Advice
Evidence review
Evidence review
Clinical and technical evidence
Regulatory bodies
A search of the Medicines and Healthcare Products Regulatory Agency website revealed that no manufacturer Field Safety Notices or Medical Device Alerts for this device. No reports of adverse events were identified from a search of the US Food and Drug Administration (FDA) database: Manufacturer and User Device Facility Experience (MAUDE).
Clinical evidence
A literature search for evidence identified 7 studies that used the i‑STAT analyser with either the CG4+ or CHEM8+ cartridges for point‑of‑care tests. Of these studies, 5 used the CG4+ cartridge (Shapiro et al. 2010; Rossi and Khan 2004; Thomas et al. 2009; Karon et al. 2007; Singer et al. 2014), 1 used the CHEM8+ cartridge (Jarvis et al. 2015), and the other used both CG4+ and CHEM8+ cartridges (Jarvis et al. 2014). Another study (Shephard et al. 2012) also used the CHEM8+ cartridge, but was excluded because it did not report outcomes relevant to this briefing.
Four studies were identified where the papers do not report which tests were used but where it is likely, based on information from the company and from specialist commentators, that either the i‑STAT CHEM8+ or CG4+ cartridges were used. For completeness, a brief summary of each study is in table 15.
The Shapiro et al. (2010) study (tables 1 and 2) was a prospective cohort study in an urban tertiary care ED in the USA, carried out between May 2006 and March 2007. It aimed to study the feasibility and accuracy of the i‑STAT using CG4+ cartridges for bedside serum lactate measurements, and to determine if other measurements (pH and base excess) are predictive of mortality. It was off-label use of the i‑STAT CG4+ cartridge, because serum is not an approved sample type for the i‑STAT CG4+ cartridge. A convenience sample of 699 adults attending the ED with suspected sepsis during the study period was included. Lactate measurements were taken using the CG4+ cartridge and a mandatory confirmatory lactate measurement was done by the hospital's clinical laboratory.
Of the 699 patients in the cohort, 34 (4.9%) died. The area under the curve in receiver operating curve (AUROC) analysis for mortality prediction was 0.72 for i‑STAT lactate, 0.70 for laboratory lactate, 0.60 for pH measurement and 0.60 for base excess. A Bland–Altman plot showed that the lactate measurement using i‑STAT was accurate for clinical decision‑making compared with the laboratory test. The i‑STAT lactate measurement was on average 0.32 mmol/l lower than laboratory lactate (standard deviation 0.45; 95% confidence interval [CI] ‑0.35 to 0.98) with the limits of agreement ranging from −1.1 to 0.50. The i‑STAT lactate was highly correlated with the laboratory lactate with an r value of 0.97.
The study by Rossi and Khan (2004; tables 3 and 4) was a before‑and‑after study conducted in a cardiac intensive care unit in a paediatric hospital in the USA. It aimed to evaluate the combination of 2 strategies, goal‑directed therapy and point‑of‑care blood lactate testing using the i‑STAT CG4+ cartridge, in improving outcomes for young children (aged under 1 year) and babies (under 1 month) after congenital heart surgery. Group A (851 patients) had surgery before the i‑STAT was implemented in the care unit and group B (378 patients) had surgery after implementation. Measurements included overall mortality at 30 days after surgery, blood lactate levels, cardiopulmonary bypass times and aortic cross‑clamp times.
The study found that overall mortality was significantly lower for group B (2.4%) compared with group A (6.2%; p=<0.007). A significant reduction in mortality between group B and group A was observed in babies (4.3% compared with 12%, p=0.008), but the reduction did not reach significance in young children (0.9% compared with 2.6%, p value not reported).
The turnaround time for lactate was 120 seconds using the i‑STAT system and 15 minutes to 2 hours with the laboratory test.
The study by Karon et al. (2007; tables 5 and 6) was a cohort study conducted in the USA. It compared lactate values obtained from laboratory (plasma‑based assays) and point‑of‑care (whole blood) platforms to determine whether clinically relevant discrepancies might occur between the values obtained from the 2 methods.
Whole‑blood specimens were obtained from patients in the ED and intensive care unit (n=90), and were analysed using 3 different methods: the Radiometer ABL 715 blood gas analyser, the i‑STAT (with CG4+ cartridge), and the Nova analyser. All tests were done within 1–2 minutes of each other and within 1 hour of the blood sample being taken. Within 5 minutes of whole‑blood analysis, the blood specimens were centrifuged. The plasma was separated and kept on ice. Plasma lactate was analysed using the Integra Roche analyser and the Vitros analyser within 1 hour of plasma separation. It was unclear how the samples and patients were selected.
The authors found that correlation between lactate methods was good, with slopes of best fit of 0.87–1.06 and intercepts of 0.1–0.2 mmol/l lactate for all 4 methods compared with the Vitros.
At high lactate values (>6 mmol/l), the i‑STAT system showed negative bias (relative to the Vitros), and reported lower lactate results compared with the Vitros and Integra.
Of the 90 samples tested, the i‑STAT lactate values for 85 of the samples (94%) fell within the same risk category as the Vitros value.
The study by Thomas et al. (2009; tables 7 and 8) was a prospective cohort study conducted in a level 1 trauma centre in the USA. It evaluated the 'measure of treatment agreement' – the number of standard clinical laboratory arterial blood gas measurements that prompted changes in mechanical ventilator support therapy compared with the number of portable device measurements that would have prompted the same or different changes. The study included 446 intubated adult intensive care unit patients. Measurements taken with the i‑STAT system (using the CG4+ cartridge for arterial O2 saturation, PO2, pH and PCO2) and 2 other test devices (for peripheral capillary O2 saturation and end‑tidal CO2) were compared with paired standard laboratory measurements for arterial CO2, PO2, pH and PCO2.
Testing for equivalence found that the i‑STAT PO2, i‑STAT pH and i‑STAT PCO2 measurements were deemed 'equivalent' surrogates to paired laboratory analysis.
The measure of treatment agreement between the i‑STAT and paired laboratory blood gas values was 73% for arterial O2 saturation, 97% for PO2, 88% for pH and 95% for PCO2. Based on a minimum of 95% treatment agreement, only the i‑STAT PO2 and the i‑STAT PCO2 measurements were considered acceptable surrogates to those done in the laboratory.
The study by Singer et al (2014; tables 9 and 10) was also a before‑and‑after study done in an ED in the US. It assessed the effects of bedside point‑of‑care lactate measurement using the i‑STAT CG4+ cartridge on the time to administration of intravenous fluids and antibiotics in adult ED patients with suspected sepsis. Bedside point‑of‑care lactate measurement in a convenience sample of 80 patients with suspected sepsis presenting in the ED was compared with laboratory lactate measurement in the first 80 consecutive patients with suspected sepsis presenting the ED 12 months before the introduction of the bedside lactate testing. Only patients who had an initial lactate level of ≥2 mmol/l were included in the study. Of these, patients whose initial lactate level was 4 mmol/l or greater were transferred to the critical care area for further evaluation and management. The primary outcome measure was time to intravenous fluid and time to intravenous antibiotics.
The study found that introducing the bedside point‑of‑care lactate testing had a statistically significant reduction in median (interquartile) time to intravenous fluid administration compared with the laboratory lactate testing (55 [34–83] minutes compared with 71 [42–110] minutes; p=0.03). No statistically significant difference in median (interquartile) time to intravenous antibiotics administration was observed between the two groups (89 [63–182] minutes in the point‑of‑care testing group compared with 97 [55–160] minutes in the laboratory testing group; p=0.59).
The study by Jarvis et al. (2014; tables 11 and 12) was a before‑and‑after study conducted in an ED in a district general hospital in the UK with approximately 65,000 ED attendances a year. The study assessed the introduction of a rapid consultant‑led assessment model supported by point‑of‑care testing (phase 2, between 30 September and 18 October 2013, n=787) and how it affected the time patients spent in the ED, when compared with nurse‑led triage (phase 1, between 1 April and 24 May 2013, n=3835). The rapid assessment used point‑of‑care testing for the analysis of renal function (using the i‑STAT CHEM8+ cartridge), blood gases (using the i‑STAT CG4+ cartridge) and full blood counts (using another assay).
The study found that there was a significant reduction of 53 minutes (or 41.1%) in the median time for patients to be declared ready to leave the ED in phase 2 compared with phase 1 (p=0.0025).
The authors conducted another very similar study, Jarvis et al. (2015; tables 13 and 14), which was a before and after study in a seemingly identical setting. It assessed the impact of introducing point‑of‑care testing for renal function on the length of time patients spend in the ED. It consisted of 2 consecutive phases: phase 1 (between 1 April and 24 May 2013, n=3835), during which renal function was tested using the hospital's centralised laboratory analyser (which seemed to be identical to the phase 1 in the Jarvis et al. [2014] study), and phase 2 (between 28 May 2013 and 29 September 2013, n=7033) during which renal function was tested using the i‑STAT with the CHEM8+ cartridge.
The study found that there was a significant reduction of 20 minutes (or 15.5%) in the median time for patients to be declared ready to leave the ED in phase 2 compared with phase 1 (p=0.0025).
Recent and ongoing studies
One in‑development trial of i‑STAT for point‑of‑care testing was identified in the preparation of this briefing (Clinicaltrials.gov identifier: NCT02189096). The trial is not yet open to participants. The condition is sepsis and the interventions are the use of a standard single National Early Warning Score and sepsis screening, and point‑of‑care lactate measurement.
Costs and resource consequences
No published evidence on resource consequences was identified. Savings with point‑of‑care testing could be achieved by improving patient flow, ED throughput and clinical decision‑making. In practice, point‑of‑care testing is often introduced as part of complex ED service redesign.
Two reports of what appears to be the same service improvement project (Jarvis et al. 2014 and Jarvis et al. 2015) report before‑and‑after results when a rapid consultant‑led assessment model supported by point‑of‑care testing with i‑STAT was introduced in a UK district general hospital ED. The rapid assessment used point‑of‑care testing for the analyses of renal function (using the i‑STAT CHEM8+ cartridge), blood gas (using the i‑STAT CG4+ cartridge) and full blood counts (using another assay) with a median reduction of 53 minutes (or 41.1%) in the time for patients to be declared ready to leave the ED. Analysis of renal function using the i‑STAT CHEM8+ cartridge resulted in a median reduction of 20 minutes for patients to be declared ready to leave the ED.
Strengths and limitations of the evidence
Seven relevant studies were identified: 5 used the i‑STAT CG4+ cartridge, 1 used the CHEM8+ cartridges, and the other used both the i‑STAT CG4+ and CHEM8+ cartridges.
All the studies were cohort studies, with 3 being prospective single arm and the other 4 using a before‑and‑after comparison model. Of the 3 single‑arm cohort studies, 1 study used convenience sampling, which may not be representative of the study population (Shapiro et al. 2010). It was also unclear how the test samples and patients were selected in the Karon et al. (2007) study.
Of the 5 studies that used the CG4+ cartridge, 3 studies assessed the correlation between different test methods (Karon et al. 2007; Thomas et al. 2009; Singer et al. 2014). One study assessed mortality following the implementation of a patient management strategy based on lactate measurements using the i‑STAT CG4+cartridge compared with the strategy implemented without the point‑of‑care test (Rossi and Khan 2004). Only 1 study evaluated the diagnostic test accuracy of the test, by measuring the AUROC analysis for mortality prediction against a laboratory test (Shapiro et al. 2010).
In the study by Singer et al. (2014), sample size was calculated for the 2 comparison groups. The investigators who determined ultimate diagnosis and severity of sepsis, source of infection, Sequential Organ Failure Assessment (SOFA) scores, Modified Early Warning Scores (MEWS) scores, length of stay, and in‑hospital mortality were masked to study group and lactate levels. There were no statistically significant differences between the comparison groups in the majority of baseline demographic and clinical characteristics. However, the reported time from arrival to laboratory testing results and the time from order to laboratory testing results were significantly shorter in the i‑STAT CG4+ group than in the laboratory group, so there may have been influences other than the testing on the general care and clinical process (for example, improved time to decide to take a test). The paper also reported a statistically significant lower mortality rate in the 'after' group than in the 'before' group. However, like other before‑and‑after studies in the briefing, it is possible that any difference noted between the groups was because of other unmeasured confounding variables rather than introduction of the point‑of‑care testing. No studies on the diagnostic test accuracy or performance characteristics of the i‑STAT CHEM8+ cartridge were identified.
Both the study that used CHEM8+ and CG4+ cartridges (Jarvis et al. 2014) and the study that used only CHEM8+ cartridges (Jarvis et al. 2015) evaluated the impact of introducing point‑of‑care testing on the length of time patients spent in ED. This also involved a redesign of the service. Although these studies had positive results, the use of the i‑STAT for point-of-care testing was only 1 part of a complex service redesign and the impact of i‑STAT alone cannot be evaluated. Both studies were single institution before‑and‑after studies which provide relatively weak elements.
In 4 studies (Shapiro et al. 2010; Rossi and Khan 2004; Karon et al. 2007; Singer et al. 2014), the CG4+ cartridge was used for testing blood lactate levels. Currently, there is no reference standard for lactate measurement. In the Karon et al. (2007) study, the Vitros assay was used as the reference method, which might not be a perfect reference standard to assess diagnostic test accuracy. However, the use of this reference standard was not problematic because the aim of the study was to assess the agreement and discrepancies between different test methods.
In the Rossi and Khan (2004) study, lactate values using the i‑STAT CG4+ cartridge were included as part of the post‑operative management strategy for the patients after congenital heart surgery. Post‑operative mortality was the primary outcome measure. The duration of the study from start (phase 1, before i‑STAT was introduced) to finish (phase 2, with i‑STAT) was nearly a decade. It was unclear whether there might have been confounding factors contributing to the difference in the mortalities observed between the 2 phases, other than the introduction of the point‑of‑care lactate measurement.
Both Jarvis et al. studies (2014 and 2015) were conducted in a UK district general hospital, indicating that their results are likely to be generalisable to the NHS. The other 5 studies were conducted in the USA and may not be so reflective of NHS practice.
The Shapiro et al. (2010) and Singer et al. (2014) studies were funded by the manufacturer. For the Singer et al. (2014) study, the manufacturer was also consulted during the design of the study; furthermore, the first author of this study is on the speaker's bureau of the manufacturer. In the Jarvis et al. (2014) study, the manufacturer donated the i‑STAT CG4+ and CHEM8+ cartridges used. One of the authors of this study also served as an expert speaker and received honoraria from the manufacturer, and the Jarvis et al. (2015) study was supported by a grant from the manufacturer.
With regards to the 4 studies/articles that were outlined for completeness, it is uncertain whether they contribute to the evidence base due to uncertainty about whether the point‑of‑care tests used are within the scope of the MIB. In studies where the i‑STAT testing was used with other devices as part of service redesign package, the i‑STAT's individual contribution to the overall service improvement is uncertain. Furthermore, with the exception of the randomised controlled trial, these articles reported only a limited amount of data. The briefing was restricted to exclude studies in which the cartridges were unspecified; selection bias could be introduced if only a number of selected studies were included from those studies in which the cartridges were unspecified.
Overall, current published evidence on the diagnostic accuracy of the i‑STAT CG4+ and i‑STAT CHEM8+ tests is sparse although the manufacturer provides some information on test performance for each individual test on its website. Each of the identified studies had diverse patient groups, settings, test‑specific indicators tested, reference standards or comparison tests used, treatment strategies based on the test results, and outcome measures.