The Wits Transplant unit, which forms part of the Wits Medical Centre, is the largest transplantation unit in sub-Saharan Africa and the only centre that performs Living Donor Liver transplants (LDLT) in the region. To combat the global shortage of cadaveric donors, a problem exacerbated during the COVID-19 pandemic, the Wits transplant unit makes use of split liver grafts, ABO incompatible transplants as well as running an LDLT programme. Living Donor Liver transplants allow healthy individuals (related or not) to donate part of their liver to patients with end-stage liver disease (ESLD). Despite this, in 2020, there was a 20% paediatric and 11% adult mortality rate of patients on the transplant waiting list.^1,2

To ensure favourable transplant outcomes in LDLT, the donor needs to retain 30% - 40% of their total liver mass and the recipient needs a graft recipient weight ratio (GRWR) of at least 0.8%.^3,4,5,6,7 This is to mitigate the risk of complications, such as small-for-size in the recipient or liver insufficiency in the donor. Both conditions arise because of insufficient liver tissue to meet the body’s metabolic and synthetic demands.³

Computed tomography (CT) is utilised preoperatively to determine the eligibility of potential donors and to predict liver graft weight, both being pivotal for successful outcomes in LDLT.^7,8 The estimated graft volume (EGV) is computed using liver volumetry software and is then converted into an estimated graft weight (EGW) in a 1:1 ratio. This, however, is based on the water displacement of cirrhotic livers and therefore, does not hold true for donors of healthy liver tissue.⁹

It is well established that CT-derived EGV predicts actual graft weight (AGW) with acceptable accuracy as there is a linear correlation between the two. Many studies have sought to further improve EGV predictions of graft weight using formulas and coefficient factors.¹⁰ Saleem et al.¹⁰ found that in Egyptian living donors, 1 mL of liver volume equated to 0.96 g of liver weight, thereby making the CT volumetry predictions more accurate. Similarly, Yoneyama et al.⁹ found that the density of liver in right lobe grafts was 0.84 g/mL and 0.85 g/mL in left lobe grafts in Japanese donors.

In this study, the aim was to assess the relationship between EGV and AGW in a South African context and to test if combining the CT volumetry values with a population-specific correlation coefficient will be more accurate in predicting EGW.

In total, 117 liver donors were included in the study (47 males and 70 females); the demographics are summarised in Table 1. Of all the LDLT, 83% were adult to child donations and 17% adult to adult donations. The most frequent graft type harvested was LL (73.5%) followed by R (12.8%) (Figure 1).

The EGV of the grafts averaged 388 mL ± 249 mL with a range of 127 mL – 1610 mL. The AGW averaged 353 g ± 184 g with a range of 160 g – 1300 g. The mean EGV was statistically significantly low compared to that of the AGW (p < 0.001).

The relationship between EGV and AGW was significantly linear and statistically significant (r = 0.95, p < 0.001). The scatter plot for EGV and AGW is demonstrated in Figure 2 with a trend line equation of y = 0.71x, R² = 0.91, p < 0.001. This implies that multiplying the CT-generated graft volume by 0.71 will more accurately estimate the graft weight. Overall, looking at percentage deviation, CT volumetry overestimated the AGW 61% of the time and underestimated the graft weight 39% of the time.

Many methods of predicting graft weight have been proposed over the years; however, CT volumetry remains the gold standard. The Donald Gordon Transplant unit, like many other transplant institutions, operates under the premise that 1 mL of liver volume equates to 1 g of liver weight. Many studies have, however, shown this not to be the case, as liver tissue from healthy donors tends to have a density of less than 1 g/mL.

In this study, the liver density of healthy donors was 0.71 g/mL, and overall, CT volumetry was much more likely to overestimate the graft weight (61% frequency). Specifically looking at graft type, 100% of LC and 93% of R grafts were overestimated, whereas LL had almost equal amounts of over- and underestimation. Overestimation can be problematic in cases where the predicted graft size is borderline, and the recipient may subsequently end up with a GRWR of less than 0.8%.

Kiuchi et al.¹¹ looked at the impact that graft mismatching had on graft survival. Small for size grafts, with a GRWR of < 1%, had a significantly lower graft survival than larger ones. Recipient mortality or graft loss in these cases was attributed to sepsis, graft non-function, cerebral infarction associated with immunosuppressive drugs and chronic rejection. In addition, recipients presenting in acute fulminant liver failure or advanced chronic liver disease may not have the functional reserve to cope with smaller grafts in the acute postoperative period.

Of the 117 transplants, there were four recipients (3.4%) who ended up with a GRWR of less than 0.8 (Table 2). These cases involved LL and LC grafts. In such instances, there are surgical techniques that can be used to moderate portal venous inflow such as a hemi-portocaval shunt or splenic artery ligation.¹² These procedures mitigate the damage done by the shearing forces associated with portal hyperperfusion and provide the graft with a better chance of survival.

Underestimation and large graft size can be problematic in paediatric recipients who weigh less than 10 kg. In this population, the native portal vein may be too small to perfuse a large graft. Large grafts also present difficulty in the primary closure of the abdominal wall post-transplant and result in raised intra-abdominal pressure.¹³ This is the so-called ‘large for size’ syndrome. Another consideration is the shape of the graft. Schukfeh et al. showed that graft weight, particularly a thicker ventral dorsal diameter, had a decreased patient survival.¹³ It is recommended that a GRWR of less than 3–3.5 is used to mitigate the risk of large for size syndrome in this patient population.¹²

Besides the conversion error of assuming 1 g/mL in healthy liver tissue, there may be several other factors that account for the discrepancies between the predicted and the AGW. These can be broadly categorised into radiological and surgical factors that affect the EGV and AGW, respectively.

The radiological factors that affect the EGV include the scanning protocol and acquisition, the software used to segment the liver and interobserver variation. Lim et al.³ found that CT volumetry calculated using non-contrasted versus contrasted venous phase sequences leads to less overestimation of size. Despite this, the portal venous phase is still preferred by radiologists as it better delineates the vascular and hepatic anatomy needed for segmentation. Additionally, the slice thickness also plays a role. The smaller the slice, the more accurate the volume assessment. This, however, is more time-consuming, especially if using manual volumetric techniques as there are many more slices to mark and contour.^3,14,15 It is generally accepted that the use of 6 mm slices has adequate accuracy without taking too much time to segment.³ There is segmentation software that includes and others that exclude vasculature in the assessment of volume. Software that includes vessels tends to overestimate, whereas vessel-free volumetry tends to underestimate weight.¹ Lastly, interobserver variation naturally plays a role in volume prediction in everyday practice as there are several radiologists who report on these cases.

Surgically, the transection plane of the graft may not correlate exactly to the segmentation line on CT leading to EGV/AGW mismatch. There has also been a suggestion that the preservation solution used to flush the graft may influence the graft weight.^9,14 Histidine-tryptophan-ketoglutarate solution used by Yoneyama et al. as well as by the Wits Donald Gordon transplant unit causes graft oedema. However, given that the graft is weighed soon after being flushed, this effect is likely to be negligible.⁹

In the context of LDLT in South Africa, a correlation coefficient of 0.71 multiplied by the EGV will lead to more accurate predictions of graft weight at our institution. Incorporating this into clinical practice should assist with donor and graft selection and decrease the chance of inappropriate GRWR ratios. The findings of this study are only applicable to transplants at our institution and in a South African population because of biases introduced by different CT machines and protocols, differing segmentation software, surgical protocols and population characteristics.