Platelet count can predict the grade of esophageal varices in cirrhotic patients: a cross-sectional study

Background: Bleeding from esophageal varices is a life-threatening complication in cirrhosis. Screening endoscopy is recommended in cirrhotic patients to identify patients at risk of variceal hemorrhage, but this is an invasive procedure and has limitations. Therefore, thrombocytopenia has been proposed to predict the existence and grade of esophageal varices. The aim of the current study was to determine a correlation between platelet count and grades of esophageal varices in patients with liver cirrhosis. Methods: This cross-sectional study was conducted at the POF Hospital, Wah Cantt from 1 st October, 2017 to 30 th May, 2018. Newly diagnosed cases of cirrhosis having varices of any grade on endoscopy were included. Endoscopic findings of patients were standardized using Paquet grading system. On the basis of platelet count, patients were divided into four subgroups. Platelet count groups were correlated with grading of esophageal varices using Spearman rank correlations. Chi Square test was used to see association between the platelet count and grade of esophageal varices. Results: 110 patients were included in the study, 55.5% (n=61) were male. Mean age of the patients was 59.89±9.01 years. Platelet count was <50,000/uL in 35.5% patients, 50,000-99,000/uL in 26.4%, 100,000-150000 in 12.7%, and >150,000/uL in 25.5% patients. Grade I esophageal varices were found in 23.6% of patients, whereas grade II, III and IV were found in 24.5%, 33.6% and 18.2% of patients, respectively. Mean platelet count was 213884.62/mm 3 in patients with grade I varices, whereas it was 119518.52/mm 3, 58386.49/mm 3 and 21600.00/mm 3 in patients with grade II, III and IV varices, respectively (p=<0.0001). A significant negative correlation between platelet count and grades of esophageal varices was found (p<0.001). Conclusion: Platelet count can predict the grade of esophageal varices in cirrhotic patients. There is significant negative correlation between platelet count and grades of esophageal varices.


Amendments from Version 1
We had made few changes as suggested by reviewers. We applied one way ANOVA (Analysis of Variance) test and added in results. Reviewer raised an important point regarding the imminent bleeding signs, which we missed during data collection. As it is not possible to include it in article (we do not have data), we had mentioned it in limitations of our study in revised version so researchers in future can include this point.

Introduction
Portal hypertension is a common complication of liver cirrhosis which can lead to esophageal varices that may rupture and bleed 1 . Bleeding from esophageal varices is life threatening and comprises 10% of all cases of upper gastrointestinal (GI) bleeding 2 . Esophageal varices are present in approximately one third of patients at diagnosis of cirrhosis and incidence increases to 90% in 10 years 3 . The rate of progression from small to large varices is estimated to be 8-10% per year and the annual rate of esophageal hemorrhage is 5% for small varices and 15% for large varices 3,4 . Screening endoscopy is therefore recommended for early detection of esophageal varices and grading of varices in cirrhotic patients to identify patients with risk of variceal hemorrhage and administer prophylactic treatment if required 5 . Screening endoscopy is however an invasive procedure and has limitations 6 . Diagnosis and grading varices by endoscopy is operator dependent. This approach also places a heavy burden upon endoscopy units and repeated testing over time may affect patient compliance, and endoscopic screening also increases the associated health care costs 7 . Due to these problems regarding endoscopy, some noninvasive means have been proposed for prediction of esophageal varices and their grades in order to restrict endoscopy to the population with high risk of variceal bleeding 7 . Various clinical and biochemical predictors have been studied to predict the existence and grade of esophageal varices in cirrhotic patients 8 .
Thrombocytopenia (platelet count <150,000 /uL) is one such predicator. It is found in approximately 64-76% of patients with portal hypertension and cirrhosis 9 . The pathogenesis of thrombocytopenia in cirrhosis is multi-factorial and includes decreased thrombopoietin production, sequestration of platelets in spleen and direct myelosuppression due to hepatitis C virus 10 . As thrombocytopenia and esophageal varices are common findings in cirrhosis, and portal hypertension is associated with both of these findings, some studies have been conducted to see the relation of thrombocytopenia as a non-invasive marker with grades of esophageal varices 10,11 . It was found in different studies that platelet count was inversely correlated with grades of esophageal varices 11 . Large esophageal varices also appeared to have a lower platelet count in a study conducted at Shanghai East Hospital 4 In resource poor developing countries, endoscopy of every cirrhotic patient to grade the esophageal varices to select the patients for prophylactic therapy is not possible due to the limited number of endoscopes and endoscopists in government hospitals. We need to identify non-invasive and reliable markers to predict the grades of esophageal varices in our population in Pakistan, so that endoscopists can select the patients at increased risk of bleeding for endoscopy on priority basis. Therefore, we conducted this study to determine correlation between platelet count and grades of esophageal varices in our population and thus assess the possibility of using platelet count to predict the grades of esophageal varices.

Study design and participants
This cross-sectional descriptive study was conducted at the Department of Medicine, POF Hospital, Wah Cantt, Pakistan from 1 st October, 2017 to 30 th May, 2018. Newly diagnosed cases of cirrhosis having varices of any grade on endoscopy, irrespective of the cause of cirrhosis, with or without ascites and splenomegaly, from 18 to 70 years were included in the study. Diagnosis of cirrhosis was made using data obtained from clinical findings, laboratory investigations and ultrasonographic findings of the liver. Patients with hematological disorders, portal vein thrombosis, on B-blocker prophylactic therapy and those who had undergone endoscopic band ligation or sclerotherapy were excluded.
The sample size calculated was 110 using sample size calculator n = (Z 2 × P × (1 -P))/d 2 for cross sectional studies, taking a confidence level 95%. In total, 110 patients were included by consecutive sampling.
Informed written consent was taken from patient or relative wherever relevant (consent was taken from relatives in patients with hepatic encephalopathy). Ethical approval was obtained from POF Hospital Research Ethics Committee before the start of the study (vide letter no. POFH/ERC/99053/03).

Data collection
Endoscopy was performed by the same endoscopist using the Olympus GIF type Q260 endoscope. All patients were kept NPO for 8 hours before endoscopy and the endoscopic findings of the patients were standardized using the Paquet grading system for esophageal varices dividing esophageal varices by grades I to IV 12 . Blood samples of all patients were taken by the staff nurse before endoscopy and sent to the laboratory of POF Hospital on the same day to calculate platelet count. Platelet count was calculated by automatic hematology analyzer (Sysmex XN-1000) and confirmed by manual method (small volume of whole blood was treated with RBC lysing reagent ammonium oxalate, was put in hemocytometer and platelet counting was done by hematologist using phase contrast light microscopy). All details including demographic details, endoscopic findings and platelet count were obtained from medical records of the selected patients. On the basis of platelet count, patients were divided into four subgroups: group 1,platelet count <50,000/uL; group 2,platelet count 50,000-99000 /uL; group 3,platelet count 100,000-150,000/uL; and group 4,normal platelet count >150,000/uL.

Data analysis
All data was entered and analyzed using SPSS version 19.0. Mean and standard deviation was calculated for quantitative data, such as age. Frequency and percentage was calculated for gender, platelet count and grade of esophageal varices. Platelet count group was correlated with grading of esophageal varices using Spearman rank correlations. Chi square test was used to calculate the association between platelet count and grade of esophageal varices. P-value ≤0.05 was considered significant.
Spearman's correlation showed a significant negative correlation between platelet count and grades of esophageal varices (Table 2).
Mean platelet count was also significantly lower in patients with grade IV varices as compared to patients with grade I varices (p=0.0001; Table 3).
Analysis of variance also showed significant relation between platelet count and grades of esophageal varices. (p <0.0001; Table 4; Figure 1). Lower platelet count is significantly associated higher grades of esophageal varices.

Discussion
Our results showed that there is a significant negative correlation between the platelet count and grades of esophageal varices. Patients with a lower platelet count had varices of higher grades. In group 1 patients (platelet count <50,000//uL; n=39), 18 and 16 patients were found to have grade IV and grade III varices, respectively, as compared to only 2 and 3 patients having grade I and II varices, respectively. In contrast, in group 4 patients (platelet count >150,000/uL; n=28), 22 patients had grade I varices and no patients had grade III or IV varices. This suggests that platelet count can be used as a predictor of grades of esophageal varices without frequent upper GI endoscopy.
In our study, 55.5% (n=61) patients were male and 44.5% (n=49) were female. Studies have shown that the incidence of cirrhosis is lower in females as compared to males due to slower progression of fibrosis 13 . In a study conducted in Taiwan, 71% cirrhotic patients were male 14 . These findings are consistent with our findings of male predominance. Another study also supports our findings, as it shows that men had a higher incidence of cirrhosis in all age groups as compared to women of the same age group 15 . A local study also showed that 64% of cirrhosis patients were men 16 .
Mean age in our study was 59.89±9 years. Although cirrhosis of the liver can develop at any age, it is usually common in old age with a higher mean age of incidence 15 . Mean age was found to be 40.5 years in cirrhosis patients in a study conducted by Devrajani et al. 16 . This is comparable with our results. A study conducted in China showed a mean age of 50.29±7.03 years in cirrhosis patients 17 , which is similar to a mean age of 54.39±7.46 years in a study conducted by Abd-Elsalam et al. 10 , almost the same mean age as found in our study.
A significant inverse correlation was found between platelet count and grades of esophageal varices in our study. Lower  Many studies earlier used platelet count along with other non invasive parameters like platelet count/ spleen diameter ratio, AST Platelet Ratio Index (APRI) to predict the grades of esophageal varices and need of endoscopy in cirrhotic patients 20 . It was found that these markers were useful in predicting the grades of esophageal varices 20,21 . We only used platelet count, platelet count is not only noninvasive, it is also inexpensive, resource effective, does not need special expertise, and is easily available.
Our study has some limitations: it was a single center study including all patients of cirrhosis irrespective of the cause; we only used platelet count to predict the grade of varices, some studies have shown that platelet count and spleen size ratio is more accurate in predicting the size and grade of varices. We did not look for impending or recent bleeding signs like cherry red spots along with grades of varices. In future, multicentre studies with specific cause of cirrhosis, using imminent bleeding signs also along with grades of varices and using platelet count and spleen size ratio are suggested.

Conclusion
There is significant negative correlation between platelet count and grades of esophageal varices. Low platelet count is associated with higher grades of esophageal varices. This suggest that platelet count can be used as a predictor to predict the grade of esophageal varices.

Alexandru Constantin Moldoveanu
Gastroenterology and Internal Medicine II Clinic -University Emergency Hospital Bucharest, Carol Davila University of Medicine and Pharmacy, Bucharest, Romania Esophageal varices are a frequent complication of cirrhosis and can be a life-threatening complication if ruptured. The risk of variceal bleeding is, however, not equal between patients, and the current gold standard method for assessing bleeding risk is upper gastrointestinal endoscopy -an invasive procedure, that is not easily tolerated by the patient. There are multiple non-invasive methods that were proposed over the years as either an alternative to endoscopy or as a method of reducing the number of endoscopies by performing them selectively only in patients with a high probability of also requiring endoscopic treatment.
The authors' study presents a very simple parameter, thrombocyte count, as a non-invasive and very cost-effective alternative method of identifying patients with different grades of esophageal varices. Considering that both hypersplenism (and thrombocytopenia) and esophageal varices are results of portal hypertension, it appears to be a reasonable presumption that the two would be correlated.
The results indeed show as a very good correlation between the size of varices and platelet count, in both group analysis as well as continuous variable analysis. What is most interesting to me is that the platelet count has a very good sensitivity (approximatively 93% based on the results) for predicting patients with large varices (grades III and IV) requiring prophylactic endoscopic treatment. No patient with normal thrombocytes has had large varices in this study, which suggests that these patients might not require diagnostic endoscopy.
Regarding the statistical analysis methodology, I believe that the methods could be expanded upon. An analysis of variance type test between groups would probably reveal statistically significant results given the values presented. Another possible analysis method is classifying patients into two groups: 1. needing endoscopic treatment and 2. not needing endoscopic treatment, and plotting an ROC curve, which could be used to determine sensitivity and specificity for an optimal cut-off.
One criticism that I could make is that imminent bleeding signs, such as red wale signs, were not assessed. While these signs appear more frequently in large varices, they have frequently been reported in grade 2 varices as well, making it more difficult to assess which patients do not require endoscopy using the grade of the varices alone.
Another criticism of the article is that, while the conclusions are very interesting, the information is not entirely new. The Baveno VI consensus has already proposed thrombocyte count as one of the most important factors to determine which patients can be ruled out from screening endoscopy. The authors' study still has practical value for medical centres where measurement of the other main parameter in the Baveno VI consensus, liver stiffness, is either unavailable or not sufficiently cost-effective.

If applicable, is the statistical analysis and its interpretation appropriate? Partly
Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Partly 1. We applied one way ANOVA (Analysis of Variance) test and added in results.
2. Reviewer raised an important point regarding the imminent bleeding signs, which we missed during data collection. As it is not possible to include it in article (we do not have data), we had mentioned it in limitations of our study in revised version so researchers in future can include this point.
Another concern of reviewer was "information is entirely not new". We agree with this statement but most of the studies used other parameters as well along with platelet count like spleen diameter, liver stiffness as mentioned by reviewer also. We used only platelet count to predict the varices size which is valuable for centres in developing world where other facilities are not available or cost is a problem, this was the main purpose of the study and also appreciated by reviewer.