Correction to: J Nucl Cardiol https://doi.org/10.1007/s12350-022-03099-x
For the article “Integration of coronary artery calcium scoring from CT attenuation scans by machine learning improves prediction of adverse cardiovascular events in patients undergoing SPECT/CT myocardial perfusion imaging,” by Feher et al. (J Nucl Cardiol. 2022 Oct 4. https://doi.org/10.1007/s12350-022-03099-x., PMID: 36195826) 11 patients included in the analysis had false data. While doing a routine data surveillance, we found out, that during the de-identification process 11 images were matched with wrong patient information. As we were not able to confidently identify the identity of these 11 images, due to this significant error we needed to exclude these 11 patients from the analysis. We have repeated all analyses after excluding these 11 patients to ensure that the results are not affected by including these patients. The hazard ratio minimally changed for Figure 4B from 5.3 (95% CI 4.3-6.5) to 5.2 (95% CI 4.2-6.5) with changes noted in NRI from overall NRI of 0.09 (95% CI 0.02, 0.17) to 0.05 (95% CI 0.003, 0.10), otherwise the results were not affected in any way after excluding these 11 patients including data represented in the text and in all other figures. Corrected versions of Figure 4 and the updated Tables 1 and 2 appear below; the authors sincerely regret these errors.
Figure 4.
Kaplan–Meier curves of cardiac events with a high versus low machine learning (ML) risk score (Panel A) and high versus low coronary artery calcification (CAC)-ML risk score. CI, confidence interval.
Table 1.
Baseline characteristics
| N | Overall n = 4759 | MACE n = 475 | No MACE N = 4284 | P value |
|---|---|---|---|---|
| Age, years | 64 (56-73) | 69 (60-78) | 64 (56-72) | < .001 |
| Female | 2115 (44%) | 148 (31%) | 1967 (46%) | < .001 |
| BMI, kg/m2 | 29.3 (25.5-33.7) | 28.3 (24.3-32.6) | 29.4 (25.6-33.9) | < .001 |
| Family history of CAD | 680 (14%) | 43 (9%) | 637 (15%) | < .001 |
| Smoking | 950 (20%) | 113 (24%) | 837 (20%) | .03 |
| Hypertension | 3050 (64%) | 331 (70%) | 2719 (64%) | .008 |
| Dyslipidemia | 2544 (53%) | 276 (58%) | 2268 (53%) | .03 |
| Diabetes | 1233 (26%) | 169 (36%) | 1064 (25%) | < .001 |
| PAD | 1247 (26%) | 225 (47%) | 1022 (24%) | < .001 |
| History of MI | 392 (8%) | 72 (15%) | 320 (8%) | < .001 |
| History of PCI | 532 (11%) | 104 (22%) | 428 (10%) | < .001 |
| History of CABG | 264 (6%) | 67 (14%) | 197 (5%) | < .001 |
| Resting SBP, mmHg | 138 (125-153) | 140 (126-158) | 138 (124-152) | .02 |
| Resting DBP, mmHg | 80 (73-86) | 77 (70-85) | 80 (73-86) | < .001 |
| Resting HR, beats/min | 71 (63-80) | 70 (62-79) | 71 (63-80) | .52 |
| LVH on resting ECG | 307 (7%) | 43 (9%) | 264 (6%) | .02 |
Table 2.
Stress test and imaging characteristics
| N | Overall n = 4759 | MACE n = 475 | No MACE N = 4284 | P value |
|---|---|---|---|---|
| Stress type | ||||
| Exercise | 1764 (37%) | 68 (14%) | 1684 (39%) | < .001 |
| Regadenoson | 2995 (63%) | 407 (86%) | 2600 (61%) | |
| Stress HR, beats/min | 107 [89-144] | 91 (80-110) | 110 (90-146) | < .001 |
| Stress SBP, mmHg | 153 (131-175) | 142 (120-164) | 154 (132-176) | < .001 |
| Stress DBP, mmHg | 80 (71-88) | 73 (64-82) | 80 (72-88) | < .001 |
| Exercise duration | ||||
| ≤6 min | 452 (26% exercise) | 22 (32% exercise) | 430 (26% exercise) | .20 |
| 7-9 min | 842 (48% exercise) | 34 (50% exercise) | 796 (47% exercise) | |
| ≥10 min | 470 (27% exercise) | 12 (18% exercise) | 458 (27% exercise) | |
| ECG response to test | ||||
| Negative | 3330 (70%) | 298 (63%) | 3032 (71%) | < .001 |
| Positive | 473 (10%) | 40 (9%) | 433 (10%) | |
| Equivocal | 279 (6%) | 20 (4%) | 259 (6%) | |
| Non-diagnostic | 665 (14%) | 115 (24%) | 550 (13%) | |
| ECG ST slope | ||||
| Upsloping | 188 (4%) | 10 (2%) | 178 (4%) | .15 |
| Down-sloping | 134 (3%) | 12 (3%) | 122 (3%) | |
| Horizontal | 423 (9%) | 40 (8%) | 383 (9%) | |
| Stress end-diastolic volume, mL | 88 (68-115) | 83 (59-113) | 87 (68-113) | < .001 |
| Stress end-diastolic wall volume, mL | 128 (110-151) | 102 (77-137) | 127 (109-149) | < .001 |
| Stress end systolic volume, mL | 31 (20-49) | 42 (26-70) | 30 (19-47) | < .001 |
| Stress TPD (after AC), % | 2.34 (0.78-5.14) | 3.89 (1.61-7.94) | 2.22 (0.71-4.80) | < .001 |
| Stress TPD (no AC), % | 2.21 (0.76-5.02) | 3.55 (1.16-8.04) | 2.12 (0.71-4.75) | < .001 |
| Stress quality control | 1.62 (1.27-2.03) | 1.76 (1.35-2.25) | 1.61 (1.26-2.00) | < .001 |
| Calcium score | 88 (0-602) | 564 (130-1364) | 62 (0-521) | < .001 |
The abstract should read:
“From the REFINE SPECT Registry 4759 patients with SPECT/CT performed at a single center were included (age: 64 ± 12 years, 45% female).”
“On survival analysis patients with high CAC-ML score (> 0.091) had higher event rate when compared to patients with low CAC-ML score (CI 5.2, 95% CI 4.2-6.5, P > .001).”
The results should read:
“The final study population comprised 4759 patients after exclusion of 228 studies without CTAC or with non-diagnostic CTAC from the total of 4987 Yale New Haven Hospital studies included in the REFINE-SPECT registry. Out of the 4759 included studies 4122 were performed on the Discovery 570c with the remaining performed on NM 530c.”
“These results were also confirmed in NRI analysis where the model with CAC had overall NRI of 0.05 (95% CI 0.003, 0.10). The positive and negative NRI were 0.06 (95% CI 0.02, 0.1) and − 0.01 (95% CI − 0.02, − 0.004), respectively”.
“Patients with high CAC-ML score (> 0.091, had higher event rate when compared to patients with low CAC-ML score (< 0.091, HR 5.2, 95% CI 4.2-6.5, P < .001) (Figure 4B).”
Footnotes
The original article can be found online at https://doi.org/10.1007/s12350-022-03099-x.