Table 3.
Probability of detecting true resistance and true non-resistance in bacterial isolates using broth microdilution and disk diffusion susceptibility testing
| Non-type-specific E. coli | M. haemolytica | |||||||
|---|---|---|---|---|---|---|---|---|
| Ampicillin (n = 2279) | Ceftiofur (n = 2295) | Streptomycin (n = 2302) | Sulfisoxazole (n = 2316) | Tetracycline (n = 2238) | TMP-SMX (n = 2308) | Ampicillin (n = 783) | Tetracycline (n = 783) | |
| Probability of detecting true resistance (sensitivity)* | ||||||||
| Disk diffusion (95% PI) | 84·3% | 37·1% | 70·6% | 93·9% | 94·9% | 81·9% | 34·9% | 59·9% |
| (59·4–98·9) | (4·8–94·1) | (52·1–79·7) | (84·5–98·5) | (89·7–96·9) | (49·1–98·7) | (9·6–88·0) | (34·1–91·4) | |
| Broth microdilution (95% PI) | 94·7% | 79·4% | 87·3% | 96·1% | 92·6% | 81·5% | 79·4% | 89·6% |
| (76·7–99·1) | (45·7–96·4) | (65·5–93·8) | (88·1–98·7) | (87·8–94·7) | (58·8–95·9) | (39·4–96·1) | (57·8–98·3) | |
| Probability that sensitivity A > B‡ | 0·90 | 0·88 | 0·99 | 0·83 | 0·97 | 0·49 | 0·64 | 0·15 |
| Probability of detecting true non-resistance (specificity)† | ||||||||
| Disk diffusion (95% PI) | 99·9% | 99·9% | 99·7% | 99·7% | 98·0% | 99·9% | 99·7% | 99·8% |
| (99·5–99·9) | (99·8–99·9) | (98·7–99·9) | (98·8–99·9) | (94·8–99·4) | (99·7–100·0) | (98·9–99·9) | (98·8–99·9) | |
| Broth microdilution (95% PI) | 99·1% | 99·6% | 97·9% | 98·7% | 96·3% | 99·8% | 98·5% | 98·3% |
| (98·3–99·8) | (99·3–99·9) | (96·4–99·3) | (97·6–99·4) | (93·5–97·7) | (99·5–99·9) | (96·7–99·8) | (95·5–99·8) | |
| Probability that specificity A > B‡ | 0·98 | 0·97 | 0·99 | 0·99 | 0·97 | 0·83 | 0·51 | 0·79 |
TMP-SMX, Trimethoprim-sulfamethoxazole; PI, probability interval.
*
The probability of classifying isolates as resistant among isolates that are truly resistant.
†
The probability of classifying isolates as non-resistant (susceptible or intermediate) among isolates that are truly non-resistant.
‡
One-sided probability comparing values for disk diffusion and broth microdilution: A = value that is numerically greater; B = value that is numerically lower.