CEM is an emerging breast imaging modality, introduced in 2011 by GE Healthcare, but is currently being offered by four vendors as commercial systems . Many studies have shown that CEM’s diagnostic accuracy is superior to FFDM, even matching the performance of breast MRI –. However, these results are mainly based on studies performed on GE Healthcare systems. In this study, we demonstrated that an increase in diagnostic performance can also be achieved with a Hologic system when compared to FFDM. When using Hologic CEM, sensitivity significantly increased from 74.3–87.6%, and specificity from 87.8–94.6%.
Although several reviews and meta-analyses have been published ,,, we believe that the one published by Zhu et al. is the most comprehensive one. In our opinion, their review provided the most extensive and up-to-date overview, including 18 studies published between 2014 and 2017. Seventeen of the 18 studies were performed on systems by GE Healthcare and only one study on a Hologic system. No studies using equipment of other vendors were included. In this review, the pooled sensitivity and specificity for CEM was 89% and 84%, respectively. The AUC value based on their summary ROC curve was 0.957. The combination of sensitivity and specificity observed in our study falls within the 95% confidence interval of the summary ROC curve given by Zhu et al. . Based on these findings, we concluded that sensitivity and specificity for CEM on a Hologic system are in line with prior findings for sensitivity and specificity of CEM studies that were mainly performed on GE Healthcare equipment.
Reproducibility is one of the pillars in science. In clinical studies, this means that certain results obtained can be achieved again (‘reproduced’) when the study is repeated with a similar methodology but by other, independent researchers. Reproducibility is important, since no claim can be proven by a single published study. Despite the utmost efforts of investigators to perform a study to the best of their ability, reproducible results provide the scientific community the only true transparency in results, giving us confidence in the results as well. In the context of the current study, it is important to reproduce the claims of CEM superiority over FFDM mainly made on GE Healthcare equipment, as this creates confidence that the results stand even when other equipment is used.
The number of studies using Hologic CEM units is still limited, often focusing on equally important aspects of CEM, such as studies on radiation exposure or evaluation of disease extent. To the best of our knowledge, there are currently eleven studies published that assessed diagnostic accuracy ,,,–. However, these studies often stated only partial diagnostic performance, for example only sensitivity. These studies were also relatively small, with sample sizes varying from 34 to 100 patients, with two exceptions up to 208 and 326 patients. One study provided a case-by-case description of the performance of CEM , but in the other ten studies, sensitivity of CEM ranged from 81.5–100% [16, 17, 20–27], with specificity available in just three studies and ranging from 42.6–79.6% –. However, the differences in methodologies used in these studies hamper a robust comparison of our results with these eleven studies.
Our study has some limitations. Firstly, the dominating indication for the CEM examinations in this population was the evaluation of recalls from breast cancer screening. However, a sub analysis of these (recalled) women versus all other indications showed that there was no statistically significant difference in sensitivity and specificity (data not shown). Secondly, the follow-up period used to determine the true negative disease status of cases deemed to be benign or negative could be considered to be short. A follow-up period of at least two years is preferable, which is a commonly accepted time interval used in screening programs. However, the follow-up period used was considered the minimum time interval. Half of the cases had a follow-up period of more than two years, while the shortest follow-up period lasted 17 months. In addition, the prevalence of interval carcinomas is low (0.21%), rendering it unlikely that an extra year of follow-up would reveal a substantial number of missed cancer cases . Consequently, prolonging follow-up will not result in significant alterations of the study conclusion. Finally, one might consider the sample size to be small, leading to underpowered results. However, increasing the sample size would only result in a smaller variance. Since the level of significance was already very high in this small sample size, it is unlikely that it is based on chance and will change with a larger sample size.
In summary, this study showed that CEM performed on a Hologic system improved both sensitivity and specificity when compared to FFDM. Our observed diagnostic performance falls within the 95% CI of summary ROC curves of a previously published meta-analysis. Thus, it can be safely assumed that the type of CEM unit does not seem to be a decisive factor in studies that are using equipment from different vendors.