6.1.1. Type of evaluation

A cost-effectiveness analysis was conducted to compare combined ESR + CRP testing with either ESR or CRP alone. The cost-effectiveness outcome evaluated depended on whether a positive combined test was based on both ESR and CRP being positive (ESR + CRP) or either ESR or CRP being positive (ESR/CRP).

If a combined positive test was based on both tests being positive (ESR + CRP), it was assumed that the benefit of a combined test compared with a single test would be to increase specificity and reduce the number of false-positive results. Therefore, the cost-effectiveness outcome for this definition was the incremental cost per false-positive avoided.

If a combined positive test was based on either test being positive (ESR/CRP), it was assumed that the benefit of a combined test compared with a single test would be to increase sensitivity and reduce the number of false-negative results. Therefore, the cost-effectiveness outcome for this definition was the incremental cost per false-negative avoided.

6.1.2. Target population

There were a number of different target patient populations for this economic analysis. These were based on the patient populations in which comparative diagnostic accuracy data were found in the clinical review of this report. Specifically, the four target populations of the analyses were:

• patients suspected of periprosthetic infections
• children suspected of having orthopedic infections
• patients suspected of having inflammatory bowel diseases
• patients suspected of having giant cell arteritis.

As described in the clinical evaluation, because of the methodological differences in the two studies that evaluated giant cell arteritis,^19,28 pooling of diagnostic accuracy data was not completed. Therefore, two cost-effectiveness analyses were conducted for the giant cell arteritis population: one used the diagnostic accuracy and prevalence data from Kermani et al.,¹⁹ while the other used diagnostic accuracy and prevalence data from Hayreh et al.²⁸

6.1.3. Comparators

The comparators in the evaluation are combined ESR and CRP testing; ESR testing alone; and CRP testing alone. In this analysis, pairwise comparisons of ESR + CRP versus ESR alone and ESR + CRP versus CRP alone were conducted. Because it was not intended to be part of this review, cost-effectiveness comparisons of ESR alone versus CRP alone were not reported.

6.1.4. Perspective

A third-party payer perspective such as that of a provincial ministry of health was undertaken.

6.1.5. Model structure

A graphical representation of the model structure is provided in Figure 1. The model begins with patients being tested for one of the conditions of the four target populations. There is an underlying prevalence of disease that categorizes patients as either having the condition being tested for or not having the condition being tested for. Based on the test results, patients having the condition can either be diagnosed correctly as a true-positive or diagnosed incorrectly as a false-negative. Similarly, patients without the condition are either correctly diagnosed as a true-negative or incorrectly as a false-positive. The diagnostic status (i.e., true-positive, false-negative, true-negative, and false-positive) of patients is dependent on the prevalence of the disease, along with the sensitivity and specificity of the diagnostic test. As indicated in Figure 1, the proportion of false-positives can be calculated as 1 minus condition prevalence multiplied by 1 minus test specificity. Similarly, the proportion of false-negative results is calculated as condition prevalence multiplied by 1 minus the sensitivity of the test.

Figure 1

Graphical Representation of the Model Structure. ESR = erythrocyte sedimentation rate; CRP = C-reactive protein.

6.1.6. Clinical model inputs

As shown in Figure 1, to estimate the diagnosis status of patients for each testing strategy, the prevalence of the condition along with the sensitivity and specificity of each testing strategy are required. The prevalence rates used in the model for each of the four populations are shown in Table 10, and are based on the disease frequencies found in the included studies (see Appendix 11). The prevalence for the periprosthetic infection population was based on the weighted average disease frequency reported in the four studies evaluating this population.^18,21,23,25 For the pediatric orthopedic infections and the inflammatory bowel disease populations, there was only a single study in which disease frequency could be derived. For giant cell arteritis, prevalence rates specific to the studies by Kermani et al.¹⁹ and Hayreh et al.²⁸ were used in the analysis.

Table 10

Prevalence Rates by Population.

The sensitivity and specificity for each test strategy were based on the findings in the clinical review (see Appendix 12 and Table 4). Tables 11 and 12 present the sensitivity and specificity for each test by population. Table 4 shows data when a positive combined test is based on both ESR and CRP being positive. Table 5 shows diagnostic accuracy data when a combined test is defined as either ESR or CRP being positive. For the perioperative infection population, sensitivity and specificity are based on the pooled analysis in the clinical review (see Table 4). For the other populations, sensitivity and specificity were based on single studies (see Appendix 12). For giant cell arteritis, separate analyses were conducted based on data from by Kermani et al.¹⁹ and Hayreh et al.²⁸ Data were available for only two populations when the combined test was based on either test being positive. For the pediatric infection population, Robinson et al.²⁰ informed diagnostic accuracy when the combined positive test was defined as both tests being positive, while Paakkonen et al.²² informed data when the combined positive test was defined as when either test was positive.

Table 11

Sensitivity and Specificity by Patient Population When a Positive Combined Test Assumes Both ESR and CRP Are Positive (ESR + CRP).

Table 12

Sensitivity and Specificity by Patient Population When a Positive Combined Test Assumes Either ESR or CRP is Positive (ESR/CRP).

6.1.7. Resource use and costs

CADTH conducted a survey to estimate the costs of ESR and CRP testing across Canada. At least one contact in each jurisdiction was asked to complete the survey. Contacts were asked the cost of each test, and to indicate the various components included in the test costs, along with utilization data. A summary of the survey results is presented in Table 13. As shown, there was much variation in the reported test costs and included components. For example, the cost of CRP was reported to be $0.67 in New Brunswick, $1.60 in Nunavut, and $10.31 in British Columbia. Much of this variation may have been due to differences in the components included in the cost estimates. For example, only the reagent costs were included in New Brunswick and Nunavut’s CRP costs. Because of this variation, the costs of the tests in the model were assumed to be that reported by British Columbia, where it was reported that all components were included.

Table 13

Results From Survey on Costs and Included Components of ESR and CRP Tests Across Jurisdictions in Canada.

As such, the model assumed the cost of ESR and CRP to be $10.61 and $10.31, respectively. The cost of the combined test was assumed to be $20.92, the sum of the individual ESR and CRP costs.

Another source of data for test costs was provincial benefit schedules, which indicated fees for both ESR and CRP. Table 14 shows costs derived from these fee schedules, along with average fees of $5.83 for ESR and $9.57 for CRP. The cost of combined testing ($15.40) was assumed to be the sum of the individual test costs. These alternative costs were assumed in sensitivity analysis.

Table 14

Costs of ESR and CRP Across Provinces.

6.1.8. Sensitivity analysis

Sensitivity analyses were conducted on all variables that affected the model. Specifically, one-way sensitivity analyses were conducted on the incremental cost of the combined test, the diagnostic accuracy (sensitivity and specificity) of the test strategies, and the prevalence of conditions for the four different populations. The model was run using alternative testing costs based on the average from various provincial fee schedules. For diagnostic accuracy the model was run using the lower and upper confidence intervals for sensitivity and specificity. Similarly, the model was run using the lower and upper confidence intervals for the prevalence of the condition for each population. In addition, prevalence rates for all conditions varied from 0.01 to 0.95, due to the concern that prevalence rates from the reported studies may not reflect those found in clinical practice.

6.2.1. Base-case results

Periprosthetic infection

Tables 15 and 16 present cost-effectiveness results for the periprosthetic infection population when the combined positive test is defined as ESR + CRP and ESR or CRP, respectively. False-positives and misdiagnoses are presented per 100 patients tested. As shown in Table 13, the number of false-positives per 100 patients tested for ESR, CRP, and ESR + CRP per patient are estimated to be 17.3, 15.6, and 8.7, respectively. This means a combined ESR + CRP would lead to 8.6 fewer false-positives per 100 patients tested than ESR alone and 6.9 fewer false-positives per 100 patients tested than CRP alone. When cost differences are taken into account, the incremental cost per false-positive avoided for ESR + CRP compared with ESR alone is estimated to be $118.85. When compared with CRP alone, the cost per false-positive avoided is $152.89. The cost per total misdiagnosis avoided for ESR + CRP is estimated to be $157.02 and $240.62, respectively.

Table 15

Cost-Effectiveness Results for Periprosthetic Infection Population, ESR + CRP.

Table 16

Cost-Effectiveness Results for Periprosthetic Infection Population, ESR/CRP.

Table 16 presents cost-effectiveness results for the periprosthetic infection population when a combined positive test is defined as either ESR or CRP being positive (ESR/CRP). As shown under this assumption, ESR/CRP is estimated to produce 1.7 fewer false-negatives per 100 patients tested than ESR alone and 1.3 fewer false-negatives per 100 patients tested than CRP alone. The cost per false-negative avoided for ESR/CRP is estimated to be $611.22 compared with ESR alone, and $838.68 compared with CRP alone. ESR/CRP is dominated (higher costs, less effectiveness) by both ESR alone and CRP alone when all misdiagnoses are considered. The combined test leads to 5.6 more total misdiagnoses per 100 patients tested than CRP alone and 8.3 more misdiagnoses per 100 patients tested than CRP alone.

Pediatric orthopedic infections

Table 17 presents cost-effectiveness results for the pediatric orthopedic infection population, when the combined positive test is defined as both ESR and CRP being positive (ESR + CRP). As shown, the proportions of false-positives for ESR, CRP, and ESR + CRP per 100 patients are estimated to be 31.8, 12.1, and 9.3, respectively. This means a combined ESR + CRP would lead to 22.5 fewer false-positives per 100 patients tested than ESR alone and 2.8 fewer false-positives per 100 patients tested than CRP alone. When cost differences are taken into account, the incremental cost per false-positive avoided for ESR + CRP compared with ESR alone is estimated to be $45.97. When compared with CRP alone, the cost per false-positive avoided is $378.50. The cost per total misdiagnosis avoided for ESR + CRP is estimated to be $49.29 and $440.32, respectively.

Table 17

Cost-Effectiveness Results for Pediatric Orthopedic Infections, ESR + CRP.

Table 18 presents cost-effectiveness results for the pediatric orthopedic infection population, when a combined positive test is defined as either ESR or CRP being positive (ESR/CRP). ESR/CRP is estimated to produce 0.3 fewer false-negatives than ESR alone, and 0.2 fewer false-negatives than CRP alone. The cost per false-negative avoided for ESR/CRP is estimated to be $3,929 compared with ESR alone and $5,391.26 compared with CRP alone. Because specificity was not reported in the single study informing this analysis, total misdiagnoses could not be estimated.

Table 18

Cost-Effectiveness Results for Pediatric Orthopedic Infections, ESR/CRP.

Inflammatory Bowel Disease

Table 19 presents cost-effectiveness results for the inflammatory bowel disease population when the combined positive test is defined as both ESR and CRP being positive (ESR + CRP). The proportions of false-positives for ESR, CRP, and ESR + CRP per 100 patients tested are estimated to be 25.1, 22.9, and 12.2, respectively. This means a combined ESR + CRP would lead to 12.9 fewer false-positives per 100 patients tested than ESR alone and 10.7 fewer false-positives per 100 patients tested than CRP alone. When cost differences are taken into account, the incremental cost per false-positive avoided for ESR + CRP compared with ESR alone is estimated to be $79.60. When compared with CRP alone, the cost per false-positive avoided is $99.47. The cost per total misdiagnosis avoided for ESR + CRP is estimated to be $170.49 and $250.36, respectively.

Table 19

Cost-Effectiveness Results for Inflammatory Bowel Disease, ESR + CRP.

Giant Cell Arteritis

Table 20 presents cost-effectiveness results for patients suspected of having giant cell arteritis, using prevalence and diagnostic accuracy data from Kermani et al.¹⁹ The proportion of false-positives for ESR, CRP, and ESR + CRP in this population are estimated to be 53.3, 52.5, and 44.9 per 100 patients tested, respectively. Therefore, a combined ESR + CRP would lead to 8 fewer false-positives per 100 patients tested than ESR alone and 7.6 fewer false-positives per 100 patients tested than CRP alone. When cost differences are taken into account, the incremental cost per false-positive avoided for ESR + CRP compared with ESR alone is estimated to be $123.18. When compared with CRP alone, the cost per false-positive avoided is $139.44. The cost per total misdiagnosis avoided for ESR + CRP is estimated to be $134.73 and $165.43, respectively.

Table 20

Cost-Effectiveness Results for Giant Cell Arteritis, ESR + CRP (Kermani et al.).

Table 21 presents cost-effectiveness results for patients suspected of having giant cell arteritis, using prevalence and diagnostic accuracy data from Hayreh et al.²⁸ The proportion of false-positives for ESR, CRP, and ESR + CRP in this population are estimated to be 23.4, 12.7, and 5.7 per 100 patients tested, respectively. Therefore, a combined ESR + CRP would lead to 17.7 fewer false-positives per 100 patients tested than ESR alone, and 7.1 fewer false-positives per 100 patients tested than CRP alone. When cost differences are taken into account, the incremental cost per false-positive avoided for ESR + CRP compared with ESR alone is estimated to be $58.25. When compared with CRP alone, the cost per false-positive avoided is $149.86. The cost per total misdiagnosis avoided for ESR + CRP is estimated to be $57.30 and $163.33, respectively.

Table 21

Cost-Effectiveness Results for Giant Cell Arteritis, ESR + CRP (Hayreh et al.).

6.2.2. Sensitivity analysis

Table 22 presents cost-effectiveness results using cost estimates derived from provincial benefit fee schedules. As shown, the incremental cost per false-positive avoided of the combined test compared with ESR alone is very similar, using the alternate cost source compared with the base-case assumption across the various populations. However, compared with CRP alone, the incremental cost per false-positive avoided is nearly half of what it is using the base-case assumption. A similar pattern emerges for the incremental cost per false-negative for ESR/CRP compared with ESR alone and CRP alone.

Table 22

Sensitivity Analysis of Cost-Effectiveness Using Alternative Estimates for ESR ($5.83), CRP ($9.57), and Combined Testing ($15.40), Based on Provincial Fee Schedules.

Cost-effectiveness results are shown in Table 23, where the lower and upper confidence intervals for sensitivity and specificity for the testing strategies are used in the model. The confidence intervals for sensitivity and specificity for each test by population were based on data provided in Appendix 12 and Table 4. As shown, varying the diagnostic accuracy produces small variations in the incremental cost per false-positive avoided for ESR + CRP in most populations. The cost per false-positive avoided was most affected in the giant cell arteritis population when using sensitivity and specificity estimates from Kermani et al.¹⁹ More variation is seen for the incremental cost per false-negative avoided for ESR/CRP compared with ESR or CRP alone. In the pediatric orthopedic population, the cost per false-negative avoided for ESR/CRP compared with ESR alone is estimated to be $2,619 when the lower confidence intervals for diagnostic accuracy for all test strategies are used. The cost per false-negative avoided is estimated to be $5,239 if the upper 95% confidence intervals for diagnostic accuracy are used. The cost per false-negative avoided for ESR/CRP compared with CRP alone is estimated to be $3,235 if the lower 95% confidence intervals for diagnostic accuracy are used as estimates for diagnostic accuracy. If the higher 95% confidence interval for sensitivity is used for all testing strategies, the incremental cost per false-negative avoided for ESR/CRP compared with ESR alone becomes $8,087.

Table 23

Sensitivity Analysis of Cost-Effectiveness Using Lower and Upper Confidence Intervals for Diagnostic Accuracy for All Testing Strategies.

Table 24 presents cost-effectiveness results when the lower and upper confidence intervals for prevalence of disease for each population are used in the model. Little variation in the cost per false-positive avoided is found for ESR + CRP. Lower prevalence leads to lower cost per false-positive avoided for all populations, as lower prevalence rates lead to more false-positive test results and therefore more false-positives avoided for ESR + CRP because of its higher specificity. Greater variation in costs is found in the incremental cost per false-negative avoided for ESR/CRP. Higher prevalence leads to lower cost per false-negative avoided for all populations, because higher prevalence rates lead to fewer false-negative test results and therefore fewer false-negatives avoided for ESR/CRP.

Table 24

Sensitivity Analysis of Cost-Effectiveness Using Lower and Upper Confidence Intervals of Condition Prevalence.

Table 25 presents sensitivity analysis using the upper and lower 95% confidence intervals for both diagnostic accuracy and condition prevalence as model values for all testing strategies. As shown, there is not much difference in the cost per false-positive avoided for ESR + CRP between the lower and upper confidence sensitivity analysis, except in the giant cell arteritis population when diagnostic accuracy data from Kermani et al.¹⁹ are used There is also not much difference in the cost per false-negative of ESR/CRP versus either CRP or ESR alone between the lower and upper 95% confidence interval sensitivity analysis.

Table 25

Sensitivity Analysis of Cost-Effectiveness Using Lower and Upper Confidence Intervals of Both Diagnostic Accuracy and Condition Prevalence.