As discussed in the previous section, there are many reasons why histopathology is an imperfect measure of liver stiffness against which to validate MR elastography. The effects of some of these factors can be managed by controlling the relevant biomarker and reference standard conditions to make them both as likely as possible to reflect the desired underlying process. As an example for MRE, excluding livers from validation studies where there is known iron accumulation (Fig. 12.4) or regions of focal confluent fibrosis (on imaging) would reduce sampling variation. Avoiding cases with other obvious biomarker-confounding factors (Fig. 12.5) such as acute inflammation (e.g. acute hepatitis) or where elevated vascular or biliary tract pressure changes are likely (e.g. right heart failure, bile duct obstruction) could also improve the validation process. Using pathologist consensus and a single “fibrosis” grading scale may improve reliability and repeatability of the standard. Given the subjective and non-linear nature of the established grading scales for fibrosis, an alternative approach to improve the reference standard that has been tried is semiautomated quantitative collagen estimation [16, 46] within a selected microscope field of view. This is achieved using specific collagen stains such as sirius red and automated microscopy and image analysis methods to quantify the proportion of stain present as a percentage of the whole field of view [21]. Several authors have considered direct mechanical stiffness measurements in post-mortem livers, but clearly the lack of vascular perfusion and alteration of tissue mechanical properties with temperature make this approach unlikely to be comparable to in vivo measurements. A method for mechanically assessing the liver stiffness at open surgery has been developed [44], but as the authors discuss in some detail, their results do not correlate well with other approaches including MR methods. This is partly owing to their tube aspiration method being influenced by the constraining liver capsule. It is also not yet fully understood how liver stiffness is influenced by the “closed” abdomen and intra-abdominal pressure changes quite apart from possible influences of anaesthetic drugs on hepatic haemodynamics. Finally statistical approaches to utilising more than one imperfect reference standard have been applied to achieve liver fibrosis validation with mixed results [53, 54].

Training is of crucial importance in maintaining a high adenoma detection rate, and improving training can have a positive effect [31]. In the UK, the Joint Advisory Group (JAG) on GI endoscopy sets standards for competence in GI endoscopy and quality assures endoscopy units, training and services. As an example of how JAG monitors standards, research using the JAG Endoscopy Training System (JETS) database has provided evidence that the current requirement for trainee colonoscopists to complete 200 procedures prior to assessment may be too few [70]. In the future, this research is likely to drive up the minimum requirements for independent practice.

Various technical improvements in colonoscopy have been used in attempts to drive up the adenoma detection rate. Chromoendoscopy describes the use of dye spraying during colonoscopy and is a way of improving sensitivity for diminutive polyps [33]. Technique changes, for example, changing the patient’s position during the examination [33], as well as the use of accessories such as the “Endocuff” [35], can also improve the adenoma detection rate.

In addition to improving the “gold standard”, the new biomarker can be validated against itself over time. For example, a polyp seen at CTC in the same place but larger over time is likely to be a genuine finding even if not seen at colonoscopy. This can be used as justification for repeating invasive tests such as colonoscopy [54].

CT perfusion biomarkers are largely based on pharmacokinetic models, and because there are several different models, different analysis software and potentially different methods of data acquisition, it is important to control for these. Most research groups have developed a specific technique often using a specific CT system to ensure acceptable repeatability and reliability of the biomarker metrics. It is of interest that the simplest least processed metric – the integrated area under the time-concentration curve (IAUCC) – is often the metric with highest correlation in many research studies rather than the model-based calculated parameters such as Ktrans, distribution volume, etc.

Having established a repeatable CT perfusion biomarker then, the problem of validation arises as the best techniques are not practical or ethical in humans in vivo. Animal-based validation has been performed successfully using microsphere techniques giving some confidence in the technique as discussed earlier, but in man, imperfect standards such as ¹⁵O or ⁸²Rb tracer methods must be used. As these typically use different tracers (e.g. freely diffusible H₂ ¹⁵O vs extracellular iodine-based CT tracers), models and different metrics, direct comparison for validation is very difficult, and unsurprisingly results often vary substantially [22, 23]. In addition other factors may influence the reference standard results, for example, radioactive tracer dose variation, cardiac output, arterial blood sampling variation and filtering and post-processing algorithms. Instead of absolute metrics, those based on pre- and post-intervention ratio changes of the biomarker can be compared with the same changes in a reference standard, and this may prove to be the best in vivo validation that is available. As a result of these challenges and practical issues, the validation of CT perfusion in man has been severely limited. However, based on the evidence from animal work and clinical studies, and despite the relatively poor validation, there are many advocating [33] that the technique can still be successfully “qualified” for use as a tumour biomarker.

MRE has been extensively validated using histopathologic fibrosis stage as a gold standard, but awaits qualification for its use in predicting clinical endpoints such as development of HCC, liver decompensation or death. Given that the histopathologic fibrosis stage has only relatively recently been qualified against these clinical endpoints [14], it is perhaps not surprising that MRE awaits this step. There is some initial evidence that MRE-measured liver stiffness may be a risk factor for developing HCC in patients with chronic liver disease referred for MRI following an abnormal ultrasound [44]. Although this does not qualify the technique for the prediction of HCC development, it is at least encouraging. Further support can be drawn from qualification of similar biomarkers, in this case the demonstration that ultrasound elastography is predictive of clinical endpoints [50, 63], but this does not obviate the need for separate qualification of MRE. Several studies have also demonstrated that liver MRE could in future be qualified as a predictor of portal hypertension and variceal haemorrhage [59, 62, 67]. Given the limitations of histopathological staging, it is probably appropriate that qualification of hepatic MRE is explored until a better reference standard for validation is developed.

CTC has been extensively validated in patients undergoing both CTC and optical colonoscopy (using optical colonoscopy as the gold standard) [3, 27, 30, 38]. Data to support qualification against hard clinical endpoints are emerging but as yet are incomplete. In patients undergoing CTC to clear the colon proximal to an obstructing tumour, there is evidence using the resection specimen as the gold standard that CTC is 100 % sensitive for malignancy but only 70 % sensitive for lesions ≥ 6 mm [51].

Qualification of CTC for its use more widely in symptomatic patients or as a screening tool is lacking. The endoscopic removal of colonic adenomas has been shown to reduce deaths from colorectal cancer [60]. The effect of carrying out CTC (and subsequent endoscopic polypectomy) on cancer-specific or all-cause mortality in symptomatic or asymptomatic patients has yet to be reported – this is likely a reflection of the relative novelty of the technique and the time it would take for meaningful results to emerge (i.e. at least 5 years). As there are ongoing trials evaluating CTC in screening, it may be that this is the next context for its qualification [11].

Owing to the difficulties of validating CT perfusion metrics in vivo in humans and the complexity of the various models and metrics employed, there has been a tendency to qualify CT perfusion metrics against clinical endpoints. Often these studies have involved relatively small cohorts and several authors [19] consider that further studies using much large cohorts are needed to properly qualify CT perfusion metrics for clinical use. In order to try and standardise as well as reduce the complexity and variability of CT perfusion metrics, several groups have also issued guidelines on the technique and analysis [41].

Many of these relatively small studies have demonstrated that CT perfusion metrics that indicate “high” levels of tumour perfusion at baseline correlate with a relatively good response to therapy and overall outcomes, for example, in breast cancer [37]. The survival and response of liver metastases to radioembolisation has been correlated with baseline CT perfusion metrics [43]. Perfusion metrics have also been used to predict survival in gliomas [73] and outcomes in head and neck tumours [5]. Other qualification studies include stroke outcomes which have been correlated with CT perfusion [52] and the risk of hyperperfusion syndrome after carotid stenting [75].