We recruited 109 DC patients (95Males:14Females; mean age-51 ± 10 years) randomly assigned to either G-CSF + SMT or SMT (standard medical therapy) in the ongoing studies (25, 26) at a tertiary care center between January 2016 to March 2019. Patient selection criteria and study protocols (NCT02451033, NCT03415698) have been published previously (Supplementary Figures S1, S2).

Briefly, we included DC patients aged 18–80 years with a history of ascites and/or variceal bleeding and/or encephalopathy and/or jaundice. The cirrhosis was diagnosed through the clinical presentation, biochemical tests, radiology, endoscopy, and/or histopathology. We excluded patients with acute-on-chronic liver failure, sepsis, recent variceal bleed, grade III-IV hepatic encephalopathy, acute kidney injury, hepatocellular carcinoma or active malignancy, spleen>18 cm, portal vein thrombosis, cardiac dysfunction, recent alcohol abuse in 3 months, or alcoholic hepatitis, seropositivity for H.I.V., pregnancy, and hypersensitivity to G-CSF. We obtained written informed consent from all participants, followed GCP guidelines, and the protocol was approved by the Institute Ethics Committee (N.K./2,722/Ph.D./1,224).

The eligible patients were randomized using a computer-generated random number table with allocation concealment in the two groups (Supplementary Figure S2). The patients (n = 68) of group-A received standard medical therapy (SMT) plus G-CSF treatment at a dose rate of 5.0 μg/kg subcutaneously every 12.0 h for five consecutive days. This group received four similar cycles of G-CSF treatments at 3.0 months intervals over 1.0 years. The patients (n = 41) of group B received only standard medical therapy (SMT), which included: nutritional support, rifaximin, lactulose, bowel wash, albumin, diuretics, multivitamins, and antibiotics as per guidelines. Fresh frozen plasma and packed red-cell transfusions were administered as and when indicated. Stem cell mobilization was tested with CD34 cell estimation in peripheral blood at baseline and 6 days after therapy in both groups. The patients were followed up till 12 months, death or transplant (whichever was earlier) with physical examination, biochemical parameters, and CTP/MELD scoring. The patients were classified on CTP and MELD-based prognostic categories as CTP-A (n = 10), CTP-B (n = 56), and CTP-C (n = 43) and MELD of ≤15.0 (n = 77) and > 15.0 (n = 32). Study subjects underwent ^99mTc-SC liver and spleen SPECT and ICG tests at baseline and follow-up. The Q-SPECT parameters were evaluated and correlated with the CTP/MELD score and ICG-R15 in both groups of patients.

A series of phantom (Jaszczak SPECT phantom, Biodex Medical Systems, Inc., NY, United States) studies were performed using varying volumes (6.0 mL to 3,200 mL) and different radioactivity concentrations (28.5 kBq/mL to 151.33 kBq/mL) of pertechnetate (^99mTcO₄⁻) to develop and standardize the optimal processing protocol for accurate quantification (Supplementary Figures S3–S6; Supplementary Table S1). Data were acquired on a dual-headed gamma camera (Symbia- T16, SIEMENS, Erlangen, Germany) in 120 projections (20s/projection) in a 128 × 128 matrix with a zoom factor of 1.0. The data acquired was reconstructed by iterative reconstruction using a butter worth filter (cut off 0.5 cycles/cm) on a dual-head gamma camera. The threshold values varied from 19.0 to 52.0 to obtain the best volume estimation. Counts/voxel were converted into concentration units (μCi/mL) using the regression line, which was used to calculate liver/spleen volume, quantitative liver /spleen uptake, the percentage of injected dose per milliliter of liver/spleen tissue (%ID/mL) respectively.

Briefly, 4.0–5.0 mCi (148- to 185-MBq) radioactivity of ^99mTc-SC was injected intravenously in each patient (Supplementary Figure S7). Planar liver imaging was done at 20-min post-injection, followed by SPECT/CT acquisition at 30-min using the same gamma camera described above for the phantom standardization. The camera peaked at 140.0 keV with a window of ±20.0%. SPECT data were acquired over 360-degree rotation in 120 projections (20 s/ projection) in a 128 × 128 matrix (zoom, 1.5). After SPECT acquisition, low-dose CT (60 mAs and 140 kVP) was performed for the liver region. The images were reconstructed using an iterative reconstruction algorithm and Butterworth smoothing filter. The reconstructed images were analyzed using 3D volumetric analysis software 8.5.10.1 (Symbia.net, Munich, Germany). R.O.I.s were drawn coronal slices using a fixed threshold iso-contouring method to obtain volume and radioactive concentration.

ICG (0.3 mg/kg of body weight) was administered intravenously via a peripheral vein. Blood samples were collected at 0, 5-, 10-, 15-, and 20-min post-injection from the contralateral vein. A standard dilution curve was obtained by using varying ICG concentrations. The same was used to estimate the unknown blood ICG levels to measure the dye retention in the liver. The ICG concentrations in standard solution and blood were recorded at 805 nm using a spectrophotometer (Systronics, UV/VIS spectrophotometer118, Gujarat, India). Retention value and plasma disappearance rate were calculated for all the patients. The standard reference ICG retention value was taken as less than 10.0% at 15-min and plasma disappearance rate of more than 18.0 mL/min.

Fluorescence-activated cell sorting (FACS) technique was used to estimate the circulating hematopoietic progenitor cells (CD34+) in peripheral venous blood on day 0 and at day 6 of initiation of G-CSF therapy using a flow cytometer (BD FACS Canto II, San Jose, California, U.S.A.) (Supplementary Figures S8, S9). Precisely, 2.0 mL of blood was collected in EDTA vacutainer for total nucleated cell count and CD34+ cells estimation.

Numerical values were expressed as mean ± SD or median (range) as appropriate. The unpaired t-test or chi-square was used to compare numerical and categorical variables between groups. A paired t-test or u-test was used to compare the pre-and post-treatment numerical parameters. ANOVA was used to compare numerical data between the three groups. The correlation between numerical variables was expressed through the Pearson correlation coefficient (r) at 95% confidence intervals. Receiver operating characteristics (R.O.C.) curves were used to calculate the area under the curve (AUC), sensitivity, specificity, positive and negative predictive values, and accuracy for differentiating severity grades in liver cirrhosis. The optimal cut-offs of ICG and Q-SPECT/CT for defining severity and mortality were obtained from maximum Youden’s index. Kaplan Meier survival and cox-regression were conducted to evaluate survival estimates and hazards of mortality between groups, and the groups were compared with the Log-Rank test. Transplanted and lost to follow-up patients were considered an event for a 12-month transplant-free survival analysis. All statistical tests were two-sided and performed using IBM SPSS v.22 and RStudio v.1.4.1103 at a significance level of p < 0.05.

The best threshold value of Q-SPECT was calculated from the phantom imaging data using 3D volumetric analysis (Supplementary Figures S5, S6; Supplementary Table S1). These values were found to be 41.0, 38.0, and 33.0 for activity volume ranges of 6.0–30.0 mL, 500.0–1500.0 mL, and 2,400–3,200 mL, respectively. The estimated threshold values provided significant (p = 0.001) correlation of 0.90, 0.98 & 0.97 for the three reference ranges. Further, by using the threshold of 38.0%, the SPECT reconstructed data was used to calculate the different range (28.5 kBq/mL to 151.33 kBq/mL) of radioactivity concentrations and compared with the actual concentrations used for this standardization procedure. For calculating the tracer radioactivity, the regression line equation on attenuation corrected images used was: counts/cc = 6807.3 × μci/cc + 39.0. We used the best-fitted regression line to calculate the volume and radioactivity concentration at baseline and post-G-CSF treatment SPECT data in all the patients (Supplementary data).

The Q-SPECT and ICG parameters differed in CTP and MELD-based prognostic categories, as illustrated in Table 1. The liver volumes (LV) observed in CTP-C patients were significantly lower than that in CTP-B and A patients (p = 0.028) (Supplementary Figure S8A). The fractional liver uptake (%QLU) was substantially lower in CTP-C patients than in CTP-B and A (p = 0.001) (Supplementary Figure S8B). Likewise, liver volumes and %QLU were lower in the MELD>15 groups than MELD≤15 group, with p = 0.003 and p < 0.001 (Supplementary Figures S9A,B). Quantitative spleen parameters were not different between various Child and MELD classes (p > 0.05) (Table 1). ICG-R15 was significantly lower in CTP-A and B than in C and MELD≤15 than >15 groups (p < 0.001) (Supplementary Figure S10).

Table 2: The %QLU exhibited a strong inverse correlation with CTP (r = −0.728, p < 0.001), MELD (r = −0.743; p < 0.001) and reference standard ICG-R15 (r = −0.720, p < 0.001) values, respectively. The percentage of injected dose/mL also showed a significant but moderate correlation with CTP, MELD, and ICG-R15, followed by LV. However, no significant correlation was observed for CTP and MELD scoring with any quantitative spleen parameters. Analysis between CTP, MELD, and ICG parameters (ICG R-15 and P.D.R.) also demonstrated a negative correlation between CTP and ICG R-15 (r = −0.670; p = <0.001); CTP and P.D.R. (r = −0.670; p = <0.001); MELD and ICG R-15 (r = −0.521; p = <0.001), MELD and P.D.R. (r = −0.550; p = <0.001), respectively.

On R.O.C. analysis (Figure 1), compared to ICG and Liver volume assessment, %QLU showed the highest area under the R.O.C. curve (AUC) for discriminating CTP-A from B (0.920), CTP-B from C (0.891), and MELD≤15 from MELD>15 groups (0.906). The %QLU above 37.2% could differentiate CTP-A from B cirrhosis with sensitivity and specificity of 88.9 and 90.7%. A value above 22.2% could discriminate CTP-B from C cirrhosis with sensitivity and specificity of 88.9 and 82.9%. A value above 18.7% could classify low MELD from high MELD patients with sensitivity and specificity of 90.3 and 81.2%. Patients with %QLU > 25 at baseline were associated with significantly better 12-month survival than those with %QLU ≤25 (83% vs. 49%, p < 0.001) (Figure 2A).

Following the first cycle of G-CSF treatment, there was a significant (p < 0.001) increase in %CD34+ cell count (0.8 to 16.4, p < 0.001) following G-CSF treatment on day 6 as compared to the “day 0” counts (Supplementary Table S3; Supplementary Figure S11). However, no significant rise was seen in the %CD 34+ cell count (0.85 to 0.90, p > 0.05) in the SMT group (Supplementary Table S3; Supplementary Figure S12). The 12-month survival was significantly better in the G-CSF group (75%) than in the SMT group (51%), p = 0.008 (Figure 2B). Better survival in G-CSF-treated patients was associated with a significant increase in % QLU at 12 months compared with the baseline (26.1 ± 10.6 to 31.2 ± 12.2, p = 0.001) (Table 3). While in group B, the %QLU at 12 months did not change from the baseline value. The observed liver SPECT findings corroborated a significant improvement in CTP and MELD scores only in the G-CSF group. No significant change was noted in any other parameter, including the ICG-R-15 value (p = 0.07) at 12 months versus the baseline value (Table 3).

On adjusting for disease severity (MELD), baseline %QLU [H.R.: 0.234, 76.6% lower] and G-CSF treatment [H.R.: 0.414, 58.6% lower] were independently associated with a lower hazard of mortality in DC (Figure 2C). A representative ^99mTc-SC SPECT scan demonstrating an increase in %QLU and LV in a DC patient following G-CSF treatment at 12 months is shown in Figure 3A, while no change in the parameters was noted in a patient who received SMT (Figure 3B).