Cirrhosis is the final stage of any chronic liver disease. Gastrointestinal bleeding by rupture of esophageal varices (EV) is one of the most feared complications of cirrhosis because of its high mortality. All patients with cirrhosis should undergo endoscopic screening at the time of diagnosis to identify those at high risk of bleeding who may benefit from primary prophylaxis  . Upper gastrointestinal fibroscopy (UGF) is a relatively expensive, invasive and restrictive practice  . In Africa, and in Togo in particular, the follow-up of cirrhotic patients by UGF remains a challenge for clinicians because of the absence or insufficiency of digestive endoscopy units  . The appearance of ascite is the most common complication during cirrhosis  . The serum-ascites albumin gradient (SAAG) is a minimally invasive method that is very precise, already described as allowing to classifying the ascites fluid according to the absence or the presence of a portal hypertension (PHT)    . According to Starling, SAAG is the best indicator of PHT   and several studies have shown that there is a direct relationship between SAAG and the different measures of PHT   . Studies on non-invasive methods of EV including platelet ratio and spleen diameter have been conducted in african cirrhotic subjects   . In Ivory Coast, a study evaluating SAAG and the endoscopic parameters in 2015 showed a link between a high threshold of SAAG and the presenceof EV  . The use of SAAG can be considered as an indirect parameter for the detection of EV and be useful in regions where there is a shortage of human and material resources for the realization of the UGF. This study was conducted to find a more accessible and less restrictive mean such as SAAG to predict EV in cirrhotics with ascit.
2. Patients and Methods
This was a descriptive and analytical study of data from the hospitalized patient’s records in the hepatogastroenterology unit of the Campus University Hospital Center conducted from January 1, 2008 to March 31, 2018 (10 years and 3 months). Included were ascitic cirrhotic patients over the age of 15 who had performed cytochemical analysis of ascites fluid and UGF and had a protidogram. Patients hospitalized for degenerate cirrhosis were not included. SAAG was calculated by the following formula: SAAG (g/dl) = [Albuminemia (g/l) − Ascites fluid protein ratio (g/l)]/10. Normal albuminemia is between 30 and 40 g/l. The low SAAG was SAAG <1.10 g/dl (exudate); high SAAG was SAAG ≥1.10 g/dl (transudate). A protein content in the ascites fluid <30 g/l corresponded to a transudate and ≥30 g/l, corresponded to an exudate. The UGF has allowed according to the French classification based on that of the Japanese research company on portal hypertension  to diagnose the presence of EV, PHT gastropathy orcardio-tuberosity varices, and specify the grade of EV (I, II, and III) and objectify the presence of red signs. The statistical analysis was performed with the R Studio software version 3.4.2. The statistical tests used were the Pearson Chi-square test or the Fisher’s exact test for qualitative variables and the Student’s test for quantitative variables. A univariate and multivariate logistic regression was performed to investigate the factors associated with EV. The dependent variable was the existence of a PHT sign coded as 1 if the patient presented endoscopic signs of PHT (EV, cardio-tuberosity varices or PHT gastropathy) and 0 if not. The SAAG variable was the main exposure variable. In the absence of significance, we adjusted the final model to this variable. The variables statistically associated with the endoscopic signs of PHT during the univariate analysis with a degree of significance p < 0.25 were introduced in the initial model. Multivariate analysis made it possible to estimate the Adjusted Coast Ratio (aCR) and its 95% confidence interval (CI) for each selected variable. The adequacy of the model has been verified, based on the R2 value. Patients who did not die during the data collection phase, had given their oral consent for the completion of this study.
Of 1416 hospitalized cirrhotic patients, 125 medical records were included.
3.1. Characteristics of Cirrhotic Patients
The average age was 48.70 ± 14.89 years (extremes 15 to 85 years). The most represented age group was 36 to 54 years old. There was a male predominance (70.40%, male/female sex ratio of 2.38). Ascites fluid was transudative in 111 patients (88.80%). Albuminemia <35 g/l was noted in 92.80% of cases The median SAAG was 1.40 g/dl. Thrombocytopenia was noted in 75 patients (60%). Esophageal varices were observed in 81 patients (64.6%). Child Pugh B and C scores were noted in 62% and 29% of cases. The main causes of cirrhosis were hepatitis B and C viruses (63.60%) and alcohol (14.40%).
3.2. SAAG and Presence of EV
Esophageal varices were more frequent in case of high SAAG ≥1.10 g/dl, p = 0.0208 (Table 1).
3.3. SAAG and Other Endoscopic Parameters of Portal Hypertension
There was no association between SAAG and EV grade, red signs,
Table 1. Serum-ascites albumin gradient (SAAG) and presence of esophageal varices.
++ = Independence Chi-2 test.
cardio-tuberosity varices (VCT) and PHT gastropathy (Table 2).
3.4. SAAG and Ascitic Fluid Chemistry
There was an association between high SAAG (≥1.10 g/dl) and transudative ascites fluid, and between low SAAG (<1.10 g/dl) and exudative ascites fluid (p < 0.0001).
3.5. Predictive Factors of Esophageal Varices (Univariate Analysis)
The predictive factors for EV were mainly high vs low SAAG (OR = 2.45, 95% CI [1.14 - 5.32], p = 0.0223), and a SAAG threshold of 1.50 g/dl (OR = 2.42, 95% CI [1.14 - 5.27], p = 0.0232) as shown in Table 3.
3.6. Predictive Factors of Esophageal Varices (Multivariate Analysis)
After adjustment on sex, protein content in ascites fluid and platelet count, patients with SAAG ≥1.01 g/dl had 1.63 (95% CI [0.63 - 4.15]) times the risk of having EV (p = 0.3020). Being male (OR = 2.71 95% CI [1.09 - 6.88], p = 0.0321), having a protein level in ascites fluid <30 g/l (OR = 12.77 95% CI [2.71 - 94.24], p = 0.0033) and thrombocytopenia (OR = 4.30 95% CI [1.83 - 10.47], p = 0.0010), predisposed to having EV (Table 4).
3.7. Diagnostic Performance of SAAG for the Prediction of Esophageal Varices
At the threshold of 1.1 g/dl, the sensitivity (Se) was 72.83%, the specificity (Spe) was 47.72%; the positive predictive value (PPV) was 71.6%. When SAAG ≤0.02 g/dl, Se was 100% and Spe ≤18.16%. When SAAG ≥3.16 g/dl, Spe was 100% and Se ≤1.23%. 0.91 g/dl corresponded to the Youden index (better sensitivity, better specificity), with Spe of 40.90% and Se of 83.95%. The area under the curve was
Table 2. Serum-ascites albumin gradient (SAAG) and other endoscopic parameters of portal hypertension.
aPHT: Portal hypertension; bEV: Esophageal varices; + = Fisher’s exact test; ++ = Independence Chi-2 test.
Table 3. Predictive factors of esophageal varices (univariate analysis).
95% CI: 95% Confidence Interval; OR: Odds Ratio; SAAG: Serum-Ascites Albumin Gradient.
Table 4. Predictive factors of esophageal varices (multivariate analysis).
95% CI: 95% Confidence Interval; OR: Odds Ratio; SAAG: Serum-Ascites Albumin Gradient.
0.61 (Figure 1).
4.1. Limitation of Our Study
The main limitation of our study is that the collection of data was retrospective, a source of bias related to this type of study, including missing informations in medical records.Some patients did not performed cytochemical analysis of ascites fluid, protidogramand UGF. The main reasons mentioned were the lack of financial means and or lack of access to UGF. This reduced the number of included cirrhotic patients in our study. The other limitation was the fact that
Figure 1. ROC curve corresponding to the evaluation of serum-ascites albumin concentration gradient (SAAG) in the diagnosis of esophageal varices.
cytochemical analysis of ascites fluid and protidogram were not always performed in the same laboratory and that also UGF was not always performed by the same operator. However, our study, the first one in Togo, a country with limited resources, keeps its interest because it enable us to investigate the existence of an association between SAAG and EV in cirrhotic patients with ascites.
4.2. Serum-Ascites Albumin Concentration Gradient
In our series, the median SAAG was 1.40 g/dl; the one found by Okon et al. in Ivory Coast  was −0.30 g/dl. The majority of patients in our study had a SAAG high ≥1.10 g/dl (65.6%) as in the studies of Gurubacharya et al.  in Nepal (78.13%), and of Suresh et al.  in India (79%). Shahed et al.  in Bangladesh had only included patients with high SAAG in their study. On the other hand, in the Okon et al. study in Ivory Coast  , the majority of patients (68.2%) had low SAAG. For these Ivorian authors, the condition of malnutrition of their patients causing hypoalbuminemia, and the infection of ascites fluid could explain the decrease in the value of SAAG.
4.3. SAAG and Presence of Esophageal Varices
Esophageal varices were significantly rated for SAAG high ≥1.10 g/dl. From a SAAG ≥1.50 g/dl, 3 out of 4 patients had significant EV. For Okon et al.  , EV were significantly more prevalent in SAAG high ≥1.10 g/dl (71.95%) and were almost constant from SAAG ≥1.50 g/dl. In the work of Torres et al.  , Demirel et al.  , Suresh et al.  , Gurubacharya et al.  , and Shahed et al.  , EV were consistently present in patients at the threshold of 2 g/dl. Apart from Shahed et al.  who did not include patients with low SAAG <1.10 g/dl in their study, none of the patients had EV at a low SAAG in the Torres et al.  , Demirel et al.  , and Gurubacharya et al.  studies contrary to the study of Okon et al.  , and ours where the probability of presence of EV is not zero when the SAAG is low. Low SAAG < 1.10 g/dl was not a parameter to rule out the existence of EV in our study; on the other hand, high SAAG ≥ 1.10 g/dl significantly predicted the presence of EV.
4.4. SAAG and Esophageal Varices Grade
There is controversy over the correlation between SAAG and the size of varicose veins. In our study, no correlation was noted between SAAG and EV grade; it is the same in the study of Okon et al. in Ivory Coast  . Suresh et al.  , Torres et al.  , Gurubacharya et al.  and Demirel et al.  found a correlation between SAAG and the grade of EV. Shahed et al.  noted in their study that the increase in the value of SAAG correlated with the increase in the frequency of patients with grade III esophageal varices. On the other hand, this observation was not made in the case of grade I and II esophageal varices.
4.5. SAAG and Other Signs of Portal Hypertension
It was not noted in our study, an association between SAAG and the PHT gastropathy, the red signs on the EV and the cardio-tuberosity varices. Unlike our study, Okon et al.  observed that at a threshold of 1.50 g/dl of SAAG, there were significantly 50% red signs, more cardio-tuberosity varices and gastropathy PHT. However, ivorian authors have not specifically identified a correlation between the level of SAAG and other PHT parameters  . Shahed et al.  noted that as the value of SAAG increased, so did the proportion of patients with severe PHT gastropathy; this finding was not made in the case of moderate PHT gastropathy.
4.6. Predictive Factors of Esophageal Varices
In multivariate analysis, being male, having a protein content of less than 30 g/l in ascites fluid and thrombocytopenia, were the main predictive factors of EV. Patients with high SAAG ≥ 1.10 g/dl were more likely to have EV. For Torres et al.  , SAAG predicted EV from a threshold ≥ 1.435 ± 0.05 g/dl.
4.7. Diagnostic Performance of SAAG for the Prediction of Esophageal Varices
The area under the ROC curve was 0.61 indicating a fairly good diagnostic performance of SAAG. Okon et al.  found 0.80; they had noted at the threshold of 1.10 g/dl a Se of 53.3% and a Spe of 80%. Suresh et al in India  found at the same threshold an Se of 81%, a Spe and a PPV of 100%. From 1.50 g/dl of SAAG, Spe was 65.90% greater but 30.40% were false negatives (1/3 of patients who developed EV were not diagnosed by the SAAG method with this threshold). At a threshold value of 0.91 g/dl of SAAG (Youden index) in the patients of this study; 10.40% of false negatives, or one tenth of patients with EV, were not diagnosed. Okon et al.  found a Youden index of −1.640 g/dl with a Se of 96%, a Spe of 55%, a PPV of 70% and a false negative rate of 3%. These different findings reflect the difficulty of accurately finding a threshold value of SAAG, which can be used to diagnose patients with EV.
Since SAAG measurement is simple, less costly, non-invasive and easily achievable, it can be used as a tool for indirect diagnosis of EV and to allow the setting up of an early prophylactic treatment in the regions of Togo lacking digestive endoscopy centers; this can help prevent cataclysmic gastrointestinal bleeding. However, the upper gastrointestinal fibroscopy remains the most effective mean of diagnosing EV.