5 Outcomes

The Diagnostics Advisory Committee considered evidence from a number of sources. Full details are in the project documents for this guidance.

How outcomes were assessed

5.1

The External Assessment Group did a systematic review of the evidence on the clinical effectiveness of the use of procalcitonin testing with standard clinical practice to guide antibiotic therapy for:

  • patients with confirmed or highly suspected sepsis in intensive care settings

  • people presenting to the emergency department with suspected bacterial infection.

5.2

Studies were included in the review if they contained information on the following:

  • Adults or children with confirmed or highly suspected sepsis, in whom antibiotic therapy is indicated, who are being treated in intensive care units; or, adults or children presenting to the emergency department with suspected bacterial infection.

  • Treatment decisions based on standard clinical practice with laboratory procalcitonin testing (using any of the 5 tests described in sections 4.1 to 4.5) compared with treatment decisions based on standard clinical practice without procalcitonin testing.

  • At least 1 of the following outcomes:

    • antibiotic exposure (initiation or duration of antibiotic therapy)

    • resource use (number of hospital admissions, length of hospital or intensive care unit stay, costs)

    • adverse clinical outcomes (such as in‑hospital mortality, condition‑specific outcomes, antibiotic‑related adverse events).

Overview of included studies

5.3

In summary, 18 studies were included in the review; 8 studies were done in intensive care unit settings and 10 studies were done in emergency department settings.

5.4

Most of the included studies measured procalcitonin levels using the BRAHMS PCT Sensitive Kryptor assay. Two studies measured procalcitonin levels using the VIDAS BRAHMS PCT assay. The remaining 4 studies used quantitative procalcitonin assays, but did not specify the assay manufacturer.

5.5

There were 12 studies done in Europe (mainly Switzerland), 3 in China, and 1 in Brazil; no UK studies were identified. There were 2 studies (conference abstracts) that did not specify location.

5.6

The methodological quality of all included studies was appraised using the Cochrane risk of bias tool. Three studies were judged as having a high risk of bias and 1 as having a low risk of bias. All other studies were judged as having an unclear risk of bias because insufficient information was reported to make a judgement on 1 or more bias domains.

Clinical effectiveness in intensive care unit settings

Overview of studies

5.7

There were 8 randomised controlled trials that provided data on the effectiveness of using procalcitonin testing with standard clinical practice to guide antibiotic therapy in intensive care unit settings. All studies were done in adult populations. No studies done in paediatric intensive care unit settings met the inclusion criteria for the review.

5.8

There were 4 studies done in adults with confirmed or highly suspected sepsis, in whom antibiotic therapy was indicated. One study included adults being treated in an intensive care unit for suspected bacterial infection, or who developed sepsis during their stay. Two studies included adults being treated in intensive care unit settings who were considered to be at increased risk of developing sepsis (1 study in adults with acute pancreatitis and 1 study in adults with ventilator‑associated pneumonia). The final study included adults who were being treated for suspected bacterial infections in intensive care unit settings.

5.9

There was 1 study that assessed the effectiveness of using procalcitonin testing with standard clinical practice to guide the start of antibiotic treatment. All other studies assessed the effectiveness of using procalcitonin testing with standard clinical practice to decide when to stop antibiotic treatment.

Antibiotic duration

5.10

There were 4 studies that reported data to allow the calculation of mean difference in the duration of antibiotic therapy between study arms. Two of these studies were done in patients with suspected or confirmed sepsis, 1 was done in patients with acute pancreatitis and 1 was done in patients with suspected bacterial infection and those who developed sepsis while in the intensive care unit. Three of these studies found that the addition of procalcitonin testing to standard clinical practice resulted in a statistically significant reduction in the mean duration of antibiotic therapy (Bouadma et al. 2010; Liu et al. 2013; Qu et al. 2012). The fourth study found that the addition of procalcitonin testing to standard clinical practice was associated with a trend towards reduction in the duration of antibiotic therapy, which was not statistically significant (Nobre et al. 2008). The summary effect estimate showed that the use of procalcitonin testing in addition to standard clinical practice was associated with a statistically significant reduction in the duration of antibiotic therapy (weighted mean difference –3.19 days; 95% confidence interval [CI] −5.44 to −0.95). However, between‑study heterogeneity was high.

5.11

When the meta‑analysis was restricted to the 2 studies done in populations with suspected or confirmed sepsis (Liu et al. 2013; Nobre et al. 2008), the summary effect estimate still showed that the addition of procalcitonin testing to standard clinical practice was associated with a statistically significant reduction in the duration of antibiotic therapy (weighted mean difference −1.20 days; 95% CI −1.33 to −1.07).

5.12

There were 3 further studies that reported median duration of antibiotic therapy. Two of these studies were done in people with suspected or confirmed sepsis, and found that the addition of procalcitonin testing to standard clinical practice had no statistically significant effect on the duration of antibiotic treatment (Annane et al. 2013; Deliberato et al. 2013). The third study was done in adults with ventilator‑associated pneumonia and found that the addition of procalcitonin testing to standard clinical practice was associated with a statistically significant reduction in the median duration of antibiotic therapy from 15 days to 10 days (Stolz et al. 2009).

Duration of hospital stay

5.13

There were 7 studies that reported lengths of hospital stay ranging from 11 to 27 days in the intervention groups (procalcitonin testing plus standard clinical practice) and from 11 to 33 days in the control groups (standard clinical practice alone). Four of these studies reported data to allow the calculation of mean difference in the duration of hospital stay between study arms. Two of these studies found that the addition of procalcitonin testing to standard clinical practice resulted in a statistically significant reduction in the mean duration of hospital stay (Liu et al. 2013; Qu et al. 2012). One study found that the addition of procalcitonin testing to standard clinical practice was associated with a trend towards reduction in the duration of hospital stay, which was not statistically significant (Nobre et al. 2008). The fourth study found that the addition of procalcitonin testing to standard clinical practice did not reduce the duration of hospital stay (Bouadma et al. 2010). The summary effect estimate showed that the use of procalcitonin testing with standard clinical practice was associated with a statistically significant reduction in the duration of hospital stay (weighted mean difference −3.85 days; 95% CI −6.78 to −0.92). However, between‑study heterogeneity was high.

5.14

Only 2 of the 4 studies were done in populations with suspected or confirmed sepsis (Liu et al. 2013; Nobre et al. 2008). When the meta‑analysis was restricted to the 2 studies, the summary effect estimate showed that the addition of procalcitonin testing to standard clinical practice was associated with a greater reduction in duration of hospital stay (weighted mean difference –4.32 days; 95% CI –6.50 to –2.14).

5.15

There were 3 further studies that reported median duration of hospital stay. Of these, 2 were done in people with suspected or confirmed sepsis and 1 was done in people with ventilator‑associated pneumonia. All found that the addition of a procalcitonin algorithm had no statistically significant effect on the duration of hospital stay (Annane et al. 2013; Deliberato et al. 2013; Stolz et al. 2009).

Duration of intensive care unit stay

5.16

There were 6 studies that reported lengths of intensive care unit stay ranging from 3.5 to 22 days in the intervention groups (procalcitonin testing plus standard clinical practice) and from 3 to 23 days in the control groups (standard clinical practice alone). Four of these studies reported data to allow the calculation of mean difference in the duration of intensive care unit stay between study arms. Two of these studies found that the addition of procalcitonin testing to standard clinical practice resulted in a statistically significant reduction in the mean duration of intensive care unit stay (Liu et al. 2013; Qu et al. 2012). One study found that the addition of procalcitonin testing to standard clinical practice was associated with a trend towards reduction in the duration of intensive care unit stay, which was not statistically significant (Nobre et al. 2008). The fourth study found that the addition of procalcitonin testing to standard clinical practice did not reduce the duration of intensive care unit stay (Bouadma et al. 2010). The summary effect estimate showed that the addition of procalcitonin testing to standard clinical practice was associated with a trend towards decreased duration of intensive care unit stay, which did not reach statistical significance (weighted mean difference –2.03 days; 95% CI –4.19 to 0.13). However, between‑study heterogeneity was high.

5.17

As with previous outcomes, only 2 of the 4 studies were done in populations with suspected or confirmed sepsis (Liu et al. 2013; Nobre et al. 2008). When the meta‑analysis was restricted to the 2 studies, the summary effect estimate showed that the addition of procalcitonin testing to standard clinical practice was associated with a statistically significant reduction in the duration of intensive care unit stay (weighted mean difference –2.31 days; 95% CI –3.97 to –0.65).

5.18

There were 2 further studies that reported median duration of intensive care unit stay. Both of these studies were done in people with suspected or confirmed sepsis and both found that the addition of procalcitonin testing to standard clinical practice had no statistically significant effect on the duration of intensive care unit stay (Annane et al. 2013; Deliberato et al. 2013).

Adverse clinical outcomes

5.19

There were 5 studies that reported 28‑day all‑cause mortality (Bouadma et al. 2010; Liu et al. 2013; Nobre et al. 2008; Qu et al. 2012; Stolz et al. 2009). All found no statistically significant difference in mortality rates between patients in the intervention group (procalcitonin testing plus standard clinical practice) and those in the control group (standard clinical practice alone). The summary relative risk was 0.98 (95% CI 0.76 to 1.27). This finding was consistent when the meta‑analysis was restricted to studies done in people with suspected or confirmed sepsis (relative risk 1.07; 95% CI 0.54 to 2.12).

5.20

There was 1 study that reported mortality at 60 days and found no statistically significant difference between the intervention and control groups (relative risk 1.15; 95% CI 0.89 to 1.48; Bouadma et al. 2010). One further study reported mortality at 5 days and found no statistically significant difference between the intervention and control groups (relative risk 1.0; 95% CI 0.25 to 4.04; Annane et al. 2013).

5.21

There were 3 studies that reported intensive care unit mortality. All found no statistically significant difference in the intensive care unit mortality rate between the intervention and control groups (Annane et al. 2013; Deliberato et al. 2013; Layios et al. 2012). The summary relative risk was 0.87 (95% CI 0.55 to 1.37). This finding was consistent when the meta‑analysis was restricted to studies done in people with suspected or confirmed sepsis (relative risk 0.59; 95% CI 0.27 to 1.28).

5.22

There were 4 studies that reported rates of infection relapse or recurrence. All found no statistically significant difference in the infection relapse or recurrence rate between the intervention and control groups (Bouadma et al. 2010; Deliberato et al. 2013; Liu et al. 2013; Nobre et al. 2008). The summary relative risk was 1.37 (95% CI 0.77 to 2.44). This finding was consistent when the meta‑analysis was restricted to the 3 studies done in people with suspected or confirmed sepsis (relative risk 1.89; 95% CI 0.47 to 7.59).

5.23

A variety of other general and disease‑specific adverse clinical outcomes were reported by 1 or more studies. These included multi‑drug‑resistant infection, sepsis‑related mortality, multiple organ dysfunction syndrome, ventilator‑associated pneumonia‑related clinical deterioration, duration of mechanical ventilation, and Sequential Organ Failure Assessment score at various time points. No study reported a statistically significant difference between the intervention and control groups for any adverse clinical outcome assessed. None of the studies reported antibiotic‑related adverse events.

Clinical effectiveness in emergency department settings

Overview of studies

5.24

There were 10 randomised controlled trials that provided data on the effectiveness of using procalcitonin testing with standard clinical practice to guide antibiotic therapy in emergency department settings. Two studies were done in children and the rest in adults. Most studies were done in people with respiratory presentations.

5.25

Of the adult studies, 2 were done in people with lower respiratory tract infection, 3 were done in people with community‑acquired pneumonia, 1 included people with chronic obstructive pulmonary disease exacerbations, 1 included people with suspected asthma exacerbations, and 1 was done in people with urinary tract infection. Of studies done in children, 1 included children with lower respiratory tract infection and 1 included children with community‑acquired pneumonia.

Antibiotic initiation

5.26

There were 7 studies, done in adults, which assessed the effectiveness of using procalcitonin testing with standard clinical practice to guide the start of antibiotic treatment. All of these studies found that the addition of procalcitonin testing to standard clinical practice was associated with a reduction in antibiotic use (relative risk 0.77; 95% CI 0.68 to 0.87; Christ‑Crain et al. 2004; Christ‑Crain et al. 2006; Roh et al. 2010; Roh et al. 2013; Schuetz et al. 2009; Stolz et al. 2007; Tang et al. 2013).

5.27

There were 2 studies done in children that reported contradictory results for the proportion of patients in the intervention and control groups who had antibiotic treatment. The study done in children with community‑acquired pneumonia found that the addition of procalcitonin testing to standard clinical practice to decide whether to start antibiotic treatment was associated with a statistically significant reduction in antibiotic use (relative risk 0.85; 95% CI 0.79 to 0.91; Esposito et al. 2011). Subgroup analyses showed that procalcitonin testing was associated with a greater reduction in antibiotic use for children with mild community‑acquired pneumonia (relative risk 0.69; 95% CI 0.59 to 0.80) than for children with severe community‑acquired pneumonia (relative risk 0.96; 95% CI 0.92 to 1.01).

5.28

The study done in children with lower respiratory tract infection (including community‑acquired pneumonia and non‑community‑acquired pneumonia) reported a trend towards increased antibiotic use when procalcitonin test results were added to standard clinical practice (relative risk 1.12; 95% CI 0.94 to 1.35; Baer et al. 2013). Subgroup analyses showed that for children presenting with non‑community‑acquired pneumonia, the addition of procalcitonin testing to standard clinical practice was associated with a statistically significant increase in antibiotic use (relative risk 2.71; 95% CI 1.46 to 5.01). But for children presenting with community‑acquired pneumonia the addition of procalcitonin testing was associated with a trend towards reduction in antibiotic use (relative risk 0.92; 95% CI 0.79 to 1.08). When data from the 2 studies on children presenting with community‑acquired pneumonia were combined the summary relative risk was 0.86 (95% CI 0.80 to 0.93).

Antibiotic duration

5.29

There were 2 studies done in adults that reported data to allow the calculation of mean difference in the duration of antibiotic therapy between study arms. Both of these studies found that the addition of procalcitonin testing to standard clinical practice resulted in a statistically significant reduction in the mean duration of antibiotic therapy (Christ‑Crain et al. 2004; Christ‑Crain et al. 2006). The summary effect estimate showed that the addition of procalcitonin testing to standard clinical practice was associated with reduction in the duration of antibiotic therapy, which did not reach statistical significance (weighted mean difference –4.49 days; 95% CI –9.59 to 0.61). However, these studies included patients who did not have antibiotics in their estimates of mean duration. Therefore an additional meta‑analysis was done, excluding patients who did not have antibiotic treatment. The summary effect estimate for patients who had antibiotic treatment (that is, weighted mean difference conditional on having antibiotics) was 1.48 days (95% CI –13.64 to 16.59).

5.30

There were 4 studies, done in adults, which reported median duration of antibiotic therapy, or mean with no estimate of variance. The results of these studies were consistent with the 2 studies included in the meta‑analysis, showing that the addition of procalcitonin testing to standard clinical practice was associated with a reduction in the duration of antibiotic therapy (Drozdov et al. 2014; Roh et al. 2010; Roh et al. 2013; Schuetz et al. 2009).

5.31

There was 1 study, done in children, which reported data on the duration of antibiotic therapy. This study found the addition of procalcitonin testing to standard clinical practice was associated with a statistically significant reduction in the duration of antibiotic therapy (mean difference –1.8 days; 95% CI –3.1 to –0.5; Baer et al. 2013).

Duration of hospital stay

5.32

There were 6 included studies in adults that reported duration of hospital stay ranging from 8 to 12 days in the intervention groups (procalcitonin testing plus standard clinical practice) and from 8 to 16 days in the control groups (standard clinical practice alone). Two of these studies reported data to allow the calculation of mean difference in the duration of hospital stay between study arms. Neither study found a statistically significant difference between groups (Christ‑Crain et al. 2004; Christ‑Crain et al. 2006). The summary effect estimate showed that addition of procalcitonin testing to standard clinical practice was associated with a non‑significant trend towards reduction in the duration of hospital stay (weighted mean difference –0.80 days; 95% CI –2.37 to 0.78).

5.33

There were 4 studies done in adults that reported median duration of hospital stay, or mean with no estimate of variance. Two of these studies, both done in people with community‑acquired pneumonia, reported results showing that addition of procalcitonin testing to standard clinical practice was associated with a reduction in the duration of hospital stay (mean duration 9.2 days and 14.6 days [Roh et al. 2010]; mean duration 14.6 days and 16.0 days [Roh et al. 2013]). The remaining 2 studies, (1 done in people with lower respiratory tract infection and 1 done in people with chronic obstructive pulmonary disease exacerbations) found that use of a procalcitonin algorithm did not affect the median duration of hospital stay (Schuetz et al. 2009; Stolz et al. 2007).

5.34

Both studies done in children reported data to allow the calculation of mean difference in the duration of hospital stay between study arms. These studies reported lengths of hospital stay ranging from 2.5 to 5.0 days in the intervention groups (procalcitonin testing plus standard clinical practice) and from 2.3 to 5.9 days in the control groups (standard clinical practice alone). When data on children presenting with community‑acquired pneumonia were combined, the summary effect estimate showed that the addition of procalcitonin testing to standard clinical practice was associated with a small reduction in the duration of hospital stay (weighted mean difference –0.74 days; 95% CI –1.17 to –0.31; Baer et al. 2013; Esposito et al. 2011).

5.35

One study reported data on duration of intensive care unit stay. This study was done in adults with chronic obstructive pulmonary disease exacerbations. It reported no statistically significant difference in the mean duration of intensive care unit stay between the study groups (mean difference –0.40; 95% CI –1.06 to 0.26; Stolz et al. 2007).

Adverse clinical outcomes

5.36

There were 2 studies in adults that reported hospital re‑admission rates. One study was in people with acute asthma exacerbations and the other was in people with urinary tract infection. Both studies found no statistically significant difference in re‑admission rates between patients in the intervention group (decision to stop antibiotics based on procalcitonin test result plus clinical judgement) and those in the control group (decision to stop antibiotics based on clinical judgement alone; Drozdov et al. 2014; Tang et al. 2013).

5.37

There were 2 studies (1 in adults with acute asthma exacerbations and 1 in adults with chronic obstructive pulmonary disease exacerbations) that reported the rate of second emergency department visits. Both found no statistically significant difference between the intervention and control groups (Stolz et al. 2007; Tang et al. 2013).

5.38

There were 6 studies done in adults that reported all‑cause mortality at various time points, ranging from 14 days to 6 months (Christ‑Crain et al. 2004; Christ‑Crain et al. 2006; Drozdov et al. 2014; Roh et al. 2010; Roh et al. 2013; Schuetz et al. 2009; Stolz et al. 2007; Tang et al. 2013). All studies reported no statistically significant difference in mortality rates between the intervention and control groups. The summary relative risk was 0.95 (95% CI 0.71 to 1.27). This finding was consistent when the meta‑analysis was restricted to the 2 studies reporting mortality at 6 months (relative risk 0.85; 95% CI 0.46 to 1.59). Neither of the 2 studies done in children reported mortality data.

5.39

There were 4 studies done in adults that reported data on rates of admission to the intensive care unit. All studies found no statistically significant difference in intensive care unit admissions between the intervention and control groups (summary relative risk 0.79; 95% CI 0.59 to 1.05; Stolz et al. 2007; Christ‑Crain et al. 2004; Christ‑Crain et al. 2006; Schuetz et al. 2009). Neither of the 2 studies done in children reported any information on intensive care unit admissions.

5.40

There were 2 studies, done in adults, which reported inconsistent results for rates of infection relapse or recurrence. One study, in adults with urinary tract infection, found no statistically significant difference in relapse or recurrence rates between the intervention and control groups (Drozdov et al. 2014). The second study, in adults with lower respiratory tract infection, found that the addition of procalcitonin testing to standard clinical practice was associated with a statistically significant reduction in infection relapse or recurrence rates (relative risk 0.57; 95% CI 0.36 to 0.92; Schuetz et al. 2009). One study, done in children with community‑acquired pneumonia, reported very low rates of infection relapse or recurrence and a non‑significant trend towards lower rates in the intervention group (relative risk 0.23; 95% CI 0.04 to 1.34; Esposito et al. 2011).

5.41

There was 1 study, done in adults with lower respiratory tract infection, which reported numbers of patients having antibiotic‑related adverse events. This study found that the addition of procalcitonin testing to standard clinical practice was associated with a reduction in antibiotic‑related adverse events (relative risk 0.71; 95% CI 0.58 to 0.86).

5.42

Both studies, done in children, reported the numbers having antibiotic‑related adverse events (Baer et al. 2013; Esposito et al. 2011). When data on children with community‑acquired pneumonia were combined, results showed that the addition of procalcitonin testing to standard clinical practice was associated with a non‑significant reduction in antibiotic‑related adverse events (relative risk 0.37; 95% CI 0.04 to 3.49). This finding was consistent when data for all children in both studies were included in the meta‑analysis (summary relative risk 0.40; 95% CI 0.06 to 2.78).

5.43

A variety of other general and disease‑specific adverse clinical outcomes were reported by 1 or more studies. These included composite adverse outcome measures, need for steroids, need for mechanical ventilation and complications from pneumonia. No study reported a statistically significant difference between the intervention and control groups for any adverse clinical outcome.

Costs and cost effectiveness

Systematic review of cost‑effectiveness evidence

5.44

The External Assessment Group did a search to identify existing economic evaluations of people with sepsis or bacterial infection having care in emergency departments or intensive care units. Two studies were considered eligible for inclusion in the systematic review.

5.45

There was 1 study (Michaelidis et al. 2013) that considered procalcitonin testing in 2 scenarios. The first analysis was based on adults with an acute respiratory tract infection presenting to an outpatient clinic and judged by their physicians to need an antibiotic prescription. The second analysis was based on adults with an acute respiratory tract infection presenting to an outpatient clinic before any decision to start antibiotic therapy. Procalcitonin‑guided antibiotic therapy was both more costly and more effective than standard care alone. The incremental cost‑effectiveness ratios (ICERs) were $118,828 and $575,249 per quality‑adjusted life year (QALY) gained for the first and second analyses respectively.

5.46

The second study (Smith et al. 2013) considered the cost effectiveness of procalcitonin‑guided antibiotic therapy in adults with community‑acquired pneumonia in a hospital setting. Procalcitonin‑guided antibiotic therapy was both more costly and more effective compared with standard care alone. For patients with low‑risk community‑acquired pneumonia, procalcitonin‑guided antibiotic initiation is likely to be cost effective if the maximum acceptable ICER was $90,000 per QALY gained. For the same patients, procalcitonin‑guided antibiotic initiation and monitoring is likely to be cost effective if the maximum acceptable ICER was $40,000 per QALY gained. For patients with high‑risk community‑acquired pneumonia, procalcitonin‑guided antibiotic initiation and monitoring is likely to be cost effective if the maximum acceptable ICER was $170,000 per QALY gained.

Economic analysis

5.47

The External Assessment Group developed a de novo economic model designed to assess the cost effectiveness of the addition of procalcitonin testing to standard clinical practice compared with standard clinical practice alone for:

  • adults with confirmed or highly‑suspected sepsis in an intensive care unit setting

  • adults with suspected bacterial infection presenting to the emergency department

  • children with suspected bacterial infection presenting to the emergency department.

5.48

Children with confirmed or highly‑suspected sepsis in an intensive care unit setting were not considered because there was insufficient clinical evidence.

Model structure
5.49

There were 2 decision tree models constructed:

  • 1 in the intensive care unit setting that incorporated discontinuation of antibiotics only

  • 1 in the emergency department setting that incorporated both starting and stopping antibiotics.

5.50

The structures of both models started with a decision node that denotes the use of procalcitonin testing in addition to standard clinical practice or standard clinical practice alone. The key endpoints were:

  • alive with antibiotic‑related complications

  • alive without antibiotic‑related complications

  • death.

5.51

The time horizon was 6 months (183 days), divided into an initial short‑term (28 days) phase and a subsequent phase of 155 days. This time horizon was adopted to be consistent with the outcomes reported in the studies included in the clinical‑effectiveness systematic review.

5.52

A 'lower clinical extreme' and a 'higher clinical extreme' were specified for each population and setting. For these 'clinical extremes', different baseline values were used for: mortality; duration of antibiotic therapy; probability of starting antibiotic treatment (emergency department setting only); length of hospital stay; and length of stay in an intensive care unit. The same relative risks and mean difference estimates were applied for both clinical extremes.

Model inputs − all‑cause mortality and adverse events
5.53

The baseline probabilities and relative risks for all‑cause mortality were taken from the systematic review of clinical effectiveness. Results from meta‑analysis were used where possible. When a meta‑analysis result was not available, data from the most plausible single source were chosen. Mortality rates for children presenting to the emergency department with suspected bacterial infection were not available from the systematic review. Therefore national background mortality rates were assumed, based on the expert opinion that mortality rates for children presenting to the emergency department are close to 0.

5.54

Antibiotic‑related adverse events were incorporated in the economic model through the time on antibiotic treatment, using a disutility for having antibiotic treatment. Disease‑specific complications were not included in the economic model and therefore were assumed to be equal for the intervention and control groups.

Health state utilities
5.55

Searches were undertaken to find relevant utility value studies on adults and children with sepsis or bacterial infection presenting to, or being treated at, emergency departments and intensive care units.

5.56

For adults being treated in the intensive care unit, a utility of 0.53 was used for the initial short‑term phase, and a utility of 0.68 was used for the subsequent phase (Drabinski et al. 2001).

5.57

No utility values were found for adults presenting to the emergency department with suspected infection. Therefore, utility values for adults presenting to primary care with lower respiratory tract infection were used; 0.70 for the initial short‑term phase and 0.86 for the subsequent phase (Oppong et al. 2013).

5.58

For children presenting to the emergency department, a constant base utility of 0.99 was assumed (utility for local infection; Bennett et al. 2000).

5.59

To incorporate antibiotic‑related adverse events in adults being treated in the intensive care unit, a disutility of 0.046 for being on antibiotic treatment was used (Oppong et al. 2013).

Costs and resource use
5.60

Resource use consisted of duration of hospital stay (days), intensive care unit stay (days) and antibiotic treatment duration (days). The estimates were retrieved from studies identified in the systematic review of clinical effectiveness. Results from meta‑analysis were used where possible. When a meta‑analysis result was not available, data from the most plausible single source were chosen. Where necessary, data were identified through consultation with experts for unpublished data.

5.61

Cost data were drawn from routine NHS sources (for example, NHS reference costs and British national formulary) and information provided by manufacturers of the procalcitonin tests.

5.62

An average unit price for the procalcitonin test was calculated to be £13.79, based on the list prices of the tests (excluding VAT) and with no discounts assumed. Overhead costs including capital, service and maintenance, and calibration costs were included. These were calculated from the initial capital costs, the lifetime of the assay (assumed to be 5 years) and the average number of tests per day (assumed to be 272). A similar estimation was done for the maintenance and calibration costs.

Base‑case analysis
5.63

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

  • The duration of hospital stay retrieved from the systematic review of clinical effectiveness included days in hospital after infection relapse or recurrence.

  • Relative risks for all‑cause mortality for children presenting to the emergency department were assumed to be equal to those for adults presenting to the emergency department.

  • There was no disutility for the hospital stay.

  • The baseline utility for children presenting to the emergency department was constant over time.

  • The disutility for being on antibiotic treatment was equal for all settings and populations.

  • Procalcitonin testing used for starting antibiotics in the emergency department and stopping antibiotics in the intensive care unit was used to calculate the average number of tests per day.

  • There were no costs associated with antibiotic‑related adverse events.

  • There were no differences in disease‑specific complications between the group with procalcitonin testing and the group without procalcitonin testing.

  • There were no differences in long‑term costs and effects between the group with procalcitonin testing and the group without procalcitonin testing.

5.64

Base‑case analyses indicate that procalcitonin testing with standard clinical practice dominates standard clinical practice alone for all populations, that is, it was both cost saving and more effective.

5.65

The cost savings ranged from £368 for children with suspected bacterial infection presenting to the emergency department (lower clinical extreme) to £3268 for adults with confirmed or highly‑suspected sepsis in an intensive care unit setting (lower clinical extreme).

5.66

The use of procalcitonin testing with standard clinical practice resulted in only a small QALY gain compared with standard clinical practice alone. For adults with suspected bacterial infection presenting to the emergency department this was 0.005 for the lower and higher clinical extremes, and for adults with confirmed or highly‑suspected sepsis in the intensive care unit setting it was 0.001 for both clinical extremes. For children with suspected bacterial infection presenting to the emergency department, the QALY gains were less than 0.001 for both clinical extremes.

5.67

Procalcitonin testing with standard clinical practice always has a higher probability of being cost effective than standard clinical practice alone if the maximum acceptable ICER is between £0 to £60,000 per QALY gained. If the maximum acceptable ICER is £20,000 per QALY gained the probability of procalcitonin testing with standard clinical practice being cost effective compared with standard clinical practice alone is:

  • 85% and 98% respectively for the lower and higher clinical extremes for children with suspected bacterial infection presenting to the emergency department.

  • 88% for adults with suspected bacterial infection presenting to the emergency department (both clinical extremes).

  • 97% and 95% respectively for the lower and higher clinical extremes for adults with confirmed or highly‑suspected sepsis in an intensive care unit setting.

Analysis of alternative scenarios
5.68

The following scenario analyses were performed to assess the impact of assumptions on the estimated outcomes:

  • assume no difference in mortality

  • assume an increased cost of £50 per test

  • assume no overhead costs for the tests

  • alternative utility value for adults in the intensive care unit

  • assume no disutility for being on antibiotic treatment

  • assume no difference in duration of antibiotic treatment

  • assume no difference in hospital stay (including intensive care unit stay)

  • assume lower prices for hospital and intensive care unit stay

  • assume that procalcitonin testing in the emergency department was solely used to start antibiotic treatment (not to stop antibiotic treatment).

5.69

The scenario analyses that assumed no difference in hospital stay had a substantial impact on all populations and settings. Treatment based on procalcitonin testing with standard clinical practice became more costly (incremental costs varied between £7 for adults in the intensive care unit and £25 for children in the emergency department) and remained more effective (QALY gain varied between less than 0.001 for children in the emergency department and 0.007 for adults in the intensive care unit) compared with standard clinical practice alone. For children presenting to the emergency department with suspected bacterial infection, this resulted in ICERs of £287,076 per QALY gained for the lower clinical extreme and £35,219 per QALY gained for the higher clinical extreme. For adults in both settings (and both clinical extremes), ICERs varied between £3390 and £3948 per QALY gained.

5.70

None of the other scenario analyses resulted in substantial changes to the base‑case ICERs, and use of procalcitonin testing with standard clinical practice remained cost effective compared with standard clinical practice alone.

One‑way sensitivity analyses
5.71

One‑way sensitivity analyses were performed for all stochastic input parameters between the 95% confidence intervals.

5.72

The one‑way sensitivity analysis on relative mortality risk for adults with suspected bacterial infection presenting to the emergency department resulted in substantial changes to the base‑case ICERs. Analyses showed that when using the upper bound of the 95% confidence interval (1.590; base‑case value 0.850) procalcitonin testing with standard clinical practice guided treatment was less costly (£772) and less effective (QALY loss 0.025) compared with standard clinical practice alone. This resulted in ICERs of £30,469 per QALY lost (lower clinical extreme) and £30,446 per QALY lost (higher clinical extreme).

5.73

None of the other one‑way sensitivity analyses resulted in substantial changes to the base‑case ICERs, and use of procalcitonin testing with standard clinical practice remained cost effective compared with standard clinical practice alone.