19. For the harvest plot, study quality was additionally encoded on a common 0–1 scale derived from the primary (Layer-1) appraisal: ordinal instruments (NOS cohort and cross-sectional) as the total divided by the maximum attainable score, and categorical judgements mapped to fixed values (some concerns or fair = 0.60; high risk or poor = 0.25; no study was rated low risk or good). For modelling studies, which received only the CHEC appraisal, the normalised score was the CHEC total divided by 19. This scale served the graphical encoding only and entered no analysis.

Data synthesis

Substantial clinical, methodological, and economic heterogeneity was anticipated, and a meta-analysis was pre-specified to be undertaken only if at least three studies reported a comparable outcome using a comparable effect metric. Because the reported metrics, analytic perspectives, and time horizons were not comparable even within a single outcome domain, this threshold was not met; the synthesis is therefore narrative, and the GRADE approach [23], which had been pre-specified for rating certainty contingent on a meta-analysis, was not applied. GRADE certainty ratings presuppose pooled effect estimates with quantifiable precision; because the synthesis rests on the direction of effect across heterogeneous, largely unpooled and frequently single-arm studies, several GRADE domains—notably imprecision and magnitude of effect—could not be assessed meaningfully, and a formal certainty rating would have implied a precision the evidence does not support. No GRADE certainty ratings are therefore reported anywhere in this review, and study quality is described throughout using the design-appropriate instruments named above (RoB 2, Newcastle–Ottawa Scale, the NIH before–after tool, and CHEC).

Findings were synthesised using SWiM and structured by four pre-specified dimensions: (1) screening type (comprehensive panel, targeted biomarker, or check-up with laboratory tests); (2) post-screening intervention, classified as physician-led (clinician-delivered assessment, treatment, or referral), coordinator-led (a non-physician coordinator, nurse, or case manager actively guiding employees through follow-up), self-management (results returned with self-directed materials or digital tools but no active personal support), or screening only (results returned to the employee or employer with no structured follow-up); (3) type of economic outcome; and (4) method of productivity measurement (validated instrument, administrative records, or modelled). The direction of effect for each of six outcome domains—healthcare costs, ROI, absenteeism, presenteeism, cost-effectiveness, and staff turnover—was classified as favourable, no effect, unfavourable, mixed, or not reported. Results were displayed using an effect-direction plot [24] and a harvest plot [25], in which the direction, the methodological quality, and the sample size of each study were encoded.

Two sensitivity analyses were performed: excluding modelling studies, and excluding low-quality studies. Pre-specified subgroup analyses examined screening type, post-screening intervention, and productivity-measurement method.

Use of large language models

Large language models developed by Google and Anthropic were used to assist screening, data extraction, and quality scoring. Their role was preparatory and advisory only: for each record the model returned a provisional suggestion together with the supporting text, which a human reviewer checked against the original record, full text, or source table before any decision was recorded. Both reviewers performed this verification separately, and the senior author (SA) adjudicated every discrepancy against the source. All selection, extraction, and appraisal decisions reported in this review were therefore made by a human author, and no result was accepted from a model without source verification. Every item for which a written adjudication rationale was recorded is listed in the adjudication log in Additional file 1 (S6), so that the adjudicated decisions can be audited against the source. The large language models did not meet authorship criteria and are not listed as authors. Analyses and figures were produced in Python 3.14 (point release 3.14.5).

Results

Study selection

The searches returned 2,232 records (PubMed, n = 466; Scopus, n = 1,027; Web of Science, n = 247; EconLit, n = 353; Cochrane Library, n = 105; grey literature [WHO, ILO], n = 34). After removal of 609 duplicates, 1,623 records were screened by title and abstract and 1,433 were excluded. Of the 190 reports sought for retrieval, 56 were not assessed at full text (non-English language or outside the 2000–2026 window), leaving 134 reports assessed for eligibility. Seventy-six reports were excluded at full text—no blood-based measurement / health-risk-appraisal only (n = 29), no economic or productivity outcome (n = 22), review/protocol/editorial (n = 13), infectious-disease screening only (n = 7), anthropometric measures only (n = 2), not workplace-based (n = 2), and genetic/genomic testing (n = 1); the full list with exclusion reasons is provided in Additional file 1. A structured audit of the pre-specified blood-based eligibility criterion, conducted during data extraction and blind to economic outcomes, identified four reports initially judged eligible that recorded no blood analyte—health-risk-appraisal questionnaires only (Nyman 2010; Serxner 2012), blood pressure without a blood analyte (Greene 2009), and blood testing confined to the comparator arm (Steel 2022); these were reclassified as excluded (code E1) and are listed in Additional file 1. Fifty-eight reports met all criteria, corresponding to 56 unique included studies (Fig. 1) [10, 26–80]. None of the 34 records retrieved from grey-literature sources (WHO and ILO) was eligible for inclusion.

Fig. 1.

PRISMA 2020 flow diagram of study identification, screening, eligibility assessment, and inclusion

Study characteristics

Characteristics of the 56 included studies are summarised in Table 1 (per-study details in Additional file 1). Studies were published between 2000 and 2026 and were predominantly conducted in the United States (n = 35, 62.5%); the remaining 21 studies spanned 12 countries (Canada n = 4; Japan and Germany n = 3 each; Finland and South Africa n = 2 each; and one each from the United Kingdom, Australia, Sweden, Italy, India, the Netherlands, and Russia). Sectors were heterogeneous, including manufacturing and automotive, healthcare, financial services, retail, education, and public-sector employers. The median analysed sample was 3,444 participants (IQR 425–10,103; range 20–165,770). Among the 52 empirical studies reporting follow-up, the median was 24 months (IQR 12–51; range 0–156 months, with cross-sectional studies treated as zero months’ follow-up), while two of the three modelling studies adopted explicit time horizons (10 and 40 years); the third did not report one. Funding was not reported in 12 studies (21%) and, where reported, derived from a mix of public or charitable, university, and commercial (including pharmaceutical) sources.

Table 1.

Characteristics of the 56 included studies (study, country, sector, design, sample size, follow-up, screening type, blood analytes measured, post-screening intervention, primary economic outcome, and quality rating). Per-study details are also given in Additional file 1. Studies are ordered by study design (randomised controlled trials; quasi-experimental and controlled before–after; prospective cohort; retrospective cohort; cross-sectional; single-arm before–after; and modelling studies) and alphabetically within each design, matching the order used in Fig. 2

Study	Country	Sector	Design	N	Follow-up (mo)	Screening	Blood analytes	Intervention	Primary outcome	Quality
Adams 2024 [26]	United Kingdom	Healthcare	RCT	236	6	Comprehensive panel	Cholesterol, HbA1c, eGFR, creatinine	Coordinator-led	Absenteeism	RoB2: High
Herman 2014 [45]	Canada	Postal services	RCT	156	12	Comprehensive panel	Total cholesterol, HDL cholesterol	Coordinator-led	Composite	RoB2: Some concerns; CHEC: 18/19
Kontsevaya 2010 [52]	Russia	Workplace	RCT	468	12	Targeted biomarker	Total cholesterol	Coordinator-led	Cost-effectiveness	RoB2: Some concerns; CHEC: 9/19
Robroek 2012 [70]	Netherlands	Mixed	RCT	924	24	Check-up + lab	Total cholesterol	Self-management	Composite	RoB2: High; CHEC: 12/19
Song 2019 [10]	USA	Retail	RCT	32,974	18	Comprehensive panel	Total cholesterol, HDL cholesterol, glucose	Coordinator-led	Composite	RoB2: Some concerns
Steinberg 2015 [75]	USA	Health insurance	RCT	2,835	12	Comprehensive panel	Triglycerides, HDL-C, fasting blood glucose	Coordinator-led	Composite	RoB2: Some concerns
Allen 2012 [27]	USA	Education	Quasi-exp.	55	12	Comprehensive panel	Total cholesterol, LDL-C, HDL-C, triglycerides, fasting blood glucose, hsCRP	Coordinator-led	Cost-effectiveness	NOS-C: 8/9; CHEC: 15/19
Grossmeier 2013 [42]	USA	Energy	Quasi-exp.	29,642	24	Comprehensive panel	Blood panel (analytes not specified)	Coordinator-led	Composite	NOS-C: 8/9
Johnston 2015 [48]	USA	Higher education	Quasi-exp.	28,040	24	Comprehensive panel	Blood panel (analytes not specified)	Coordinator-led	Healthcare costs	NOS-C: 6/9
Kuehl 2013 [54]	USA	Fire service	Quasi-exp.	1,369	84	Comprehensive panel	Cholesterol, lipids	Coordinator-led	Composite	NOS-C: 8/9
Merrill 2018 [60]	USA	Public sector	Quasi-exp.	6,810	36	Comprehensive panel	Cholesterol ratio, glucose, blood pressure, BMI	Coordinator-led	Healthcare costs	NOS-C: 6/9
Mukhopadhyay 2013 [61]	USA	Not specified	Quasi-exp.	2,425	36	Comprehensive panel	Blood glucose, cholesterol	Coordinator-led	Composite	NOS-C: 8/9
Boffa 2016 [31]	Australia	Mixed	Prosp. cohort	3,550	NR	Targeted biomarker	Urea, electrolytes, creatinine, full blood count, NT-proBNP	Screening only	Cost-effectiveness	NOS-C: 6/9; CHEC: 5/19
Fischer 2020 [40]	Germany	Automotive manufacturi …	Prosp. cohort	3,992	12	Comprehensive panel	Total cholesterol, HDL-cholesterol, LDL-cholesterol, triglycerides, C-reactive protein, gamma-glutamyltransferase, glycosylated hemoglobin (HbA1c)	Screening only	Absenteeism	NOS-C: 7/9
Hermansson 2002 [46]	Sweden	Transport	Prosp. cohort	989	12	Targeted biomarker	Carbohydrate-deficient transferrin, gamma-glutamyltransferase	Screening only	Absenteeism	NOS-C: 6/9
Zhang 2020 [80]	Canada	Higher education	Prosp. cohort	196	12	Comprehensive panel	HDL cholesterol, total cholesterol	Coordinator-led	Composite	NOS-C: 4/9
Anderson 2000 [28]	USA	Mixed	Retro. cohort	46,026	36	Comprehensive panel	Total cholesterol, fasting blood glucose	Screening only	Healthcare costs	NOS-C: 8/9
Boyce 2024 [32]	USA	Healthcare and educati …	Retro. cohort	9,116	60	Comprehensive panel	Total cholesterol, HDL, LDL, triglycerides, blood glucose	Coordinator-led	ROI	NOS-C: 3/9
Burton 2002 [34]	USA	Financial services	Retro. cohort	1,773	36	Comprehensive panel	Complete blood count, blood chemistries (multiphasic), thyroid function tests, triglycerides, total cholesterol, HDL-cholesterol, LDL-cholesterol (calculated), hemoglobin A1C, PSA (prostate-specific antigen), dipstick urinalysis	Physician-led	Composite	NOS-C: 8/9
Dement 2015 [39]	USA	Higher education	Retro. cohort	10,432	60	Comprehensive panel	Cholesterol, blood pressure	Coordinator-led	Composite	NOS-C: 9/9
Forster 2026 [41]	Germany	Automotive manufacturi …	Retro. cohort	71,054	156	Check-up + lab	Gamma-glutamyl transferase (GGT), serum glutamic oxaloacetic transaminase (SGOT), serum glutamate-pyruvate transaminase (SGPT)	Physician-led	Absenteeism	NOS-C: 4/9
Henke 2011 [44]	USA	Manufacturing	Retro. cohort	63,646	84	Comprehensive panel	Total cholesterol	Coordinator-led	Composite	NOS-C: 8/9
Kirkham 2015 [50]	USA	Technology	Retro. cohort	17,089	48	Comprehensive panel	Blood glucose, cholesterol, triglycerides	Screening only	Composite	NOS-C: 5/9
Mehra 2012 [57]	India	Oil and gas	Retro. cohort	NR	60	Comprehensive panel	Blood panel (analytes not specified)	Physician-led	Absenteeism	NOS-C: 4/9
Merrill 2013 [58]	USA	Religious organization	Retro. cohort	10,721	72	Comprehensive panel	Total cholesterol, HDL, LDL, triglycerides, blood glucose	Self-management	Healthcare costs	NOS-C: 9/9
Merrill 2016 [59]	USA	Education	Retro. cohort	4,133	72	Comprehensive panel	Cholesterol, glucose	Coordinator-led	Composite	NOS-C: 7/9
Musich 2000 [62]	USA	Insurance	Retro. cohort	1,272	72	Comprehensive panel	Total cholesterol	Coordinator-led	Healthcare costs	NOS-C: 9/9
Musich 2001 [63]	USA	Manufacturing	Retro. cohort	3,338	48	Comprehensive panel	Total cholesterol	Coordinator-led	Composite	NOS-C: 5/9
Naydeck 2008 [64]	USA	Financial	Retro. cohort	3,784	48	Comprehensive panel	Cholesterol, glucose	Coordinator-led	Composite	NOS-C: 9/9
Ohata 2005 [65]	Japan	Workplace	Retro. cohort	17,647	84	Targeted biomarker	Pepsinogen I, pepsinogen II, pepsinogen I/II ratio	Screening only	Cost-effectiveness	NOS-C: 7/9; CHEC: 10/19
Piha 2017 [69]	Finland	Public sector	Retro. cohort	13,314	90	Check-up + lab	Complete blood count, serum cholesterol, fasting blood sugar, urine test	Coordinator-led	Absenteeism	NOS-C: 9/9
Schultz 2002 [72]	USA	Manufacturing	Retro. cohort	4,189	72	Comprehensive panel	Total cholesterol, HDL-C	Coordinator-led	Absenteeism	NOS-C: 9/9
Serxner 2003 [74]	USA	Automotive manufacturi …	Retro. cohort	26,411	72	Comprehensive panel	Cholesterol, blood pressure	Coordinator-led	Healthcare costs	NOS-C: 8/9
Tsurugano 2012 [77]	Japan	Insurance	Retro. cohort	4,513	36	Check-up + lab	Total cholesterol, HDL-C, triglycerides, uric acid, fasting plasma glucose	Screening only	Absenteeism	NOS-C: 2/9
Wright 2004 [79]	USA	Manufacturing, Banking …	Retro. cohort	165,770	24	Comprehensive panel	Total cholesterol	Coordinator-led	Healthcare costs	NOS-C: 6/9
Burton 2008 [35]	USA	Financial services	Cross-sec.	5,512	—	Comprehensive panel	Total cholesterol, LDL cholesterol, HDL cholesterol, triglycerides, fasting plasma glucose	Screening only	Absent.+present.	NOS-CS: 8/10
Kauppi 2024 [49]	Finland	Multiple	Cross-sec.	2,990	—	Comprehensive panel	Lipoprotein A1 (LIPOA1), lipoprotein B (LIPOB), high-sensitivity C-reactive protein (hsCRP), alanine aminotransferase (ALT), triglycerides, uric acid, HDL cholesterol, total cholesterol, glucose, LDL cholesterol, cholesterol-HDL ratio (KOL-HDL)	Screening only	Presenteeism	NOS-CS: 6/10
Kolbe-Alexander 2008 [51]	South Africa	Multiple sectors	Cross-sec.	1,954	—	Comprehensive panel	Total cholesterol	Screening only	Healthcare costs	NOS-CS: 7/10
Lucini 2014 [55]	Italy	Manufacturing	Cross-sec.	411	—	Check-up + lab	Total cholesterol, HDL cholesterol, LDL cholesterol, triglycerides, fasting plasma glucose, white blood cell count, red blood cell count, hemoglobin, hematocrit, high-sensitivity C-reactive protein, aspartate aminotransferase, alanine transaminase	Screening only	Absenteeism	NOS-CS: 6/10
Schultz 2009 [73]	USA	Manufacturing	Cross-sec.	4,188	—	Comprehensive panel	Fasting glucose, triglycerides, HDL-C, total cholesterol	Screening only	Composite	NOS-CS: 9/10
Bevis 2014 [29]	USA	Not specified	Before–after	224	12	Comprehensive panel	Fasting serum glucose, insulin, HbA1c, LDL cholesterol, HDL cholesterol, triglycerides	Coordinator-led	Composite	NIH: Fair
Bodin 2018 [30]	USA	Healthcare	Before–after	1,800	36	Comprehensive panel	LDL-C, triglycerides	Coordinator-led	Composite	NIH: Poor
Burton 2000 [33]	USA	Banking	Before–after	NR	18	Comprehensive panel	Cholesterol	Coordinator-led	Composite	NIH: Fair
Burton 2014 [36]	USA	Financial services	Before–after	7,252	12	Comprehensive panel	Blood glucose, total cholesterol, HDL cholesterol	Coordinator-led	Composite	NIH: Fair
Chenoweth 2005 [37]	USA	Manufacturing	Before–after	4,284	6	Check-up + lab	Serum lipoproteins	Physician-led	ROI	NIH: Poor; CHEC: 9/19
Davis 2009 [38]	USA	Public transportation	Before–after	300	48	Comprehensive panel	Fasting plasma glucose, total cholesterol	Coordinator-led	Composite	NIH: Poor
Johannigman 2010 [47]	USA	Mixed	Before–after	216	12	Comprehensive panel	HDL cholesterol, LDL cholesterol, triglycerides, blood glucose, HbA1c	Coordinator-led	Composite	NIH: Poor
Makrides 2011 [56]	Canada	Public administration	Before–after	402	48	Comprehensive panel	Cholesterol, glucose	Coordinator-led	Composite	NIH: Poor
Ott 2004 [66]	Germany	Chemical manufacturing	Before–after	4,789	24	Targeted biomarker	IgG antibodies against H. pylori	Physician-led	Composite	NIH: Fair
Palumbo 2013 [68]	USA	Healthcare	Before–after	80	12	Comprehensive panel	Total cholesterol, HDL-C, total cholesterol to HDL ratio, random glucose	Coordinator-led	Composite	NIH: Poor
Schouw 2020 [71]	South Africa	Energy	Before–after	156	24	Comprehensive panel	Random blood glucose, total cholesterol	Coordinator-led	Cost-effectiveness	NIH: Fair; CHEC: 12/19
Tarride 2011 [76]	Canada	Public administration	Before–after	141	12	Comprehensive panel	Total cholesterol, glucose	Coordinator-led	Composite	NIH: Fair; CHEC: 10/19
Wilson 2005 [78]	USA	Paper manufacturing	Before–after	36	11	Comprehensive panel	Total cholesterol, fasting blood glucose, LDL-C, HDL-C, triglycerides	Coordinator-led	ROI	NIH: Poor; CHEC: 13/19
Harber 2014 [43]	USA	Nuclear weapons	Modelling	1,000	480 (model)	Targeted biomarker	Lymphocyte proliferation test (LPT)	Screening only	Cost-effectiveness	CHEC: 11/19
Kowada 2018 [53]	Japan	General employment	Modelling	20	NR	Targeted biomarker	Helicobacter pylori antibody	Screening only	Cost-effectiveness	CHEC: 18/19
Ozminkowski 2004 [67]	USA	Electronics	Modelling	52,124	120	Comprehensive panel	Total cholesterol, blood glucose, blood pressure	Coordinator-led	ROI	CHEC: 13/19

NOS-C, Newcastle–Ottawa Scale (cohort, /9); NOS-CS, Newcastle–Ottawa Scale (cross-sectional, /10); NIH, NIH Quality Assessment Tool for Before–After Studies; RoB2, Cochrane Risk-of-Bias tool v2; CHEC, Consensus Health Economic Criteria (/19); ROI, return on investment; NR, not reported; mo, months. The Quality column reports the Layer-1 instrument; CHEC scores are shown additionally for the 13 studies presenting a full economic evaluation. Citation numbers [n] correspond to the reference list

Designs comprised retrospective cohort (n = 19), single-arm before–after (n = 13), randomised controlled trial (n = 6), quasi-experimental or controlled before–after (n = 6), cross-sectional (n = 5), prospective cohort (n = 4), and modelling (n = 3) studies. Most programmes used a comprehensive biomarker panel (n = 43) rather than targeted biomarkers (n = 7) or a check-up incorporating laboratory tests (n = 6). Consistent with the eligibility criteria, every included programme measured at least one blood analyte; lipid fractions (45 studies) and fasting glucose or glycated haemoglobin (29 studies) were the most frequent (per-study analytes are listed in Table 1). The post-screening intervention was coordinator-led in 35 studies, screening-only in 14, physician-led in 5, and self-management in 2. Outcomes were captured predominantly from administrative records (n = 44); validated productivity instruments were used in only 5 studies (4 exclusively, 1 alongside administrative data), and outcomes were modelled in 7 studies. The primary economic outcome was a composite in 26 studies, followed by absenteeism (n = 9), healthcare costs (n = 8), cost-effectiveness (n = 7), direct ROI (n = 4), and presenteeism (n = 1); one further study had a combined absenteeism–presenteeism primary outcome.

Methodological quality

Across all independently appraised items, the two reviewers’ parallel appraisals agreed on 74.3% of quality-appraisal items and 91.1% of effect-direction judgements before adjudication; the 226 discordant items, of 1,099 (763 quality-appraisal items and 336 effect-direction judgements across the 56 studies), were resolved by the senior author against the source record. Because appraisal was performed in duplicate with large-language-model support rather than by fully independent human dual rating, we report raw percentage concordance rather than a chance-corrected agreement statistic (e.g., κ or Gwet’s AC1); an adjudication log is provided in Additional file 1 (S6), listing individually the 217 of these discordant items for which a written adjudication rationale was recorded, with the instrument, item, final rating, and the reasoning applied against the source record; the remaining discordances were resolved against the source without a separately recorded rationale. None of the 6 RCTs was rated at low risk of bias on RoB 2 (some concerns, n = 4; high, n = 2). Among the 29 studies appraised with the NOS-cohort tool, the median score was 7 of 9 (good [≥7], n = 17; fair [5, 9], n = 7; poor [<5], n = 5). The 5 cross-sectional studies scored a median of 7 of 10 (range 6–9). None of the 13 single-arm before–after studies reached “good” on the NIH tool (fair, n = 6; poor, n = 7). For the 13 full economic evaluations, the median CHEC score was 12 of 19 (range 5–18). Item-level appraisals are provided in Additional file 1.

Synthesis of findings

Because the reported metrics could not be pooled, the synthesis reports two complementary things for each study: the direction of the reported effect in each outcome domain, and the characteristics of the programme that produced it, in particular the type of post-screening intervention. Direction of effect alone shows how consistently the literature reports benefit within a domain; pairing it with programme design shows whether that consistency is confined to particular kinds of programme, which is the question of practical interest to employers and health services. Both are displayed in the effect-direction plot (Fig. 2), which arranges studies by design and quality, and the harvest plot (Fig. 3), which arranges the same studies by post-screening intervention. Findings are reported below first by outcome domain and then across the intervention dimension.

Fig. 2.

Effect-direction plot of economic and productivity outcomes for the 56 included studies across six outcome domains, grouped by study design and ordered alphabetically within each design. The direction of effect in each domain is printed as a character code: + favourable, − unfavourable, 0 no effect, ± mixed, and · not reported. Study quality is printed as H (high), M (moderate) or L (low), and the analysed sample size (N) is given numerically; these were previously encoded by symbol shading and symbol size. Cell shading repeats the printed code and carries no additional information, so the plot is legible in black and white

Fig. 3.

Harvest plot of the direction of economic effect by outcome domain and post-screening intervention intensity (k = 56). Within each domain, each bar represents one study placed under its reported direction of effect; bar height denotes study quality (0–1) and bar colour denotes the type of post-screening intervention (physician-led, coordinator-led, self-management, or screening-only)

Healthcare costs were the most frequently reported domain (29 studies): 21 reported a favourable effect, 5 no effect, 2 an unfavourable effect, and 1 a mixed effect; favourable results were observed across quality strata, including 12 of the 13 higher-quality studies reporting this domain. Return on investment was reported in 20 studies and was uniformly favourable (favourable, n = 20), with favourable results distributed across low-, moderate-, and high-quality studies. Absenteeism, reported in 23 studies, showed an inconsistent pattern (favourable, n = 14; no effect, n = 6; mixed, n = 2; unfavourable, n = 1), with higher-quality studies divided between favourable and null findings. Cost-effectiveness was assessed in 8 studies (favourable, n = 7; no effect, n = 1), generally the formal economic evaluations. Presenteeism (7 studies; favourable, n = 4; no effect, n = 3) and staff turnover (1 study; no effect, n = 1) were rarely measured. Expressed as the proportion of reporting studies with a favourable effect (95% Wilson confidence intervals in parentheses), these were ROI 20/20 (100%; 84–100), cost-effectiveness 7/8 (88%; 53–98), healthcare costs 21/29 (72%; 54–85), absenteeism 14/23 (61%; 41–78), and presenteeism 4/7 (57%; 25–84); staff turnover was reported by a single study. The wide intervals reflect small per-domain samples and, together with the vote-counting basis of the synthesis, preclude precision-weighted inference. Among the studies expressing ROI as a benefit-to-cost ratio, reported returns ranged from approximately break-even (1:1) to nearly 16:1 and clustered between about 1.5:1 and 4:1; effect metrics in the other domains were too heterogeneous in unit, perspective, and time horizon to express as a common magnitude.

Across the SWiM intervention dimension, favourable economic results clustered in programmes with a structured post-screening pathway. All 20 studies reporting ROI, and 19 of the 21 reporting a favourable healthcare-cost effect, were coordinator-led or physician-led programmes; of the 14 screening-only programmes, only one (Kowada) reported a healthcare-cost outcome and none reported ROI. Comprehensive-panel programmes, the majority, accounted for most favourable cost and ROI results; the small numbers of targeted-biomarker and check-up programmes precluded clear comparison.

Sensitivity and subgroup analyses

The principal patterns were robust. Excluding the 3 modelling studies left the direction of effect essentially unchanged (healthcare costs, 19 favourable of 27 reported; ROI, 19 favourable of 19). Restricting to moderate- and high-quality studies preserved the findings: healthcare costs remained predominantly favourable (17 favourable, 4 no effect, 2 unfavourable), ROI remained favourable (all 13), and absenteeism remained inconsistent (9 favourable, 5 no effect, 1 mixed, 1 unfavourable). Subgroup analysis by intervention type reproduced the concentration of favourable results in programmes with structured follow-up; subgroups by screening type and productivity-measurement method were too sparse for meaningful comparison.

Discussion

Principal findings

In this systematic review of 56 studies of workplace health programmes incorporating blood-based laboratory or biomarker screening, the economic and productivity evidence was uneven across outcome domains. Healthcare-cost outcomes, the most frequently reported, were predominantly favourable, and this signal persisted among higher-quality studies. Return on investment was reported less frequently but was uniformly favourable; notably, favourable ROI was distributed across low-, moderate-, and high-quality studies, with no attenuation in the direction of effect among more rigorous work. Effects on absenteeism were inconsistent, with higher-quality studies divided between favourable and null findings. Presenteeism and staff turnover were rarely measured. Across the synthesis, favourable economic results were concentrated in programmes that coupled screening to a structured post-screening intervention (coordinator- or physician-led), whereas programmes offering screening alone seldom reported—or demonstrated—an economic return.

Comparison with previous evidence

These findings both extend and qualify the existing literature. The optimistic conclusions of early meta-analytic and narrative syntheses [3, 5, 9] contrast with the null results of the two large randomised trials of workplace wellness [10–13] and with the cautious conclusions of large independent evaluations [4]. Of particular relevance, the Illinois trial demonstrated that apparent savings in observational designs are substantially attributable to the selection of healthier, lower-cost employees into participation [12]. Our ROI finding is most informative when read against this backdrop and against the demonstration that reported ROI varies inversely with methodological rigour [7]—an attenuation of effect magnitude that our direction-based synthesis was not designed to detect: the uniform favourability of ROI in our sample, the flat distribution of its direction across quality strata, and its reliance on heterogeneous and frequently single-arm designs are together more consistent with selective reporting and residual confounding than with a robust causal return. The healthcare-cost signal is more nuanced. Unlike ROI, favourable cost results persisted among higher-quality studies, which a vote-counting summary alone cannot dismiss; however, the small number of randomised data points argues for caution, since the higher-quality observational studies remain susceptible to the same selection and confounding that the Illinois trial exposed.

Read together, these comparisons define what our review adds. It complements the broad synthesis of workplace-prevention ROI by Thonon and colleagues [6] by isolating the blood-based screening subgroup that broader reviews pool with educational, behavioural, organisational, and technical interventions. Within that subgroup the economic signal is not uniform but domain-specific and conditional on programme design: favourable for healthcare costs and return on investment, inconsistent for absenteeism, and largely untested for presenteeism. How far this generalises is constrained by the evidence base itself, which is predominantly US, employer-perspective, and observational; the findings therefore characterise the literature as it currently stands rather than establishing a transferable effect.

The intervention pathway

The most consistent programmatic signal in our synthesis was the concentration of favourable economic results in programmes with a structured follow-up pathway, rather than in screening itself. This pattern is operationally coherent and, in our reading, central to interpreting the whole evidence base: a blood test changes outcomes only insofar as its results are acted upon. Detection is not an intervention. In our sample the separation was stark: every one of the 20 ROI estimates, and 19 of the 21 favourable healthcare-cost results, came from programmes with a structured coordinator- or physician-led pathway, whereas of the 14 screening-only programmes only one reported a healthcare-cost outcome and none reported ROI. Where a programme provided a structured pathway—coordinator- or physician-led follow-up that actively supports employees in obtaining treatment and sustaining behaviour change—screening could plausibly translate into improved health and, in turn, an economic return; where screening stood alone, any response was left to the discretion of the individual employee and was therefore inconsistent and largely unmanaged, and such programmes seldom reported or demonstrated a return. The implication is to reframe the policy question from whether workplace blood-based screening saves money to whether the surrounding programme acts effectively on what screening detects, and to recognise that the quality of post-screening support—not the test itself—is the more probable determinant of both health and economic impact. This interpretation must, however, be advanced cautiously. Programmes with coordinator- or physician-led follow-up were also disproportionately those that measured and reported ROI and were more often evaluated by or for the sponsoring organisation, so the association between intervention intensity and favourable economics is itself vulnerable to reporting and confounding bias; relatedly, the measurement of returns from such integrated programmes is known to be inconsistent [8].

Measurement of productivity

A striking feature of the evidence base was the near-absence of presenteeism and turnover outcomes and the overwhelming reliance on administrative records rather than validated productivity instruments; only five of fifty-six studies used a validated instrument. Presenteeism may represent a substantial component of the productivity cost of ill-health, yet it was assessed in few studies and rarely with the validated tools developed for the purpose [1, 2, 81]. The productivity case for workplace blood-based screening therefore remains largely untested, and the predominant outcome metrics may understate or misrepresent the full economic picture.

Strengths and limitations

To our knowledge, this is the first systematic review to isolate workplace programmes with a blood-based laboratory or biomarker screening component and to synthesise their economic and productivity outcomes. Its strengths include prospective registration, a comprehensive search across five databases and grey literature, duplicate study selection and appraisal with independent adjudication, the use of design-appropriate quality instruments alongside a dedicated appraisal of full economic evaluations, and a transparent synthesis [16] supported by effect-direction and harvest plots and by sensitivity and subgroup analyses that confirmed the robustness of the principal patterns.

Several limitations qualify these conclusions. First, the reported outcomes were too heterogeneous in metric, analytic perspective, and time horizon to support meta-analysis or formal certainty rating; the synthesis therefore rests on the direction of effect, which does not weight by precision or magnitude and carries the recognised limitations of vote-counting. Second, the evidence base was dominated by observational and single-arm designs, included only six randomised trials—none at low risk of bias—and was thus generally susceptible to selection and confounding. Third, the evidence base is strongly US-centric: almost two-thirds of the studies (35 of 56) were conducted in the United States, most adopted an employer perspective (39 of 56), and most reported costs in US dollars (38 of 56). In the US employer-sponsored insurance system, employers bear a substantial share of their workforce’s healthcare costs directly, which creates both the incentive to pursue cost-saving screening and the claims infrastructure to measure it; the favourable cost and return-on-investment findings may therefore not transfer to health systems with predominantly public or universal coverage—including most of Europe—where employers do not internalise these costs and the economic rationale for workplace screening differs accordingly. Fourth, the literature is likely affected by publication and selective-reporting bias, most acutely for ROI. Fifth, eligibility was restricted to English-language reports published from 2000 onwards, which may have excluded relevant evidence. Finally, heterogeneity in the composition of screening panels and in the nature of the post-screening interventions limits precise attribution of effects to specific programme components.

Implications

For employers and occupational-health services, these findings counsel against assuming that blood-based screening will, in itself, reduce healthcare costs or yield a positive return; any economic benefit appears to depend not on testing but on the quality of the support that helps employees act on their results—structured, coordinator- or physician-led follow-up—without which screening is unlikely to change health behaviour or costs, and the favourable ROI estimates that dominate the literature should be interpreted with caution given their dependence on study quality. For researchers, the priorities are rigorous controlled designs with concurrent comparison groups, the use of validated productivity instruments to capture presenteeism, economic evaluations with explicit perspectives and time horizons, evidence from outside the United States, and explicit reporting of the screening-to-intervention pathway. For policymakers, incentives for workplace prevention might more appropriately reward integrated programmes and rigorous evaluation than the provision of screening alone.

Conclusions

In this review of 56 studies, workplace health programmes incorporating blood-based screening were associated with favourable healthcare-cost and return-on-investment results in most studies that reported them. The return-on-investment signal showed no attenuation in the direction of effect with study quality, effects on absenteeism were inconsistent, and presenteeism and staff turnover were seldom measured. Favourable economic outcomes were observed almost exclusively in programmes that coupled testing to a structured, coordinator- or physician-led pathway supporting employees to act on their results, rather than in screening offered alone.

These are associations observed across a heterogeneous and largely observational literature, and should not be read causally. The uniformly favourable return-on-investment result is difficult to separate from reporting and selection bias. Because screening-only programmes were rarely evaluated economically at all, returns could not be compared directly across programme types, so the contrast between programme designs reflects what has been evaluated as much as what works.

On this evidence, blood-based workplace screening is not supported as a stand-alone cost-saving measure. We advance as a hypothesis requiring prospective evaluation, rather than as a demonstrated effect, that its value lies less in testing itself than in the support that converts detection into sustained behaviour or treatment change. Rigorous, controlled, and non-US evidence using validated productivity measures is needed to test it.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

CHEC

Consensus Health Economic Criteria

GRADE

Grading of Recommendations, Assessment, Development and Evaluations

ILO

International Labour Organization

IQR

Interquartile range

NHLBI

National Heart, Lung, and Blood Institute

NIH

National Institutes of Health

NOS

Newcastle–Ottawa Scale

OSF

Open Science Framework

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

PRISMA-P

PRISMA for Protocols

RCT

Randomised controlled trial

RoB 2

Cochrane risk-of-bias tool for randomised trials, version 2

ROBINS-I

Risk Of Bias In Non-randomised Studies—of Interventions

ROI

Return on investment

SWiM

Synthesis Without Meta-analysis

WHO

World Health Organization

Author contributions

S.A. conceived and supervised the review and acted as adjudicator. D.S. and S.M. designed and conducted the literature searches, study selection, data extraction, and quality appraisal in duplicate, with S.A. adjudicating disagreements. D.S. drafted the manuscript, and S.M. and S.A. critically revised it for important intellectual content. All authors read and approved the final manuscript.

Funding

This work was supported by the European Funds for Lower Silesia (grant no. FEDS.01.02-IP.01-0078/24). The funder had no role in the design of the review; in the collection, analysis, or interpretation of data; or in the writing of the manuscript or the decision to submit it for publication.

Data availability

All data generated or analysed during this study are included in this published article and its additional files. The study-level dataset underlying the synthesis — for each included study: design, setting, sample size, follow-up, funding source, screening and post-screening intervention type, biomarkers measured, the direction of effect in each of the six outcome domains, and quality-appraisal scores — is provided in Additional file 1 (Supplementary Tables S1—S6), together with the verbatim search strategies, the full list of full-text exclusions with reasons, the item-level quality appraisals, and the data-extraction form. The review protocol is registered on the Open Science Framework (https://doi.org/10.17605/OSF.IO/M4W8C).

Declarations

Ethics approval and consent to participate

Not applicable. This study is a systematic review of previously published data and did not directly involve human participants, their data, or biological material.

Consent for publication

Not applicable.

Competing interests

SA is a shareholder and employee of Labplus S.A., which develops LabTest Checker®, a Class IIa medical device under Regulation (EU) 2017/745 (MDR); LabTest Checker was not used in any phase of this review. DS is a shareholder and employee of Housemed S.A. SM declares that he has no competing interests.