To date, seven studies have been published as full papers concerning the non-invasive diagnosis of the presence of any OV or large OV in patients with liver cirrhosis.11–17
The reason for this effort is simple: the number of patients undergoing screening for the presence of OV is likely going to increase in the near future as a result of the growing pool of patients with chronic liver disease.33,34
Therefore, there is a particular need for non-invasive predictors of the presence of OV as they might help relieve medical, social, and economic costs. Nevertheless, most of the studies concerning the non-invasive diagnosis of OV were performed on a particular subgroup of patients (that is, patients who were going to be placed on a liver transplantation waiting list) while some of the studies lacked uniformity in OV classification or adequate statistical analysis, and only one study analysed patients with compensated disease.11–14,16,17
Almost all of the studies were retrospective, although the only prospective study obtained results that were no different from those obtained in retrospective studies.15
In general, most identified decreased platelet count and splenomegaly as non-invasive predictors of the presence of OV. In our study, we considered only simple, commonly available, reproducible parameters as we believe that other features that had previously been identified as non-invasive predictors of OV were less reproducible in clinical practice,35
were subject to interobserver variability,36
or were assessed in different ways even within the same study. Moreover, we introduced a new parameter (that is, the platelet count/spleen diameter ratio). This parameter was chosen as it allows us to identify the degree of thrombocytopenia which most likely depends on hypersplenism. We measured the spleen bipolar diameter using ultrasonography as this parameter is easily obtainable, reproducible, and non-invasive, and because liver ultrasonography is routinely performed on patients with liver cirrhosis.
The distribution of the study population was homogeneous and representative of the population of cirrhotic patients seen in clinical practice, thus biases caused by selection of particular subgroups of patients were avoided. Diagnosis and classification of OV was made in the same endoscopy centre and using a single classification. We focused on the presence of any OV rather than on the presence of large OV as this is the first step in the diagnostic/prognostic workup of the patients and allows decision making processes (surveillance, repeat endoscopy at predetermined intervals, start therapy), while focusing on patients with large OV would only miss an important subset of patients requiring medical counselling. Moreover, analysis of the presence/absence of OV prevents misinterpretation of data and allows generalisation of the results.
In the first part of the study we found that parameters linked to portal hypertension (platelet count, spleen diameter) as well as those linked to liver dysfunction or advanced disease (serum bilirubin, prothrombin activity, Child-Pugh classes) were associated with the presence of any OV, as observed in previous studies. However, the original finding of our study was that a composite parameter (that is, the platelet count/spleen diameter ratio) showed very good results in discriminating between patients with NOV and those with OV.
We believe that the use of this ratio is of interest and is not redundant, and this hypothesis is supported by a number of both clinical and statistical reasons. Firstly, from a clinical point of view, platelet count may decrease for several reasons in patients with chronic liver disease.24
Thus the use of platelet count alone as a non-invasive predictor of OV can be misleading and cannot be solely attributed to portal hypertension. Indeed, the use of the platelet count/spleen diameter ratio bypasses this possible drawback since it “normalises” platelet count to splenic sequestration, most likely representing the aliquot of thrombocytopenia caused by portal hypertension. Secondly, from a statistical point of view, the platelet count/spleen diameter ratio was the only parameter independently associated with the presence of OV that was selected by a multivariate analysis which also included the single parameters.
In the second part of the study our aim was to evaluate whether the predictive power of the platelet count/spleen diameter ratio cut off identified in the first part of the study was able to reproduce its results in a subsequent different, although related, series of patients who represented the same population. Indeed, reproducibility of a clinical decision aid “requires the system to replicate its accuracy in patients who were not included in development of the system but who are from the same underlying population”.37
Under these conditions we demonstrated that the use of a platelet count/spleen diameter ratio cut off of 909 maintained its accuracy in the non-invasive diagnosis of OV. Indeed, even in the second group, all of the patients who had a ratio >909 were free from OV while 71% of patients with a ratio below this cut off had OV.
Analysis of patients with compensated disease was of particular importance. In fact, predicting the presence of OV in patients with no signs of decompensation could be especially useful and its clinical importance has recently been emphasised.5,15,23,26,28
We observed that even among patients with compensated disease, platelet count/spleen diameter ratios were significantly different between NOV and OV patients, and the platelet count/spleen diameter ratio had the highest accuracy for identifying patients with OV. Most importantly, this parameter proved to be independent of the Child-Pugh classification allowing a more confident use even in patients with compensated disease.
In this day and age, clinical decision aids must prove to be not only clinically useful but also cost effective in order to be widely accepted. In our study we showed that the use of the platelet count/spleen diameter ratio would have avoided performing unnecessary endoscopies in all patients with a cut off >909 without running the risk of not diagnosing OV. As far as cost benefit analysis is concerned, applying the “platelet count/spleen diameter ratio strategy” would lower the cost of OV screening in patients with cirrhosis (fig 3). In practice, the unitary cost per patient with OV identified during screening would be 249 Euros by applying the “scope all strategy” and 181 Euros by using the “platelet count/spleen diameter ratio strategy”. Generalising these results to the screening of 10 000 patients with liver cirrhosis per year, application of the “platelet count/spleen diameter ratio strategy” policy would save 405 000 Euros per year, and would avoid performing unnecessary endoscopies in 2700 patients. In our analysis we hypothesised no extra cost for ultrasonographic spleen diameter measurement, and this hypothesis has a solid practical basis. In fact, cirrhotic patients usually undergo annual/biannual abdominal ultrasonography as part of surveillance programmes for hepatocellular carcinoma. Therefore, follow up of cirrhotic patients to identify liver cancer allows us the opportunity to determine the platelet count/spleen diameter ratio at least annually or biannually. As the yearly incidence of OV in patients with liver cirrhosis is approximately 5%,38,39
we feel that this combined evaluation would not miss OV identification, and would be cost effective also.
Finally, this study undoubtedly has some shortcomings, such as its retrospective setting and its validation. None the less, the only prospective study performed concerning this topic obtained results that were no different from ours or from those of other retrospective studies.11–17
With regards to validation of the study, we climbed the first steps of the “reproducibility hierarchy” by using internal validation and historical transportability criteria,37
trying to avoid computer generated systems such as the jackknife cross validation method or bootstrapping technique so as to reproduce independent validation as closely as possible. However, we are also aware that only external validation will assess the usefulness of our model, and evaluate whether it can gain widespread clinical use.40
Lastly, as expected in our country, our series was skewed towards viral origin of liver disease,41
and therefore our results need to be validated in different series with a higher proportion of patients with alcohol related or cholestatic liver cirrhosis.
To conclude, the main finding of our study was that a composite parameter with biological plausibility (that is, the platelet count/spleen diameter ratio) showed the highest accuracy for non-invasively predicting the presence of OV in patients with either compensated or decompensated liver cirrhosis. The platelet count/spleen diameter ratio seems to represent an acceptable surrogate for clinically relevant portal hypertension. Applying the “platelet count/spleen diameter ratio strategy” for the detection of OV would seem to be more cost effective than the “scope all strategy”. On the basis of these data we hope that future studies will encourage other researchers to evaluate the reproducibility of the platelet count/spleen diameter ratio for the non-invasive diagnosis of OV in independent cohorts of patients with different clinical backgrounds.