Background Computed tomography is the most common technique used to estimate the number of pulmonary metastases and their resectability. A lack of agreement between radiologic and surgical pathologic findings could potentially lead to incomplete resection or to rejection of patients for potentially curative treatments. The objective of this study was to estimate the disagreement between the number of radiologic lesions and the number of histologically confirmed malignant lesions excised from patients with pulmonary metastases from colorectal cancer. Methods This was a multicenter longitudinal study using a national registry. All patients underwent open surgery for pulmonary metastasectomy. Results Radiologic unilateral involvement was documented in 345 of 404 patients (85%); 253 (73%) presented with single nodules. The radiologic and malignant pathologic findings were concordant in 316 (78%) patients. The two independent predictors of discordance between computed tomography and the number of pathologic metastases were the bilateral involvement and the number of radiologic nodules. This model explained 28% of the variability in the disagreement frequency and discriminated between agreement and disagreement in 85% of the patients. Discrepancies increased with the nodule count with an odds ratio of 6.17 (95% confidence interval, 4.08 to 9.33) per additional nodule. For similar nodule counts, a lower disagreement frequency was observed among bilateral cases (odds ratio, 0.2; 95% confidence interval, 0.07 to 0.55). Conclusions Differences between the radiologic and pathologic findings were documented in 1 of every 5 patients. The correlation was very accurate in patients with single radiologic nodules. However, half of the patients with more nodules showed discrepancies.