Use of multiple inflammatory marker tests in primary care: using Clinical Practice Research Datalink to evaluate accuracy

Jessica Watson; Hayley E Jones; Jonathan Banks; Penny Whiting; Chris Salisbury; Willie Hamilton

doi:10.3399/bjgp19X704309

Article Figures & Data

Figures

Tables

Download figure
Open in new tab
Download powerpoint
Figure 1.
Inflammatory marker tests requested. CRP = C-reactive protein. ESR = erythrocyte sedimentation rate. PV = plasma viscosity.

Tables

Figures

View popup

Table 1.

2×2 table of test results and overall disease incidence for single-versus double-tested patients, N = 136 961

Test result	Disease positive, n (%)	Disease negative, n (%)	Total, n (%)
Single inflammatory marker, N = 83 761
Raised	2999 (15.0)	16 933 (85.0)	19 932 (23.8)
Normal	3847 (6.0)	59 982 (94.0)^a	63 829 (76.2)

Multiple inflammatory markers,^b N= 53 200
Multiple raised inflammatory markers	1539 (22.6)	5264 (77.4)	6803 (12.8)
Discordant results	1174 (10.4)	10 101 (89.6)	11 275 (21.2)
All normal	2065 (5.9)	33 057 (94.1)^c	35 122 (66.0)

↵a Negative predictive value of a single normal inflammatory marker.
↵b Includes patients with two and three inflammatory marker tests. If two or more tests were raised patients were included in the ‘multiple raised inflammatory markers’; those with only one raised inflammatory marker were included in the ‘discordant results’ category.
↵c Negative predictive value of multiple normal inflammatory markers.

View popup

Table 2.

Comparison of measures of diagnostic accuracy for CRP, ESR, and PV, singly and in combination for diagnosis of any relevant disease (infection, autoimmune condition, or cancer)

Test and test combinations	Patients, n	True-positive, n (%)	False-positive, n (%)	True-negative, n (%)	False-negative, n (%)	Sensitivity (95% CI)	Specificity (95% CI)	PPV (95% CI)	NPV (95% CI)	DOR unadjusted (95% CI)	DOR adjusted^a (95% CI)
CRP	97 203	3947 (4.1)	18 745 (19.3)	69 797 (71.8)	4714 (4.8)	45.6 (44.5 to 46.6)	78.8 (78.6 to 79.1)	17.4 (16.9 to 17.9)	93.7 (93.5 to 93.9)	3.12 (2.98 to 3.26)	2.86 (2.73 to 2.99)
ESR	79 430	2780 (3.5)	15 589 (19.6)	57 221 (72.0)	3840 (4.8)	42.0 (40.8 to 43.2)	78.6 (78.3 to 78.9)	15.1 (14.6 to 15.7)	93.7 (93.5 to 93.9)	2.66 (2.52 to 2.80)	2.43 (2.30 to 2.55)
PV	13 834	536 (3.9)	3242 (23.4)	9439 (68.2)	617 (4.5)	46.5 (43.6 to 49.4)	74.4 (73.7 to 75.2)	14.2 (13.1 to 15.3)	93.9 (93.4 to 94.3)	2.53 (2.24 to 2.86)	2.32 (2.05 to 2.62)
CRP + ESR^b	43 820	1277 (2.9)	4269 (9.7)	35 574 (81.2)	2700 (6.2)	32.1 (30.7 to 33.6)	89.3 (89.0 to 89.6)	23.0 (21.9 to 24.2)	93.0 (92.7 to 93.2)	3.94 (3.66 to 4.24)	3.56 (3.31 to 3.84)
CRP\|ESR^c	43 820	2225 (5.1)	12 271 (28.0)	27 572 (62.9)	1752 (4.0)	56.0 (54.4 to 57.5)	69.2 (68.8 to 69.7)	15.4 (14.8 to 16.0)	94.0 (93.8 to 94.3)	2.85 (2.67 to 3.05)	2.61 (2.44 to 2.79)
CRP + PV	9575	267 (2.8)	974 (10.2)	7784 (81.3)	550 (5.7)	32.7 (29.5 to 36.0)	88.9 (88.2 to 89.5)	21.5 (19.3 to 23.9)	93.4 (92.9 to 93.9)	3.88 (3.30 to 4.56)	3.56 (3.02 to 4.19)
CRP\|PV	9575	495 (5.2)	3110 (32.5)	5648 (59.0)	322 (3.4)	60.6 (57.1 to 64.0)	64.5 (63.5 to 65.5)	13.7 (12.6 to 14.9)	94.6 (94.0 to 95.2)	2.79 (2.41 to 3.23)	2.59 (2.24 to 3.01)

↵a DOR = diagnostic odds ratio, adjusted for age and sex.
↵b CRP + ESR; positive test defined as both CRP and ESR positive.
↵c CRP|ESR; positive test defined as either CRP or ESR positive. CRP = C-reactive protein. ESR = erythrocyte sedimentation rate. NPV = negative predictive value. PPV = positive predictive value. PV = plasma viscosity.

View popup

Table 3.

Comparison of overall test performance among those with both CRP and ESR performed simultaneously, N = 43 820^a

Disease outcome^b	n	CRP AUC^c (95% CI)	ESR AUC (95% CI)	P-value for CRP versus ESR	CRP and ESR AUC (95% CI)	P-value for combined versus better single test
Any relevant disease	3977	0.682 (0.672 to 0.690)	0.665 (0.656 to 0.674)	<0.001	0.688 (0.678 to 0.697)	<0.001
Infections	1565	0.617 (0.601 to 0.632)	0.589 (0.574 to 0.603)	<0.001	0.619 (0.604 to 0.634)	0.018
Autoimmune conditions	1663	0.710 (0.697 to 0.724)	0.708 (0.695 to 0.721)	0.680	0.724 (0.710 to 0.737)	<0.001
Cancer	882	0.774 (0.759 to 0.788)	0.766 (0.752 to 0.781)	0.017	0.777 (0.763 to 0.791)	0.006
Polymyalgia rheumatica/giant cell arteritis	476	0.882 (0.880 to 0.900)	0.872 (0.860 to 0.890)	0.099	0.887 (0.874 to 0.900)	0.110
Rheumatoid arthritis	557	0.691 (0.670 to 0.712)	0.690 (0.669 to 0.711)	0.890	0.700 (0.679 to 0.721)	0.007
Seronegative arthritis	151	0.700 (0.653 to 0.746)	0.686 (0.638 to 0.734)	0.510	0.706 (0.659 to 0.753)	0.540
Inflammatory bowel disease	223	0.698 (0.660 to 0.737)	0.691 (0.653 to 0.730)	0.450	0.701 (0.662 to 0.740)	0.510

↵a Only 9494 (total amount in 1st column) out of the total 43 820 had a relevant disease outcome. The analysis of test performance uses those with and without relevant disease hence the total n = 43 820 is correct here.
↵b Where disease subtypes were examined, diseases other than the specified condition reported were classified as non-diseased.
↵c AUC = area under the receiver operator curve, where AUC = 0.5 is equivalent to no diagnostic utility and AUC = 1 is perfect diagnostic accuracy. AUC was calculated using logistic regression modelling with test result(s) on a log scale and age and sex as additional explanatory variables. CRP = C-reactive protein. ESR = erythrocyte sedimentation rate.

View popup

Table 4.

Comparison of overall test performance among those with both CRP and PV performed simultaneously, N = 9575

Disease outcome^a	n	CRP AUC^b (95% CI)	PV AUC (95% CI)	P-value for CRP versus PV	CRP and PV AUC (95% CI)	P-value for combined versus better single test
Any relevant disease	817	0.672 (0.652 to 0.692)	0.659 (0.640 to 0.679)	0.170	0.686 (0.667 to 0.706)	0.004
Infection	325	0.638 (0.608 to 0.670)	0.597 (0.564 to 0.628)	0.004	0.639 (0.608 to 0.669)	0.280
Autoimmune conditions	338	0.687 (0.657 to 0.717)	0.686 (0.655 to 0.717)	0.970	0.709 (0.679 to 0.739)	0.009
Cancer	183	0.753 (0.720 to 0.787)	0.760 (0.726 to 0.793)	0.490	0.764 (0.731 to 0.797)	0.170

↵a Where disease subtypes were examined, diseases other than the specified condition reported were classified as non-diseased.
↵b AUC = area under the receiver operator curve, where AUC = 0.5 is equivalent to no diagnostic utility and AUC = 1 is perfect diagnostic accuracy. AUC was calculated using logistic regression modelling with test result(s) on a log scale and age and sex as additional explanatory variables. CRP = C-reactive protein. PV = plasma viscosity.