A defensible blood gas analyzer comparison starts by separating three questions that are often conflated in procurement: whether the device can measure the analytes you need, whether its documented performance is adequate for your clinical setting, and whether you can sustain that performance across multiple instruments. None of these questions is answered by the headline feature list. A US guidance document for indwelling analyzers points to specific NCCLS methods—including EP5-A for precision, EP6 for linearity and EP9 for patient-sample comparison—as the testing framework on which a device's safety and effectiveness claims should rest [3]. That framework also gives purchasing teams a concrete way to rank candidates using evidence that is comparable across vendors, rather than relying on each manufacturer's preferred demonstration of a product. A procurement dossier should therefore request the same named studies from every bidder, then evaluate the completeness of the response before any workflow demonstration.

Anchor the Comparison in Method Studies

The most useful evidence for comparing analyzers is a head-to-head method study that applies identical protocols to all instruments. One published evaluation tested ten analytes—pH, pCO2, pO2, Na+, K+, Cl-, iCa2+, glucose, lactate and hematocrit—across six point-of-care analyzers, using the i-SmartCare 10 as the comparator and following CLSI guidelines [1]. Because i-Stat does not test chloride and lactate, those two measurands could not be evaluated across the full six-device set, a limitation that matters if your panels require both [1]. A procurement team reviewing this study can see exactly which analyzer combinations were compared and which among them were excluded for specific analytes, rather than assuming that a broad claimed menu implies identical coverage across devices.

That same study reports precision estimates against Ricos desirable specifications, carry-over calculations, linearity slopes with 95% confidence intervals, recovery percentages and Bland-Altman mean differences [1]. Some slopes had confidence intervals that failed to include 1.0, and some recovery ranges narrowly escaped acceptance criteria [1]. The document does not assert a single best device; it supplies the granular results a laboratory would need to judge each analyte. For procurement, this illustrates the right request: require bidders to identify the reference method, the guideline version, the sample types and the analytes excluded from comparison. A vendor that cannot produce this information has not yet provided the evidence basis for a comparative decision.

Use Structured Health Technology Assessment Dimensions

A method comparison answers only one part of the purchasing question. A study using the EUnetHTA core model for blood gas analyzers developed, through two Delphi rounds with 11 healthcare experts, a structure of 71 issues across 26 topics and 10 domains [6]. The ten context parameters identified in that work define the starting scenario for any technology assessment, meaning that the model is intended to be applied to a specific clinical environment rather than to produce a universal ranking. The authors themselves note that findings may be limited by the small number of participants drawn from one healthcare area or country [6].

For a procurement team, the value of that model is not the specific 71 issues, which are not fully itemized in the retrieved source text, but the demonstration that clinical requirements, contextual constraints and technical performance can be organized into a repeatable scoring structure. The paper states that the methodology could expedite purchasing by supporting detailed requests for proposal to providers during procurement [6]. A team can adopt this logic without copying the model wholesale: define the clinical question the analyzer must answer, the physical and workflow constraints of the testing site, and the evidence required for each domain. This prevents a single enticing specification from dominating the evaluation.

Request the same five named studies from every bidder

Ask vendors to submit results against NCCLS EP5-A for precision, EP6 for linearity, EP7 for interference testing, EP9 for patient-sample comparison and EP10 for preliminary evaluation, as referenced in the FDA special controls guidance [3].

Confirm the analyte list actually tested, the exclusion rules, the reference instrument and any platform-specific limitations before scoring. In the six-analyzer study, chloride and lactate were not tested on one device, so a universal menu cannot be assumed across all candidates [1].

Add a paired-analyzer maintenance plan to the contract, with defined failure limits and cartridge-lot tracking. The three-year comparison showed that initial installation testing did not catch recurring glucose discrepancies that appeared later [7].

Verify Test Menu, Workload and Regulatory Scope Separately

Vendor documentation for a cartridge-based system, such as the Siemens RAPIDLab 1200 pages, describes a medium-to-high workload setting and lists features including an automatic quality control cartridge with a 28-day use life, a 30-day reagent cartridge use life and calibration every 30 minutes [5]. These are manufacturer statements, not independent findings, and they should be treated as claims to verify at your site. The same pages describe a test menu that varies by model, including blood gas, electrolytes, metabolites, CO-oximetry and neonatal total bilirubin, with all results from a single sample [5]. A procurement team should ask which parameters are actually available on the model being offered, because the phrase "assays vary by model" shifts a significant part of the decision away from the brochure and into the configuration request.

Regulatory scope is a separate filter. The FDA guidance on indwelling blood gas analyzers identifies four risks to health associated with those devices and recommends performance studies using NCCLS documents [3]. It also states that FDA considers indwelling sensors to be of moderate concern for software review, and it expects labeling to include shelf life, bias, precision, correlation coefficient, known limitations or interferences, and a non-pyrogenic statement [3]. A comparison that ignores this distinction may treat a point-of-care benchtop analyzer and an indwelling sensor system as interchangeable, when their evidentiary and safety questions are not identical. Confirm which device class each candidate belongs to and request the documentation appropriate to that class.

Plan for Multi-Instrument Comparability, Not Just Single-Device Accuracy

Many hospitals operate more than one blood gas analyzer, and a three-year monitor of two identical analyzers offers a practical view of what can go wrong after installation. The study tested one blood sample on both analyzers at each of three shifts per day for three years and used Deming regression for pH, pCO2, pO2, sodium, potassium, chloride, ionized calcium, glucose and lactate [7]. Paired-analyzer failures were determined by performance limits from the Institute of Quality Management in Healthcare and from the manufacturer, producing different failure counts per 1,000 tests: for pO2 and glucose the IQMH and manufacturer limits gave 0.6 and 49.2, and 19.3 and 4.4, respectively [7].

The authors concluded that results were interchangeable across the two analyzers, but they also observed occasional persistent glucose discrepancies, defined as more than five failures per week, which they suggested might relate to different cartridge lots used on each analyzer and which were not caught during initial installation [7]. This is a concrete procurement lesson: the request should specify ongoing paired-analyzer comparison frequency, the performance limits to be used for flagging failures, and a mechanism for lot-to-lot cartridge monitoring after go-live. A candidate device may perform well in a single-analyzer evaluation and still create operational risk when a second unit with a different cartridge lot is placed in the emergency department.

A separate review article notes that blood gas testing programs commonly face quality assurance challenges and recommends practical approaches to method verification when choosing an analyzer type [8]. It also states that calibration verification across the manufacturer's claimed analytical measurement range is required before patient testing and every six months thereafter for moderate complexity testing under CLIA regulations, and that commercially available linearity material may not always span the claimed range [8]. A team can use this to ask bidders how their calibration verification materials behave across the full claimed range, and how long it takes to complete that requirement on multiple units.

A Worked Evaluation Example Using the Evidence Filters

Consider a hypothetical, suggested evaluation for a two-site emergency department planning to replace four benchtop analyzers. The first filter is menu coverage. The published six-analyzer comparison shows that chloride and lactate could not be compared on one device, so any candidate that cannot demonstrate those analytes is excluded or evaluated only for a limited deployment [1]. The second filter is documented method performance under CLSI protocols. A bidder should supply precision, linearity, recovery and carry-over data; if only marketing material is available and the named studies are absent, the evaluator records that gap as unresolved rather than assuming acceptable performance [1].

The third filter is multi-instrument experience. The request for proposal could specify that the vendor must describe a paired-analyzer monitoring protocol and a cartridge-lot reconciliation process, informed by the glucose discrepancies seen in the three-year comparison [7]. The fourth filter is calibration verification burden across the claimed analytical measurement range, which the review article identifies as a recurring source of practical frustration [8]. This hypothetical workflow does not choose a winner. It produces a shortlist of devices that can answer the same clinical questions with comparable evidence, which is a stronger procurement outcome than a scorecard filled with unverified vendor claims.