¹⁸F-FDG Fetal Dosimetry Calculated with PET/MRI

DISCUSSION

This study, which contains the largest population ever published of pregnant patients imaged with ¹⁸F-FDG, significantly expands the pool of available human dosimetric data. The available cases now cover the duration of pregnancy until the 34th week, with an approximate frequency of at least 1 case every 2 wk (Table 2). Notably, previously available case reports left a gap between the 12th and the 18th weeks. Because the present study includes 8 scans acquired between the 13th and the 18th weeks, a more complete picture of ¹⁸F-FDG uptake during pregnancy is now available. It is also important to note that, after the 34th week, dosimetry is not likely to change significantly because ¹⁸F-FDG uptake by the fetus tends to plateau (Fig. 4). In addition, late-pregnancy dosimetry might be clinically less valuable, because scans can more easily be postponed until delivery or labor can be induced, since the fetus is already viable.

Importantly, all women in this study were scanned with PET/MRI, which allowed a detailed delineation of the fetal body contours and hence dose estimations were arguably more accurate. Such accuracy cannot be achieved by coregistering the PET to an MRI acquired separately because the fetus would have moved in the meantime. In the present study, fetal movements were likely to be minimal because excellent coregistration of MRI and PET was observed in all cases; an example of this is visible in the superposition of the fetal heart on MRI and myocardial ¹⁸F-FDG activity in Figures 2 and 3. In addition to providing an excellent visualization of the fetal body, PET/MRI machines do not deliver to the fetus the additional dose of a CT scan or of a transmission source and therefore, if available, should preferentially be used to image pregnant women.

The MR images allowed drawing of the fetal regions of interest with a high level of confidence, so that inaccuracies due to manual segmentation could be minimized. Of course, because there is no gold standard against which the segmentations can be compared, residual segmentation errors cannot be quantified. However, it is reasonable to assume that fetal regions of interest drawn on an MRI are more accurate than those drawn directly from PET, where fetal contours are not visible (Fig. 1), or CT, where fetal soft tissues are difficult to delineate. The present dose estimates confirmed the low level of radiation absorbed by the fetus when the mother is injected with ¹⁸F-FDG. The highest estimate was observed in early pregnancy (3.2 mGy), but the average of the remaining cases was 1.1 ± 0.5 mGy. The fetus of the woman who had 2 examinations received a cumulative dose of 3.3 mGy. These values would not significantly change if different anthropomorphic phantoms were used (4,11,12) and are more than 1 order of magnitude lower than the threshold for deterministic effects. Although stochastic effects cannot technically be ruled out, no effects have ever been observed for doses this low (13). Taken together, these data suggest that the benefits for both mother and fetus of a clinically appropriate ¹⁸F-FDG PET scan outweigh the hypothetical risks to the fetus theorized by the linear no-threshold model (14,15). Notably, in the case of cervical cancers, the primary tumor size may be evaluated with MRI, but to assess lymph-node involvement the alternative to PET/MRI may be an invasive histologic verification.

To compensate for the incomplete data coming from sparse published case reports, we previously used mathematic modeling to extrapolate the time-integrated coefficients for the whole duration of pregnancy from available human cases (10). We found that the variation of the fetal time-integrated coefficients follows a sigmoid function: after a rapid increase in the second trimester and the beginning of the third, when the fetal mass rapidly increases, the function eventually tends to plateau when the fetus reaches maturity (Fig. 4). Building on that work, the present study prospectively tested whether the doses extrapolated by the function could predict the calculated doses. Our findings demonstrated that the extrapolated doses closely match the calculated doses until the 15th week of pregnancy but underestimate the doses by up to 25% between the 14th and 25th weeks (Table 3). An error of up to 25% could be considered minor compared with the uncertainties in internal dose estimations (16,17) and, indeed, larger differences are observed between the measured values in different fetuses at the same week of pregnancy (Table 2). On the other hand, the actual dose was underestimated in 8 of 11 fetuses, including all the fetuses starting at the 14th week (Table 3), which suggests the possibility of a systematic group difference. One possible source of this difference is ethnicity; specifically, this study comprised Japanese women (average weight, 59 ± 12 kg), whereas the populations used to build the mathematic function comprised American and European women (average weight, 73 ± 13 kg). Another possible reason for a systematic group difference is that previously published cases were acquired mostly with PET, where fetal contours were not visible, or PET/CT, where fetal soft tissues were sometimes difficult to delineate or to differentiate from placental uptake (18); in contrast, all the fetuses in this study were visible in detail thanks to the simultaneous MRI. For example, when creating the original sigmoid curve, 1 data point was automatically eliminated as an outlier during the fitting process (Fig. 4). Because this point was from a PET-only study, it is possible that the segmentation of that fetal body was inaccurate.

Finally, it should be noted that the cases in the present paper, as well as all the other cases published in the literature to date, consist of static images. In consequence, some (conservative) assumptions must be made to calculate the dose, such as considering the physical half-life of ¹⁸F equal to the effective half-life of ¹⁸F-FDG. It is likely that the fetal dose calculated with measured time-integrated coefficients will be slightly lower. If the PET scan of a pregnant woman is planned, we encourage nuclear medicine departments to acquire dynamic images. These would not increase the radiation dose but would allow an even more accurate assessment of fetal dosimetry as well as enable, for the first time, calculation in vivo of the metabolic rate of glucose in the various fetal tissues with full kinetic modeling. Given the absence of radiometabolites, ¹⁸F-FDG activity in the mother’s aorta would provide an excellent image-derived input function. Recent scanners—which allow fast dynamic images of excellent quality to be obtained over the whole body even with lower injected activities (19)—may yield input functions of such quality that they could reliably be used in conjunction with the gold standard of compartmental modeling rather than only with graphical analyses (20). This would enable a deeper understanding of the physiology of glucose utilization in the fetus and its evolution over the duration of pregnancy.