Pilot study of a noninvasive real‐time optical backscatter probe in liver transplantation

Transplantation of severely steatotic donor livers is associated with early allograft dysfunction and poorer graft survival. Histology remains the gold standard diagnostic of donor steatosis despite the lack of consensus definition and its subjective nature. In this prospective observational study of liver transplant patients, we demonstrate the feasibility of using a handheld optical backscatter probe to assess the degree of hepatic steatosis and correlate the backscatter readings with clinical outcomes. The probe is placed on the surface of the liver and emits red and near infrared light from the tip of the device and measures the amount of backscatter of light from liver tissue via two photodiodes. Measurement of optical backscatter (Mantel–Cox P < 0.0001) and histopathological scoring of macrovesicular steatosis (Mantel–Cox P = 0.046) were predictive of 5‐year graft survival. Recipients with early allograft dysfunction defined according to both Olthoff (P = 0.0067) and MEAF score (P = 0.0097) had significantly higher backscatter levels from the donor organ. Backscatter was predictive of graft loss (AUC 0.75, P = 0.0045). This study demonstrates the feasibility of real‐time measurement of optical backscatter in donor livers. Early results indicate readings correlate with steatosis and may give insight to graft outcomes such as early allograft dysfunction and graft loss.


SUMMARY
Transplantation of severely steatotic donor livers is associated with early allograft dysfunction and poorer graft survival. Histology remains the gold standard diagnostic of donor steatosis despite the lack of consensus definition and its subjective nature. In this prospective observational study of liver transplant patients, we demonstrate the feasibility of using a handheld optical backscatter probe to assess the degree of hepatic steatosis and correlate the backscatter readings with clinical outcomes. The probe is placed on the surface of the liver and emits red and near infrared light from the tip of the device and measures the amount of backscatter of light from liver tissue via two photodiodes. Measurement of optical backscatter (Mantel-Cox P < 0.0001) and histopathological scoring of macrovesicular steatosis (Mantel-Cox P = 0.046) were predictive of 5-year graft survival. Recipients with early allograft dysfunction defined according to both Olthoff (P = 0.0067) and MEAF score (P = 0.0097) had significantly higher backscatter levels from the donor organ. Backscatter was predictive of graft loss (AUC 0.75, P = 0.0045). This study demonstrates the feasibility of real-time measurement of optical backscatter in donor livers. Early results indicate readings correlate with steatosis and may give insight to graft outcomes such as early allograft dysfunction and graft loss.

Introduction
Demand for liver transplantation continues to rise with the increase in liver failure seen in the UK [1] being mirrored globally, with over a million people dying of cirrhosis worldwide every year [2]. Liver transplantation remains the only effective treatment for end-stage disease, providing an average of 17-22 years of additional life [1,3,4]. Despite rationing access to the UK waiting list [5], 11% of patients listed in 2016/17 had died and 7% had been removed from the waiting list within 2 years [6]. This figure may increase given the organ shortages occurring currently as a consequence of COVID-19.
The shortfall in donor organs has led to an increase in the use of extended criteria (or marginal) grafts, which are associated with higher rates of early allograft dysfunction and primary nonfunction (PNF) [7,8]. While the concept of these extended criteria livers for transplantation is less well defined than in kidney transplantation, donor factors include graft steatosis, donation after circulatory death (DCD), prolonged ICU stay and older age. The decision about whether to accept a more marginal graft for implantation for a given recipient has to be offset against an increased waiting list mortality from waiting longer for a more optimal graft [9]. Although the visual appearance of the liver (a coarse proxy for steatosis) is known to be associated with poorer outcome [8], predictive models for graft failure have erred away from its inclusion due to its subjective nature and the limitations of any categorical descriptions [8,10].
Transplantation of severely steatotic donor livers is associated with a higher incidence of postoperative complications, early allograft dysfunction (EAD), primary nonfunction (PNF), prolonged ICU stay as well as poorer 1-year graft survival [11][12][13][14][15] Preprocurement ultrasound is unable to accurately or reliably predict the degree of steatosis [22], and while cross-sectional imaging by MRI or CT may allow for more objective quantification of hepatic steatosis [12,23,24], it is difficult to envisage that this will be widely available or cost-effective for the assessment of grafts in such a time-constrained situation. This leaves a surgeon's assessment of steatosis using a combination of visual inspection and palpation, which is unreliable and open to significant bias [22]. An accurate and reproducible real-time test for assessing the degree of steatosis in a donor organ is essential to facilitate safe transplantation, aid research into new models for predicting PNF and EAD accurately, and facilitate informed discussions with patients about risk.
We have previously demonstrated in a preclinical study using optical spectroscopy techniques that backscatter of red and near infrared light from immediately beneath the liver surface showed a correlation coefficient of 0.85 in humans when referenced to clinical haematoxylin and eosin (H&E)-stained biopsies [25]. This led us to develop a portable handheld device which, when placed against the surface of the donor liver, allowed us to evaluate the degree of hepatic steatosis. Here, we report on the pilot study correlating the device's readings with liver transplant outcomes.

Materials and methods
This is a prospective observational cohort study of consecutive patients undergoing liver only transplantation at Addenbrooke's Hospital, Cambridge between August 2011 and May 2014 were recruited to participate in this study; split liver transplants were excluded. All consenting patients who underwent transplantation of a liver alone within the study period were included in the study. Outcome data were collected along with other factors that might predict EAD/PNF such as donor type, donor age and ischaemic time. EAD was defined using both the binary Olthoff criteria [26] and the continuous Model of Early Allograft Function (MEAF) scale [27,28]. Primary nonfunction (PNF) was defined as poor graft function necessitating retransplantation or culminating in death within 14 days, excluding rejection and vascular thrombosis. Cold ischaemic time was defined as the time between commencement of cold perfusion in the donor and reperfusion in the recipient.
The implanting surgeon, blinded to the optical backscatter readouts, was asked to grade the degree of steatosis as none, mild, moderate or severe based on visual inspection and palpation during preimplantation benchwork; these categories form part of the returns used by the National Health Service Blood and Transplant organisation in the UK.

The probe
In previously published work, we have described in detail the principles of and technology underpinning red and near infrared light backscatter measurements to assess hepatic steatosis [25]. Briefly, a custom-made diffuse reflectance (DF) optical fibre probe attached to a spectrometer was used to measure tissue absorbance and backscatter by the liver.
For this work, a compact and portable handheld probe was developed (Medicines & Healthcare products Regulatory Agency (MHRA) approval CI/2011/0004). The probe emits red and near infrared light from the tip of the device via two light-emitting diodes (LEDs) linked via optical fibres (see Figure S1) and measures the amount of backscatter of light from approximately 2mm into the liver tissue via two photodiodes, also coupled via optical fibres to the tip of the device. Measurements are taken by placing the device against the surface of the liver, with the amount of backscatter represented on a digital display in arbitrary units. The readings are automatically logged to memory in the device, with a timestamp, for later upload to a PC. To ensure a sterile measurement, the tip of the device includes a disposable cap, which is attached to the probe before use.
The probe was placed against the surface of the liver and readings were taken from 4 pre-specified sites (2 on the right lobe, 2 on the left) from each donor liver during retrieval, pre-implantation benchwork and following reperfusion. The mean reading from the 4 sites was used in the analysis. The absorption largely reflects blood within the liver, while scattering is specific for the size, density and cellular constituents of tissue. We have previously shown that differences in scatter between livers of different patients correlates with differences in the lipid content (steatosis) [25], but equally other cellular processes may affect backscatter.

Histopathology
Pre-implantation core biopsies of the donor livers were taken on the backtable. These biopsies were immediately split with half being snap frozen, stored at À80°C to subsequently allow frozen sections to be cut for oil red O staining and the remainder formalin-fixed and processed to paraffin with sections cut and stained with haematoxylin and eosin (H&E). As previously described [25,29], the extent of steatosis and reperfusion injury were scored by a histopathologist with a special interest in liver disease, blinded to the macroscopic description from the surgeon and optical backscatter results. Macrovesicular steatosis was further subdivided as either large or small droplet in line with others [30,31]. Large droplet macrovesicular steatosis was characterised by a single large fat droplet in hepatocyte cytoplasm, displacing the nucleus to the edge of the cell. This was quantified based on the percentage of large droplet fat occupying the surface area of the parenchyma and given a score 0 to 3 (Table S1). This was subdivided as none or mild (score 0 to 1) or moderate to severe (score greater than or equal to 2). Small droplet macrovesicular steatosis (termed vacuolation by ourselves) consists of multiple small and tiny lipid droplets in the cytoplasm, all being less than size of the nucleus, which retained its central position; this was graded using a score 0 to 2 depending on their extent (Table S1) and termed none to mild (score 0 to 1) or moderate to severe (score 2). Total Fat score was a sum of the small and large droplet scores with none to mild (score 0 to 2) or moderate to severe (score greater than or equal to 3). Microvesicular steatosis refers to the intracytoplasmic accumulation of tiny vesicles within hepatocytes and is a result of severe mitochondrial dysfunction [32]. Given that it highly unlikely that a liver from an affected patient would even be considered as a donor liver [30] and is typically seen in < 1% of transplanted livers [33], we did not formally score the degree of microvesicular steatosis.

Statistics
Groups were analysed with the aid of Prism 8 for Mac OSX (Graphpad Software, La Jolla, USA); statistical methods are referred to specifically in the results section. Briefly, transplant characteristics were compared using Fisher's exact test, chi-squared or the Mann-Whitney test, as appropriate. Concordance was correlated using the methods described by Lin [34]. Unadjusted graft and patient survival were displayed using Kaplan-Meier plots and curves compared by Mantel-Cox analysis.

Ethical approval
Prospective ethical approval was granted for the project by the Regional Ethics Committee (Ref: 10/H0308/94). The study was conducted in accordance with the Declaration of Helsinki and the principles of Good Clinical Practice. No organs from executed prisoners were used in this study. Hospital in the study period and consented to the study. No patients had undergone any form of machine perfusion in this study. Donor and recipient demographics and outcomes are summarised in Supplemental  Tables 2 and 3.

Of
The overall 1-and 5-year patient survival was 91.7% and 86.2%, respectively, while the 1-and 5-year deathcensored graft survival was 92.6% and 89.2% in this mixed cohort of DCD and DBD grafts ( Figure S2). The 1-and 5-year graft and patient survival data are summarised in Table 1.

Backscatter readings
There was excellent concordance between optical readings taken in the donor, on the backtable and postreperfusion with concordance greater than 0.85 (data summarised in Table 2). There were no grafts transplanted with a histological large droplet macrovesicular steatosis score of 3 in this study. While there was no significant difference in survival between patient or graft survival depending upon the severity of large droplet macrovesicular steatosis (Table 1 and Fig. 1a, Mantel-Cox P = 0.70 and P = 0.24), there was a significant decrease in survival of patients receiving allografts with more severe small droplet macrovesicular steatosis (Fig. 1b, P = 0.046 and P = 0.041). A higher total macrovesicular steatosis score in the donor organ was associated with poorer graft survival (P = 0.046), but not patient survival (P = 0.098).
Backscatter readings were significantly higher in grafts with more extensive large droplet (P < 0.0001) and small droplet macrovesicular steatosis (P = 0.0001) (Fig. 1). Oil Red O staining also strongly correlated with backscatter readings (Pearson's Rank 0.53 (95% CI 0.35-0.67, R 2 = 0.28, P < 0.0001). In general, increased severity of steatosis as judged by the surgeon was associated with an increase in both the large (P = 0.0002) and small droplet macrovesicular steatosis histological score (P = 0.0007, Figure S3), although the overall concordance between the surgeon and histology was relatively poor (coefficient 0.41).

Graft survival and early allograft dysfunction
In those grafts with EAD defined according to Olthoff criteria, the 1-year graft survival was 86.5% compared 94.5% (Fig. 2a); the backscatter reading was significantly greater in these livers (P = 0.0067).

Acute kidney injury after liver transplantation
The development of acute kidney injury after liver transplantation was associated with a reduction in 1year graft survival from 97.7% to 90.2% and patient survival of 93.1% to 86.8% (see Table 1 and Figure S4). There was not a significantly increased MEAF score between the 2 groups (P = 0.09), but the backscatter reading was significantly higher (P = 0.0027). Simple logistic regression to look at graft loss, using only backscatter on its own performed similarly at predicting graft loss (AUC 0.75 (0.58-0.91), P = 0.0045) (Fig. 5). The odds ratio for graft loss was 1.004 (95% CI 1.00-1.01) for every unit increase in backscatter.

Discussion
Here we demonstrate that real-time measurements of backscatter of red and near infrared light from the liver whilst in the donor, on the backtable and after implantation in the recipient is a feasible approach to assessing in real-time the degree of hepatic steatosis in the setting of liver transplantation. As we had previously seen in a preclinical study of murine and human liver specimens [25], backscatter strongly correlated with the extent of hepatic steatosis as determined by Oil Red O staining (Pearson's Rank 0.53, P < 0.0001) and as scored by a transplant histopathologist (Fig. 2). While increased severity of steatosis as judged by the surgeon was associated with an increase in both the large and small droplet macrovesicular steatosis histological score ( Figure S3), the overall concordance between the surgeon and histology was relatively poor (coefficient 0.41). This probe, therefore, may help to overcome the inherent problem of inter-observer bias seen when relying on arbitrary macroscopic inspection by a surgeon or microscopic evaluation by a histopathologist. While both may remain important within an individual centre, they prevent standardisation of reporting the degree of steatosis in research and across clinical trials, where heterogeneity in approach and inter-observer bias can make outcomes difficult to interpret [18,22].
As well as correlating with the extent of steatosis, measurement of optical backscatter correlated with early allograft dysfunction according to both the Olthoff and MEAF parameters (Fig. 2) as well as with acute kidney injury post-liver transplantation ( Figure S4). More complex multiple logistic regression analysis of this data is limited by the sample size, however, we demonstrated in principle how backscatter could be incorporated into a predictive algorithm utilising in this case donor factors identified at the time of procurement/implantation looking at graft failure as the endpoint. We demonstrated that backscatter measurements were predictive of graft loss (Fig. 5). Further evaluation will require large numbers from a multi-centre study to validate or refute these findings and incorporate both donor and recipient factors as well as other novel readouts into a highly predictive algorithm that will help quantify risk the of a given allograft to a particular recipient. Increased backscatter was not predictive of patient survival, in part due to the ready availability of early retransplantation at that time in the UK, and also potentially the small sample size.
While we utilised a categorical scoring system for assessing the extent of steatosis, others have recently developed a digital algorithm to quantify steatosis in tissue sections, which may make histological evaluation in future clinical studies more sensitive [35], though its usefulness in preimplantation decision-making may be limited by its inherent retrospective nature, and also the time taken to prepare and scan a sample.
Evers et al have also previously demonstrated good concordance with the histological quantification of fat by a similar approach in the context of liver resection surgery [36]. Fibroscan, CT and MRI have also been used successfully as a noninvasive tools for quantifying fat in the field of nonalcoholic fatty liver disease [37], but their role may be limited in the setting of liver transplantation by their portability, availability 24 hours per day across all potential donor hospitals, national laws about pre-mortem interventions in donors and cost. Other groups have also demonstrated the effectiveness of analysis of smartphone photographs and digital analysis software to assess the extent of macrovesicular Figure 2 Early Allograft Dysfunction by Olthoff Criteria and MEAF grouping. In those grafts with EAD defined according to Olthoff criteria, the 1-year graft survival was 86.5% compared to those that did not meet the criteria 94.5% (a); this difference was not statistically significant (summarised in Table 1) although this may be related to sample size. The backscatter readings were significantly higher in those livers (P = 0.0067) meeting the Olthoff Criteria (a). When looking at survival by MEAF grouping, those with a score less than 4 had a 100% 1-year patient and 96.7% 1-year graft survival compared to patient survival of 66.7% and 50%, respectively, for those with a score greater than 9 (b). Increased MEAF was associated with increased graft or patient survival (Mantel-Cox P = 0.031 and P = 0.55, respectively). A higher MEAF score was associated with a significantly greater backscatter (Kruskal-Wallis P = 0.0097) (b). Increased backscatter readings subsequently correlated with increased MEAF scores (Pearson's r = 0.33 (95% CI 0.14 to 0.49), R 2 = 0.11, P = 0.0011) (c); this was particularly true in organs from DCD donors (r = 0.77 (95% CI 0.50 to 0.90), R 2 = 0.59, P < 0.0001) compared to DBD (r = 0.23 (95% CI 0.0019 to 0.43),  While this study utilises a custom-built prototype, it is likely that a commercially available device could be developed and would prove to be cost-effective while not causing an increase in cold ischaemia (in contrast to biopsy examination) and could be utilised in centres who do not have 24-hour access to a transplant histopathologist. It also avoids the potential risk of bleeding or bile leak from the biopsy site.
While declining the offer of a steatotic liver has been shown to increase an individual's waiting list mortality [41], the unpredictable response of steatotic Figure 3 Backscatter by surgeon assessment. Surgeon assessment of grafts did not predict differences in 1-year graft (93.8%) or patient (91.1%) survival in those deemed 'healthy' compared to those that were 'suboptimal' 93.0% and 93.0%, respectively (a). There was a significant difference between the backscatter readings between 'suboptimal' and 'healthy' livers (Mann-Whitney P = 0.0004) (a). No grafts deemed to be severely steatotic by the implanting surgeon were implanted during this study. 1-year graft survival for those deemed to not be fatty was 96% compared to 93.1% (mildly steatotic) and 86.7% (moderately steatotic), with corresponding 1-year patient survival of 92.1%, 90.0% and 86.7% (b). The backscatter reading was significantly higher in the moderately steatotic livers compared to those with minimal or no fat (Kruskal-Wallis P = 0.0054) (b).

716
Transplant livers to reperfusion, with an increased severity of ischaemia-reperfusion injury (IRI) and subsequently increased rates of PNF, EAD and post-liver transplant acute kidney injury mean that there is an understandable reluctance to routinely transplant such livers [21,42]. As the demand for organ transplantation continues to remain high and the epidemic of obesity in the west is resulting in higher rates of steatosis in donor organs [43,44], we will inevitably need to implant more steatotic livers in the future. Others have shown that one possible solution to overcoming the excess risk of a steatotic organ is to allocate steatotic organs to 'preferred recipients' (defined as first-time recipients with a MELD 15-34, without primary biliary cirrhosis and not on life support prior to transplantation), as these recipients have no significant increase in mortality or graft loss when receiving steatotic compared to nonsteatotic livers [45]. While their study was necessarily performed retrospectively on biopsies using registry data, one could envisage that backscatter in the donor or on the backtable could be used prospectively to guide these decisions, without the need to wait for biopsy results. The continuous nature and spread of the potential backscatter data is also such that the allocation process could be less dichotomous and identify a greater range of potential recipients that would benefit from a given organ and help identify risk in each individual.
As well as matching the 'high risk' organ to the 'low risk' recipient, another strategy to mitigate the excess risk of steatotic livers is to utilise novel technology or therapeutics to identify which livers are safe to implant and identify those which need some other intervention, such as ex situ machine perfusion [46][47][48][49][50], that is personalising or targeting therapy for a given donor liver. Assessment of backscatter seems to be one such potential objective strategy to stratify risk and rather than being used in isolation, we envisage that optical backscatter measurements would be incorporated in a more complex model utilising all available data on the donor and recipient (Fig. 6) to really inform the patient and surgeon about personalised risk, especially when coupled with better modelling of individual waiting list mortality. With the advent of machine perfusion, these readings could be used by the retrieval centre and/or national organ allocation service to stratify organs into 'safe to transplant', 'not safe to transplant', 'needs further viability testing' [48] and in the future an Figure 4 Backscatter readings and survival. Kaplan-Meier plots of graft and patient survival comparing survival in cohorts of patients backscatter readings greater or less than 100. Backscatter readings > 100 were associated with worse graft (Mantel-Cox, P < 0.0001), but not patient survival (P = 0.085) (summarised in Table 3). additional arm that would recommend directed therapy (see Fig. 6). Furthermore, this technology may help validate the effectiveness of ex situ 'defatting' strategies in a given liver undergoing machine perfusion that are currently being developed [51,52], without the need for serial biopsies.
In conclusion, the data from this pilot study are promising, but needs more extensive validation alongside other novel noninvasive real-time approaches to generate robust data to support their further use and/or generate more accurate predictive models. If further validated, measuring optical backscatter in donor livers may have a role in the safe allocation of livers for transplantation and inform discussions between clinicians and patients about the risk of a given donor organ [53]. In addition, it may allow increased utilisation by helping to determine a subset of livers requiring specific pre-treatments before transplantation, such as targeted drug therapy or defatting during ex situ perfusion [54,55].
Authorship JR, LR, AB, CW and PR have made a substantial contribution to the conception, data collection, analysis and writing of this manuscript. PR designed and produced the probe. SD performed histological analyses and aided with analysis and writing of this manuscript. JM and AF made a significant contribution to the data collection.

Funding
We would like to acknowledge support from the Evelyn Trust who funded this study as well as the National Institute for Health Research (NIHR) Blood and Transplant Research Unit (NIHR BTRU) in Organ Donation and Transplantation at the University of Cambridge in collaboration with Newcastle University and in partnership with NHS Blood and Transplant (NHSBT). JR was also supported by a NIHR Academic Clinical Lectureship. The Human Research Tissue Bank is supported by the NIHR Cambridge Biomedical Research Centre. The University of Cambridge has received salary support in respect of CJEW from the NHS in the East of England through the Clinical Academic Reserve.

Conflict of interest
While working on this study, LR was wholly employed by the University of Cambridge / Cambridge University Figure 5 Logistic regression analysis of graft loss. Multiple logistic regression modelling of factors known preimplantation (donor age, donor BMI, donor type and backscatter reading) was performed and used to generate model to predict graft loss (Logit[P (graft loss)] = (donor age x 0.05853) + (donor BMI x À0.01520) + (Donor type (0 = DBD, 1 = DCD) x À0.01323)+ (Backscatter x 0.009219) -5.817). Receiver operating characteristic (ROC) curves of this analysis were plotted (a), and area under the curve (AUC) analysis showed the model to perform reasonably well (AUC 0.75 (95% CI 0.60-0.90), P = 0.0072). Simple logistic regression to look at graft loss, using only backscatter alone was also performed and ROC curve was plotted (b). This performed similarly at predicting graft loss (AUC 0.75 (95% CI 0.58-0.91), P = 0.0045). Hospitals. She has subsequently moved to work for the machine perfusion device company OrganOx, who have no commercial or other interest in this body of work and as such LR feels that there is no significant conflict of interest to declare. The remaining authors have no other conflicts of interest to declare.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article. Figure S1. Diffuse reflectance (DF) optical fibre probe. Figure S2. Overall Graft and Patient survival. Figure S3. Histological Scoring and Surgeon's Assessment of steatosis. Figure S4. Histological Scoring and Surgeon's Assessment of steatosis. Table S1. Histological Scoring. Table S2. Donor demographics. Table S3. Recipient details.