- Email: [email protected]
Comparison of heartworm antigen test kit performance in dogs having low heartworm burdens
Veterinary Parasitology 96 (2001) 317–322
Comparison of heartworm antigen test kit performance in dogs having low heartworm burdens Charles H. Courtney∗ , Qi-Yun Zeng Department of Pathobiology, College of Veterinary Medicine, University of Florida, Box 100125, Gainesville, FL 32610-0125, USA Received 31 August 2000; received in revised form 13 December 2000; accepted 13 December 2000
Abstract Sensitivity and specificity of four in-clinic heartworm antigen test kits, AbboScreen (Abbott Laboratories), Snap PF (IDEXX Laboratories), Solo Step (HESKA Corporation), Witness (Synbiotics Corporation) and two heartworm antigen microwell plate assays, DiroCHEK (Synbiotics) and PetChek PF (IDEXX) were compared in a blinded study using serum or plasma drawn from 237 random source dogs, including 140 with necropsy-confirmed, low worm burden infections (minimum 1 worm, maximum 10, mean 2.3, median 3) and 97 confirmed heartworm-free at necropsy. In general, microwell format tests were more sensitive than membrane format tests and tests using ELISA technology were more sensitive than tests using lateral flow immunochromatographic technology. Percent sensitivity and specificity, respectively, were PetChek PF 76 and 97, DiroCHEK 71 and 94, SNAP PF 67 and 98, Solo Step 60 and 98, and AbboScreen 52 and 96. The Witness test protocol was changed by the manufacturer midway through the study, and the newer version of this test kit arrived containing a package insert alerting the user to a change in procedure, which purportedly resulted in improved sensitivity. PetChek was significantly more sensitive than all other tests except DiroCHEK and the new version of Witness. DiroCHEK was significantly more sensitive than all tests except PetCheck, SNAP and the new version of Witness. Snap was more sensitive than AbboScreen and the old version of Witness. Differences in specificity were not significant (P > 0.05). © 2001 Elsevier Science B.V. All rights reserved. Keywords: Dirofilaria immitis; Diagnosis; Antigen test
1. Introduction This laboratory has evaluated a number of heartworm antigen test kits over the years, and in general the various test kits have improved in sensitivity, specificity and convenience ∗ Corresponding author. Tel.: +1-352-392-4700, ext. 5111; fax: +1-352-392-8351. E-mail address: [email protected] (C.H. Courtney).
0304-4017/01/$ – see front matter © 2001 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 4 - 4 0 1 7 ( 0 1 ) 0 0 3 7 4 - 0
318
C.H. Courtney, Q.-Y. Zeng / Veterinary Parasitology 96 (2001) 317–322
of use over time. However, one flaw in our system of evaluation lies in the population of dogs from which test specimens were drawn — mostly dogs euthanized at animal shelters in Florida. This population is characterized by a high prevalence of heartworm infection, approaching 50%, and high worm burdens, often 30 or more worms per dog (Courtney and Zeng, 1989). In contrast the population of dogs seen in a typical suburban veterinary practice is likely to be characterized by a low prevalence of infection and lower worm burdens. This means that sensitivity values for the various antigen tests as determined by studies previously done in this laboratory are likely to overestimate the sensitivity actually encountered in typical veterinary practices. Also, given the relatively high worm burdens used in this laboratory’s test system, all antigen tests tended to have similar and high levels of sensitivity. In the present study we compare canine heartworm antigen test kits commercially marketed in the US using a sera only from dogs having necropsy-confirmed infections with 10 or fewer worms. By doing so, we hope to better define differences in sensitivity between test kits and present veterinarians with more realistic values of test kit sensitivities.
2. Materials and methods Six commercially available heartworm antigen test kits were evaluated in a blinded study using 237 sera, 140 from dogs having necropsy-confirmed infections with 10 or fewer heartworms and 97 from dogs confirmed heartworm free by necropsy. Testing was accomplished on two different dates, with 96 sera tested on 11 March 2000 and 141 tested on 15 April 2000. All testing took place in facilities in Westbrook, Maine provided by IDEXX Laboratories. Test kits evaluated included two microwell ELISA tests, DiroCHEK (Synbiotics Corporation, San Diego, CA, lot number 11176) and PetChek PF (IDEXX Laboratories, Westbrook, ME, lot number 544MS). These test kits commonly are used in batch testing of multiple specimens by referral laboratories and practices having high volumes of testing. The remaining kits were membrane format tests designed for rapid in-clinic use. These included one membrane ELISA test, SNAP PF (IDEXX Laboratories, Westbrook, ME, lot number 033MS) and three lateral flow immunochromatographic tests, AbboScreen (Abbott Laboratories, Abbott Park, IL, lot number 55115M300), Solo Step (Heska Corporation, Ft. Collins, CO, lot number V010N06901 used in March and 9902963 in April), and Witness (Synbiotics Corporation, San Diego, CA, lot numbers NB037A used in March and NB046 in April). Test kits were purchased from the open market except that IDEXX directly provided the PetChek and SNAP kits from in-house stores of field-released lots. Seven animal hospital technicians regularly employed at three private veterinary practices in southern Maine, and experienced in running commercial heartworm antigen tests, conducted the testing following the manufacturer’s directions included with each test kit. Blinding of samples was accomplished by (1) concealing the true infection status of the dogs from which the sera were collected, (2) having each technician test sera from both infected and uninfected dogs, and (3) renumbering the identification codes of each technician’s allocation of sera between different test kit runs. Each technician ran all assigned sera using one test kit, then the next test kit, etc., until all test kits were run using the sera assigned to that technician. Four technicians each tested approximately one quarter of the serum specimens using each membrane format test (AbboScreen, SNAP, Solo Step, and Witness).
C.H. Courtney, Q.-Y. Zeng / Veterinary Parasitology 96 (2001) 317–322
319
Two technicians, experienced in running microwell format tests, ran the DiroCHEK and PetChek tests. Each ran approximately half of the serum specimens using both microwell format tests. Values for sensitivity and specificity were calculated and statistically analyzed as recommended by Courtney and Cornell (1990). Pairwise comparison of test kit sensitivity or specificity was made using McNemar’s paired chi-square test with the Yates correction for continuity (Armitage, 1971) and a proprietary microcomputer statistical package (Sigma Stat version 2.03, SPSS, Chicago, IL). Experimentwise error rates for the McNemar’s tests were controlled by setting the level of alpha considered to be significant at 0.05 divided by the number of simultaneous pairwise comparisons to be made. Confidence limits for sensitivity and specificity values were calculated using the Bonferoni method (Miller, 1967).
3. Results One hundred and forty of the 237 dogs, whose sera were used in this study, were infected with heartworms. Worm burdens of infected dogs ranged from 1 to 10 worms with a mean of 2.3 and a median of 3. Forty-nine of the 96 dogs tested on 11 March were infected with 1–9 worms for a mean worm burden of 1.5 and a median of 2. Ninety-one of 141 dogs tested on 15 April were infected with 1–10 worms for a mean worm burden of 2.9 and a median of 4. Percent sensitivity and specificity of the various test kits, excluding Witness, along with 95% confidence intervals for those values are shown in Table 1. In general, microwell format tests were more sensitive than membrane format tests, and tests using ELISA technology were more sensitive than tests using lateral flow immunochromatography for all infections and infections involving 1–2 adult female worms. Significant differences were not detected for infections in which adult female worms were absent or where more than two adult female worms were present, most likely due to the smaller numbers of samples tested in these two categories with a concomitant reduction in statistical power. This was further
Table 1 Comparative percent sensitivity and specificity of test kitsa Test kit
PetChek DiroChek SNAP Solo Step AbboScreen a
Sensitivityb 0 adult female wormsc (N = 20)
1–2 adult female worms (N = 89)
>2 adult female worms (N = 31)
All infections (N = 140)
45 (24–68) 40 (20–64) 35 (16–60) 35 (16–60) 30 (13–55)
77a (66–86) 71a,b (59–80) 65b,c (53–75) 56c,d (44–68) 46d (35–58)
94 (76–98) 94 (76–98) 94 (76–98) 90 (72–97) 84 (65–94)
76a (68–83) 71a,b (62–79) 67b,c (58–75) 60c (51–69) 52d (43–61)
Specificity, uninfected (N = 97) 97 (90–99) 94 (86–97) 98 (92–100) 98 (92–100) 96 (89–99)
95% confidence limits shown in parentheses. Values having different superscripts were significantly different (P < 0.05 Type I error rate for all possible paired comparisons within a column; P < 0.005 for any one individual paired comparison). c Infected only by male and/or immature and/or dead worms. b
320
C.H. Courtney, Q.-Y. Zeng / Veterinary Parasitology 96 (2001) 317–322
Table 2 Comparative percent sensitivity and specificity of the old version of the Witness test Test kit
Witness PetChek DiroChek SNAP Solo Step AbboScreen
Sensitivity 0 adult female wormsa (N = 8)
1–2 adult female worms (N = 35)
>2 adult female worms (N = 6)
All infections (N = 49)
0 50 38 25 25 25
17 71b 60b 51b 30 23
67 83 83 83 83 83
20 69b 59b 51b 36b 31
Specificity, uninfected (N = 47) 98 100 94 100 100 96
a
Infected only by male and/or immature and/or dead worms. Value is significantly different from old Witness (columnwise significance set at P < 0.05 for the five possible paired comparisons within a column between Witness and the other tests; P < 0.01 for any one paired comparison within a column). b
confounded by the reduced range of sensitivities between test kits at the extremes of female worm burden, i.e. all tests performed poorly when adult female worms were absent and all performed better when more than two adult female worms were present. Differences in specificity of test kits were not significant (Table 1). Synbiotics Corporation changed the Witness test protocol between the two study dates, for the newer version of the Witness test kit arrived containing a package insert alerting the user to a change in procedure which, according to the manufacturer’s claims, results in improved sensitivity. Percent sensitivity and specificity, respectively, of the old version of Witness was 20 and 98 (Table 2) whereas that of the newer version of Witness was 71 and 94 (Table 3), but statistical significance for the difference between the two versions of the Witness test cannot be calculated because the tests were run using sera from different
Table 3 Comparative percent sensitivity and specificity of the new version of the Witness test Test kit
Witness PetChek DiroChek SNAP Solo Step AbboScreen a
Sensitivitya,b
Specificity, uninfected (N = 50)
0 adult female wormsb (N = 55)
1–2 adult female worms (N = 12)
>2 adult female worms (N = 25)
All infections (N = 91)
42 42 42 42 42 33
69 82 78 75 73 62
92 96 96 96 92 84
71 80 78 76 74 64
94 94 94 96 96 96
Infected only by male and/or immature and/or dead worms. No significant differences were found between Witness and the other tests (columnwise significance set at P < 0.05 for the five possible paired comparisons within a column between Witness and the other tests; P < 0.01 for any one paired comparison within a column). b
C.H. Courtney, Q.-Y. Zeng / Veterinary Parasitology 96 (2001) 317–322
321
dogs. All other test kits except AbboScreen were significantly more sensitive than the old version of the Witness test. In contrast, there were no statistically significant differences between the new version of the Witness test and the other test kits although Witness was arithmetically less sensitive that all tests except AbboScreen.
4. Discussion By testing sera from dogs infected with 10 or fewer worms, significant differences in sensitivity between test kits were detected. Not surprisingly, the two microwell ELISA tests were the most sensitive. The next most sensitive test, Snap, uses a membrane ELISA format whereas the remaining membrane format tests use lateral flow immunochromatography. Thus it appears that ELISA technology, whatever the format, provides superior sensitivity than lateral flow immunochromatography, although the latter technology is simpler to use. At first glance, this study supports the manufacturer’s claim that the sensitivity of the Witness test had been improved between the two lot numbers. A major caveat remains, however, in that a comparison of the difference in sensitivity between versions of the Witness test in this study is not statistically valid because these tests were run with sera collected from different dogs (Courtney and Cornell, 1990). The dogs used in the April study had slightly higher heartworm burdens than those used in the March study, thus one could argue that the observed improvement in sensitivity could be explained by the slightly higher worm burdens in the population tested with the revised Witness test. This is not to say that there has been no improvement in the Witness test, but only that interpretation of our data as a comparison of the two Witness versions is inappropriate. Finally, all tests were run using serum or plasma as the test specimen. The membrane format tests also may be used with whole blood and indeed are more likely to be used with whole blood in everyday clinical practice. However, the design of this study precluded the use of whole blood specimens. Any nearby sources of canine blood, such as animal shelters, would not permit necropsy of the dogs from which the samples were drawn. A commercial research dog supplier was unable to collect and ship specimens that would arrive in time to be tested in accordance with some of the test kit manufacturer’s stated limits for the time interval between collection and testing of whole blood specimens. Although test kit performance using plasma or serum does not necessarily predict performance using whole blood, our experience has been that sensitivity of the membrane format tests using whole blood follows, or is slightly less than, that obtained when using serum or plasma (unpublished data, Dr. Sampath Srikanth, IDEXX Laboratories, Westbrook, ME).
Acknowledgements We gratefully acknowledge the contributions of the animal hospital technicians who gave up their saturdays to run the tests for this study: Sarah Fowler and Jennifer Peaslee of Bath Animal Hospital, Bath, Maine, Sara Daggett, Deirdre Kilgallen-Tracy, Megan Pelligrini, and Lynn Leary of the Veterinary Center of Cape Elizabeth, Cape Elizabeth, Maine, and Carol Lynn of Forest Avenue Veterinary Hospital, Portland, Maine.
322
C.H. Courtney, Q.-Y. Zeng / Veterinary Parasitology 96 (2001) 317–322
References Armitage, P., 1971. Statistical Methods in Medical Research. Blackwell Scientific Publications, London. Courtney, C.H., Cornell, J.A., 1990. Evaluation of heartworm immunodiagnostic tests. J. Am. Vet. Med. Assoc. 197, 724–729. Courtney, C.H., Zeng, Q.-Y., 1989. The structure of heartworm populations in dogs and cats in Florida. In: Otto, G.F. (Ed.), Proceedings of the Heartworm Symposium, American Heartworm Society, Washington, DC, 1989, pp. 1–6. Miller, R.G., 1967. Simultaneous Statistical Inference. McGraw-Hill, New York.
Comparison of heartworm antigen test kit performance in dogs having low heartworm burdens Charles H. Courtney∗ , Qi-Yun Zeng Department of Pathobiology, College of Veterinary Medicine, University of Florida, Box 100125, Gainesville, FL 32610-0125, USA Received 31 August 2000; received in revised form 13 December 2000; accepted 13 December 2000
Abstract Sensitivity and specificity of four in-clinic heartworm antigen test kits, AbboScreen (Abbott Laboratories), Snap PF (IDEXX Laboratories), Solo Step (HESKA Corporation), Witness (Synbiotics Corporation) and two heartworm antigen microwell plate assays, DiroCHEK (Synbiotics) and PetChek PF (IDEXX) were compared in a blinded study using serum or plasma drawn from 237 random source dogs, including 140 with necropsy-confirmed, low worm burden infections (minimum 1 worm, maximum 10, mean 2.3, median 3) and 97 confirmed heartworm-free at necropsy. In general, microwell format tests were more sensitive than membrane format tests and tests using ELISA technology were more sensitive than tests using lateral flow immunochromatographic technology. Percent sensitivity and specificity, respectively, were PetChek PF 76 and 97, DiroCHEK 71 and 94, SNAP PF 67 and 98, Solo Step 60 and 98, and AbboScreen 52 and 96. The Witness test protocol was changed by the manufacturer midway through the study, and the newer version of this test kit arrived containing a package insert alerting the user to a change in procedure, which purportedly resulted in improved sensitivity. PetChek was significantly more sensitive than all other tests except DiroCHEK and the new version of Witness. DiroCHEK was significantly more sensitive than all tests except PetCheck, SNAP and the new version of Witness. Snap was more sensitive than AbboScreen and the old version of Witness. Differences in specificity were not significant (P > 0.05). © 2001 Elsevier Science B.V. All rights reserved. Keywords: Dirofilaria immitis; Diagnosis; Antigen test
1. Introduction This laboratory has evaluated a number of heartworm antigen test kits over the years, and in general the various test kits have improved in sensitivity, specificity and convenience ∗ Corresponding author. Tel.: +1-352-392-4700, ext. 5111; fax: +1-352-392-8351. E-mail address: [email protected] (C.H. Courtney).
0304-4017/01/$ – see front matter © 2001 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 4 - 4 0 1 7 ( 0 1 ) 0 0 3 7 4 - 0
318
C.H. Courtney, Q.-Y. Zeng / Veterinary Parasitology 96 (2001) 317–322
of use over time. However, one flaw in our system of evaluation lies in the population of dogs from which test specimens were drawn — mostly dogs euthanized at animal shelters in Florida. This population is characterized by a high prevalence of heartworm infection, approaching 50%, and high worm burdens, often 30 or more worms per dog (Courtney and Zeng, 1989). In contrast the population of dogs seen in a typical suburban veterinary practice is likely to be characterized by a low prevalence of infection and lower worm burdens. This means that sensitivity values for the various antigen tests as determined by studies previously done in this laboratory are likely to overestimate the sensitivity actually encountered in typical veterinary practices. Also, given the relatively high worm burdens used in this laboratory’s test system, all antigen tests tended to have similar and high levels of sensitivity. In the present study we compare canine heartworm antigen test kits commercially marketed in the US using a sera only from dogs having necropsy-confirmed infections with 10 or fewer worms. By doing so, we hope to better define differences in sensitivity between test kits and present veterinarians with more realistic values of test kit sensitivities.
2. Materials and methods Six commercially available heartworm antigen test kits were evaluated in a blinded study using 237 sera, 140 from dogs having necropsy-confirmed infections with 10 or fewer heartworms and 97 from dogs confirmed heartworm free by necropsy. Testing was accomplished on two different dates, with 96 sera tested on 11 March 2000 and 141 tested on 15 April 2000. All testing took place in facilities in Westbrook, Maine provided by IDEXX Laboratories. Test kits evaluated included two microwell ELISA tests, DiroCHEK (Synbiotics Corporation, San Diego, CA, lot number 11176) and PetChek PF (IDEXX Laboratories, Westbrook, ME, lot number 544MS). These test kits commonly are used in batch testing of multiple specimens by referral laboratories and practices having high volumes of testing. The remaining kits were membrane format tests designed for rapid in-clinic use. These included one membrane ELISA test, SNAP PF (IDEXX Laboratories, Westbrook, ME, lot number 033MS) and three lateral flow immunochromatographic tests, AbboScreen (Abbott Laboratories, Abbott Park, IL, lot number 55115M300), Solo Step (Heska Corporation, Ft. Collins, CO, lot number V010N06901 used in March and 9902963 in April), and Witness (Synbiotics Corporation, San Diego, CA, lot numbers NB037A used in March and NB046 in April). Test kits were purchased from the open market except that IDEXX directly provided the PetChek and SNAP kits from in-house stores of field-released lots. Seven animal hospital technicians regularly employed at three private veterinary practices in southern Maine, and experienced in running commercial heartworm antigen tests, conducted the testing following the manufacturer’s directions included with each test kit. Blinding of samples was accomplished by (1) concealing the true infection status of the dogs from which the sera were collected, (2) having each technician test sera from both infected and uninfected dogs, and (3) renumbering the identification codes of each technician’s allocation of sera between different test kit runs. Each technician ran all assigned sera using one test kit, then the next test kit, etc., until all test kits were run using the sera assigned to that technician. Four technicians each tested approximately one quarter of the serum specimens using each membrane format test (AbboScreen, SNAP, Solo Step, and Witness).
C.H. Courtney, Q.-Y. Zeng / Veterinary Parasitology 96 (2001) 317–322
319
Two technicians, experienced in running microwell format tests, ran the DiroCHEK and PetChek tests. Each ran approximately half of the serum specimens using both microwell format tests. Values for sensitivity and specificity were calculated and statistically analyzed as recommended by Courtney and Cornell (1990). Pairwise comparison of test kit sensitivity or specificity was made using McNemar’s paired chi-square test with the Yates correction for continuity (Armitage, 1971) and a proprietary microcomputer statistical package (Sigma Stat version 2.03, SPSS, Chicago, IL). Experimentwise error rates for the McNemar’s tests were controlled by setting the level of alpha considered to be significant at 0.05 divided by the number of simultaneous pairwise comparisons to be made. Confidence limits for sensitivity and specificity values were calculated using the Bonferoni method (Miller, 1967).
3. Results One hundred and forty of the 237 dogs, whose sera were used in this study, were infected with heartworms. Worm burdens of infected dogs ranged from 1 to 10 worms with a mean of 2.3 and a median of 3. Forty-nine of the 96 dogs tested on 11 March were infected with 1–9 worms for a mean worm burden of 1.5 and a median of 2. Ninety-one of 141 dogs tested on 15 April were infected with 1–10 worms for a mean worm burden of 2.9 and a median of 4. Percent sensitivity and specificity of the various test kits, excluding Witness, along with 95% confidence intervals for those values are shown in Table 1. In general, microwell format tests were more sensitive than membrane format tests, and tests using ELISA technology were more sensitive than tests using lateral flow immunochromatography for all infections and infections involving 1–2 adult female worms. Significant differences were not detected for infections in which adult female worms were absent or where more than two adult female worms were present, most likely due to the smaller numbers of samples tested in these two categories with a concomitant reduction in statistical power. This was further
Table 1 Comparative percent sensitivity and specificity of test kitsa Test kit
PetChek DiroChek SNAP Solo Step AbboScreen a
Sensitivityb 0 adult female wormsc (N = 20)
1–2 adult female worms (N = 89)
>2 adult female worms (N = 31)
All infections (N = 140)
45 (24–68) 40 (20–64) 35 (16–60) 35 (16–60) 30 (13–55)
77a (66–86) 71a,b (59–80) 65b,c (53–75) 56c,d (44–68) 46d (35–58)
94 (76–98) 94 (76–98) 94 (76–98) 90 (72–97) 84 (65–94)
76a (68–83) 71a,b (62–79) 67b,c (58–75) 60c (51–69) 52d (43–61)
Specificity, uninfected (N = 97) 97 (90–99) 94 (86–97) 98 (92–100) 98 (92–100) 96 (89–99)
95% confidence limits shown in parentheses. Values having different superscripts were significantly different (P < 0.05 Type I error rate for all possible paired comparisons within a column; P < 0.005 for any one individual paired comparison). c Infected only by male and/or immature and/or dead worms. b
320
C.H. Courtney, Q.-Y. Zeng / Veterinary Parasitology 96 (2001) 317–322
Table 2 Comparative percent sensitivity and specificity of the old version of the Witness test Test kit
Witness PetChek DiroChek SNAP Solo Step AbboScreen
Sensitivity 0 adult female wormsa (N = 8)
1–2 adult female worms (N = 35)
>2 adult female worms (N = 6)
All infections (N = 49)
0 50 38 25 25 25
17 71b 60b 51b 30 23
67 83 83 83 83 83
20 69b 59b 51b 36b 31
Specificity, uninfected (N = 47) 98 100 94 100 100 96
a
Infected only by male and/or immature and/or dead worms. Value is significantly different from old Witness (columnwise significance set at P < 0.05 for the five possible paired comparisons within a column between Witness and the other tests; P < 0.01 for any one paired comparison within a column). b
confounded by the reduced range of sensitivities between test kits at the extremes of female worm burden, i.e. all tests performed poorly when adult female worms were absent and all performed better when more than two adult female worms were present. Differences in specificity of test kits were not significant (Table 1). Synbiotics Corporation changed the Witness test protocol between the two study dates, for the newer version of the Witness test kit arrived containing a package insert alerting the user to a change in procedure which, according to the manufacturer’s claims, results in improved sensitivity. Percent sensitivity and specificity, respectively, of the old version of Witness was 20 and 98 (Table 2) whereas that of the newer version of Witness was 71 and 94 (Table 3), but statistical significance for the difference between the two versions of the Witness test cannot be calculated because the tests were run using sera from different
Table 3 Comparative percent sensitivity and specificity of the new version of the Witness test Test kit
Witness PetChek DiroChek SNAP Solo Step AbboScreen a
Sensitivitya,b
Specificity, uninfected (N = 50)
0 adult female wormsb (N = 55)
1–2 adult female worms (N = 12)
>2 adult female worms (N = 25)
All infections (N = 91)
42 42 42 42 42 33
69 82 78 75 73 62
92 96 96 96 92 84
71 80 78 76 74 64
94 94 94 96 96 96
Infected only by male and/or immature and/or dead worms. No significant differences were found between Witness and the other tests (columnwise significance set at P < 0.05 for the five possible paired comparisons within a column between Witness and the other tests; P < 0.01 for any one paired comparison within a column). b
C.H. Courtney, Q.-Y. Zeng / Veterinary Parasitology 96 (2001) 317–322
321
dogs. All other test kits except AbboScreen were significantly more sensitive than the old version of the Witness test. In contrast, there were no statistically significant differences between the new version of the Witness test and the other test kits although Witness was arithmetically less sensitive that all tests except AbboScreen.
4. Discussion By testing sera from dogs infected with 10 or fewer worms, significant differences in sensitivity between test kits were detected. Not surprisingly, the two microwell ELISA tests were the most sensitive. The next most sensitive test, Snap, uses a membrane ELISA format whereas the remaining membrane format tests use lateral flow immunochromatography. Thus it appears that ELISA technology, whatever the format, provides superior sensitivity than lateral flow immunochromatography, although the latter technology is simpler to use. At first glance, this study supports the manufacturer’s claim that the sensitivity of the Witness test had been improved between the two lot numbers. A major caveat remains, however, in that a comparison of the difference in sensitivity between versions of the Witness test in this study is not statistically valid because these tests were run with sera collected from different dogs (Courtney and Cornell, 1990). The dogs used in the April study had slightly higher heartworm burdens than those used in the March study, thus one could argue that the observed improvement in sensitivity could be explained by the slightly higher worm burdens in the population tested with the revised Witness test. This is not to say that there has been no improvement in the Witness test, but only that interpretation of our data as a comparison of the two Witness versions is inappropriate. Finally, all tests were run using serum or plasma as the test specimen. The membrane format tests also may be used with whole blood and indeed are more likely to be used with whole blood in everyday clinical practice. However, the design of this study precluded the use of whole blood specimens. Any nearby sources of canine blood, such as animal shelters, would not permit necropsy of the dogs from which the samples were drawn. A commercial research dog supplier was unable to collect and ship specimens that would arrive in time to be tested in accordance with some of the test kit manufacturer’s stated limits for the time interval between collection and testing of whole blood specimens. Although test kit performance using plasma or serum does not necessarily predict performance using whole blood, our experience has been that sensitivity of the membrane format tests using whole blood follows, or is slightly less than, that obtained when using serum or plasma (unpublished data, Dr. Sampath Srikanth, IDEXX Laboratories, Westbrook, ME).
Acknowledgements We gratefully acknowledge the contributions of the animal hospital technicians who gave up their saturdays to run the tests for this study: Sarah Fowler and Jennifer Peaslee of Bath Animal Hospital, Bath, Maine, Sara Daggett, Deirdre Kilgallen-Tracy, Megan Pelligrini, and Lynn Leary of the Veterinary Center of Cape Elizabeth, Cape Elizabeth, Maine, and Carol Lynn of Forest Avenue Veterinary Hospital, Portland, Maine.
322
C.H. Courtney, Q.-Y. Zeng / Veterinary Parasitology 96 (2001) 317–322
References Armitage, P., 1971. Statistical Methods in Medical Research. Blackwell Scientific Publications, London. Courtney, C.H., Cornell, J.A., 1990. Evaluation of heartworm immunodiagnostic tests. J. Am. Vet. Med. Assoc. 197, 724–729. Courtney, C.H., Zeng, Q.-Y., 1989. The structure of heartworm populations in dogs and cats in Florida. In: Otto, G.F. (Ed.), Proceedings of the Heartworm Symposium, American Heartworm Society, Washington, DC, 1989, pp. 1–6. Miller, R.G., 1967. Simultaneous Statistical Inference. McGraw-Hill, New York.