Point-of-care creatinine devices to assess kidney function before CT imaging with intravenous contrast

Results from the StatSensor studies showed wide variation in the size and direction of measurement bias. StatSensor devices can be adjusted to correct for any bias seen, to align the POC creatinine device results with those from local laboratory methods. Only 2 StatSensor studies reported using an adjustment function for measurement bias. Although potentially important measurement bias was found in some studies of i‑STAT and ABL devices, the concordance of results for these devices was generally better than for the StatSensor devices. A smaller number of studies were available on the epoc (1 study) and Piccolo Xpress (4 studies) devices.

There were 3 studies comparing different types of POC creatinine devices. Of these, 2 studies compared StatSensor, i‑STAT and ABL800 FLEX. Both studies found that the ABL800 FLEX had the strongest agreement with laboratory serum creatinine, then i‑STAT and then StatSensor. There was 1 study comparing an ABL827 device with an i‑STAT device. It concluded that creatinine results from both devices correlated well with laboratory serum creatinine.

In some studies, measurement bias increased at higher creatinine levels (lower eGFR). This could affect care decisions for people at higher risk of kidney damage.

There were 12 studies reporting diagnostic accuracy data on eGFR thresholds. Studies were of different devices, with some studies assessing more than one device:

• 7 i‑STAT studies

• 7 StatSensor studies

• 3 studies included a Radiometer POC device (ABL800 or ABL827)

• 2 studies assessed 3 POC devices (ABL, i‑STAT and StatSensor) and

• 1 study looked at 2 devices (ABL and i‑STAT).
  
  There were no studies of ABL90 FLEX PLUS, Dri‑chem NX500, epoc Blood Analysis System and Piccolo Xpress. The eGFR equations used in the studies varied, with only 3 studies using chronic kidney disease epidemiology (CKD-EPI). There were 3 StatSensor and 2 i‑STAT studies that used an adjustment function to correct for any measurement bias between the POC creatinine test results and laboratory test results from the study sample. Adjusted and unadjusted results were reported in all 3 StatSensor studies, but only adjusted results were presented in the 2 i‑STAT studies. Most studies used an enzymatic method as the laboratory reference, but the Jaffe method was used in 2 studies and the reference method was not reported in 1 study.

There were 6 studies at low risk across all risk of bias areas, including 2 studies of ABL800, 3 studies of i‑STAT and 3 studies of StatSensor. The other 6 studies had at least one domain at unclear or high risk of bias. Risks of bias related to:

• how the adjustment function to correct for measurement bias was applied

• patient selection

• the use of a different modification of the diet in renal disease (MDRD) eGFR equation between the POC creatinine test and laboratory reference test

• the use of a Jaffe method for the laboratory reference test (compared with an enzymatic method for the POC creatinine test).

There were low concerns about the applicability of results across all domains for only 2 studies, including 1 study of ABL800, i‑STAT and StatSensor (Snaith et al. 2018), and 1 study of i‑STAT (Snaith et al. 2019). The most common concern was the use of eGFR threshold; 3 studies used an eGFR cut-off of 60 ml/min/1.73 m² or above. Several studies included disease-specific populations, therefore their applicability to a broader population of outpatients referred for CT scan without a recent eGFR may be limited.

The EAG quantitatively analysed the study results. The probabilities of being in each eGFR category were calculated from the number of people in each category reported by all included studies (regardless of the device assessed). The pooled probabilities of being in each of the 4 categories are in table 2. Most studies only included a few people in category 1 (eGFR less than 30 ml/min/1.73 m²) and more people in higher eGFR categories.

The pooled probabilities of having a classification by a POC creatinine device in each eGFR category (k) and in each laboratory-defined eGFR category (j) are given in table 3. The i‑STAT and ABL devices have higher median probabilities of correct classification in each of the 3 lowest categories (p[1,1], p[2,2], p[3,3]) compared with the StatSensor. StatSensor was particularly poor at correctly classifying category 3 (eGFR 45 to 59 ml/min/1.73 m²). However, there is considerable uncertainty in these probabilities for all devices.

Additional analyses were done to assess the effect of removing studies with limited applicability to clinical practice in the NHS. The pooled probabilities from these analyses were used in scenario analyses in the economic model:

• StatSensor devices allow a user-specified adjustment if systematic measurement bias is identified. An additional analysis including the adjusted data reported by Korpi-Steiner et al. (2009) and Shephard et al. (2010) was done. The Inoue et al. (2017) study was not included in this analysis because the reported adjustment could not be replicated in NHS practice.

• Only 2 studies used the CKD-EPI equation to calculate eGFR, all others used the MDRD equation. Of these studies, one included StatSensor, i‑STAT and ABL800 FLEX devices (Snaith et al. 2018) and the other only included an i‑STAT device (Snaith et al. 2019). An additional analysis using only the data in these 2 studies was done.

There were 6 studies reporting a relevant outcome after using a POC creatinine device. The results showed variation in practice in both the proportions of patients who do not have a recent eGFR result and in the management decisions when a POC creatinine device shows an abnormal eGFR. For example, the proportion of people offered scans with or without contrast, or offered a reduced dose of contrast. Also, many of the studies were done several years ago so the value of their results is limited because eGFR thresholds for defining an abnormal result have decreased over time. No data were available on clinical outcomes such as need for renal replacement therapy or hospital admissions.

The model assessed a cohort of outpatients presenting for a non-emergency contrast-enhanced CT scan without a recent eGFR measurement. Costs were presented from the perspective of the NHS and personal social services and were reported in UK pounds at 2018 prices. Outcomes after the first year were discounted at a rate of 3.5% per year. Most costs happened in the first year and were therefore not discounted.

The model used a decision tree cohort approach to estimate the costs and health outcomes of the different testing and treatment strategies. The model captured:

• true eGFR status (less than 30 ml/min/1.73 m² or 30 ml/min/1.73 m² and above)

• how eGFR status is classified by different testing strategies, using the eGFR cut-off value of 30 ml/min/1.73 m² and probabilities conditional on true eGFR status

• any actions to reduce post-contrast acute kidney injury (PC‑AKI) risk in patients with eGFR below the cut-off value (correct or incorrect eGFR)

• the subsequent risk of PC‑AKI (depends on eGFR status and any actions to reduce PC‑AKI risk)

• the risk of renal replacement therapy (depends on whether a patient had a PC‑AKI).

The model assessed 6 strategies to identify and manage treatment for patients with an eGFR less than 30 ml/min/1.73 m²:

• laboratory testing only

• risk factor screening with POC creatinine testing

• risk factor screening with laboratory testing

• risk factor screening with POC creatinine testing and laboratory testing

• POC creatinine testing only

• POC creatinine testing with laboratory testing.
  
  For strategies with sequential tests, only people who have a test result of eGFR less than 30 ml/min/1.73 m² would go on to receive the next test in the sequence.

For each strategy that includes POC creatinine testing, the model considered separate strategies for each of the POC devices, to give 14 alternative testing strategies. The 3 devices considered in the model were i‑STAT Alinity, ABL800 FLEX and StatSensor because only these had sufficient data available to calculate classification probabilities.

Population characteristics, including the probability of a patient being in 1 of 4 eGFR categories, are presented in table 4. The proportion of people attending a CT scan appointment without a recent eGFR measurement was used to estimate the throughput of POC creatinine devices.

The diagnostic accuracy data used for each of the tests included in the model are in table 5. The cut-off used to define a positive result is eGFR less than 30 ml/min/1.73 m2. The sensitivity of the tests is equivalent to the probability that a person with eGFR less than 30 ml/min/1.73 m2 is correctly categorised as having eGFR less than 30 ml/min/1.73 m2. The specificity of the POC creatinine devices was calculated by combining information on the distribution of population eGFR with the probability of having a classification of eGFR less than 30 ml/min/1.73 m2 for a given true eGFR category (a weighted average).

In the base-case analysis, an odds ratio of 0.97 (95% confidence interval [CI] 0.52 to 1.9) for the effect of preventative intravenous hydration was used for patients with an eGFR below 30 ml/min/1.73 m2 (Ahmed et al. 2018). It was assumed there would be no effect of intravenous hydration on risk for patients with an eGFR above 30 ml/min/1.73 m2 (AMACING trial). A scenario analysis was done using the lower bound of the odds ratio (0.52), implying a greater protective effect of intravenous hydration compared with the base-case analysis.

A fixed effects meta-analysis of 3 studies (Hinson et al. 2017; Davenport et al. 2013; McDonald et al. 2014) suggested no effect of contrast on PC‑AKI risk (odds ratio [OR] 0.98; 95% CI 0.88 to 1.08). It was therefore assumed in the base case that there was no effect of contrast on the risk of PC‑AKI. A scenario analysis exploring a greater risk of PC‑AKI in people with an eGFR of less than 30 ml/min/1.73 m² was done.

The probability of having AKI after contrast for people with an eGFR of less than 30 ml/min/1.73 m2 or 30 ml/min/1.73 m2 and above depending on whether they had intravenous hydration or not is shown in table 6.

After having a CT scan, the probability that people who did not develop AKI after contrast needed renal replacement therapy was 0.014 and for people who did develop AKI after contrast was 0.111.

The model did not consider the effect of a delay in the planned CT scan on patient outcomes because of any change in their underlying condition during the waiting period.

It was assumed that 94.5% of people were alive 6 months after they had the CT scan, based on data reported in Park et al. (2016). The health-related quality-of-life data used in the base case are shown in table 7. No disutility from PC‑AKI or intravenous hydration was included in the model.

Costs were calculated for each POC creatinine test, laboratory test, CT scan, intravenous hydration and for associated adverse events. The costs used in the model are shown in table 8. It was estimated that 92.6 patients per month would have a POC creatinine test. Risk factor screening before a POC creatinine test resulted in an estimated 32.6 patients per month having a POC test.

The following assumptions were applied in the base-case analysis:

• The laboratory test would have perfect diagnostic accuracy (100% sensitivity and specificity).

• Risk factor screening in the model would be done with a generic risk factor questionnaire.

• All patients having a laboratory test would have their CT scan cancelled and rebooked.

• A positive test result at the last step of the testing sequence resulted in the scan being cancelled and rebooked with intravenous hydration and contrast-enhanced CT scan.

• Adverse events from intravenous hydration were associated with costs but no health-related quality-of-life loss.

• Mortality in the model was the same for all patients regardless of PC‑AKI status.

• Mortality was independent of eGFR levels and PC‑AKI.

• Renal replacement therapy consisted of haemodialysis.

Deterministic and probabilistic results were presented as net monetary benefit and net health benefit using a maximum acceptable incremental cost-effectiveness ratio (ICER) of £20,000 per quality-adjusted life year (QALY) gained. Incremental net benefit was calculated for each strategy compared with laboratory testing. A fully incremental analysis was also done, but because the incremental cost and QALY differences between the strategies were so small, the ICERs are of limited use. This is because they are very sensitive to extremely small differences in the QALYs. If pairwise ICERs had been calculated, all strategies that include POC creatinine devices would cost less and be less effective than the strategy of laboratory testing for all. Full results of the base case are shown in tables 9 and 10. In general:

• Strategies that combine risk factor screening with POC creatinine testing and laboratory testing result in higher net benefit than other types of strategies, because they have a high positive predictive value. This avoids unnecessarily offering people who have false positive results intravenous hydration, which is associated with costs including cancelling and rebooking CT scans, giving intravenous hydration, treating intravenous hydration adverse events and patient follow up.

• Strategies that combine risk factor screening with POC creatinine testing, without confirmatory laboratory testing, are the next highest ranking. These have lower overall specificity and give more false positive results, which are associated with increased costs from unnecessary management for patients whose results were misclassified as eGFR less than 30 ml/min/1.73 m² (cancelling and rebooking CT scans, giving intravenous hydration, treating intravenous hydration adverse events and patient follow up).

• Strategies with POC creatinine testing that do not use risk factor screening have lower average net benefit than POC creatinine test strategies that do, because of the higher costs of testing when all patients have POC creatinine testing.

• The strategies using POC creatinine in isolation are the lowest ranking strategies involving POC creatinine testing, because they misclassify more patients' results as false positives and all patients incur the cost of POC testing.

• Laboratory testing alone and risk factor screening then laboratory testing are the lowest ranking strategies. Although they have the highest QALY gains because they give no false positives or false negatives, they are associated with the highest costs, because of cancellation, rebooking and managing treatment for people who test positive.

The strategy with the highest incremental net benefit was strategy 6 (risk factor screening plus i‑STAT Alinity plus laboratory testing). In the probabilistic sensitivity analysis, this strategy had the highest probability of being the most cost effective (79.3% for maximum acceptable ICERs of £20,000 and £30,000 per QALY gained). It was also the least costly of all strategies compared, but gave fewer QALYs than most other strategies. The corresponding strategy with StatSensor, strategy 8, only had a marginally smaller average incremental net benefit (£87.11 compared with £87.42 for strategy 6). In the probabilistic sensitivity analysis, the probability of this strategy being the most cost effective at maximum acceptable ICERs of £20,000 and £30,000 per QALY gained was 20.7%. Although ABL800 FLEX has the best diagnostic accuracy, strategies including testing with ABL800 FLEX have consistently lower net benefit than corresponding strategies with i‑STAT Alinity and StatSensor because of the higher costs of testing with this device.

The fully incremental ICER analysis showed that most strategies were dominated or extendedly dominated by strategy 6. Strategy 5 (risk factor screening plus laboratory testing) had an ICER of £3.61 billion per QALY gained compared with strategy 6.

Several scenario analyses were explored; results from most of the analyses were robust to changes in the assumptions. Some analyses caused strategy 8 (risk factor screening plus StatSensor plus laboratory testing) to become more cost effective than strategy 6 (risk factor screening plus i‑STAT Alinity plus laboratory testing). This was generally because of changes to the assumptions about the diagnostic accuracy and the costs of the POC creatinine tests. The scenario analysis in which there were no delays to CT scanning from laboratory testing with or without intravenous hydration resulted in strategy 5 (risk factor screening plus laboratory testing) and strategy 1 (laboratory testing) being more cost effective than strategies involving POC creatinine devices.

The base-case analysis was also replicated, adding 2 new strategies:

• a 'no testing' strategy when all patients had a contrast-enhanced CT scan without testing for risk of PC‑AKI

• a 'no testing' strategy combined with a greater reduction in risk of PC‑AKI from intravenous hydration.
  
  Both these strategies were associated with higher net benefit than other strategies included in the base-case analysis. That is, the no testing strategies were both less effective and cheaper than all other strategies.

An additional scenario analysis was done to consider the effect on the results if there was a higher risk of PC‑AKI than in the base case; the risk from contrast agent was increased and the protective effect of intravenous hydration was increased to give an absolute risk difference with and without hydration of 10.3%. The results of this analysis were consistent with the base case.