Association between the systemic immune-inflammation index and overall survival in patients with leukemia: a multicenter cohort study Authors: Yiyang Chen, Muchen Ren, Sihui Liu, Tianyi Wang, Heyang Zhang, Junju He, Hanping Shi Published: 2026 ABSTRACT AbstractBackgroundThe systemic immune-inflammation index (SII; platelet × neutrophil/lymphocyte) is a well-established prognostic marker in solid tumors, but its clinical significance in leukemia remains unclear, where peripheral blood cytopenias may primarily reflect bone marrow failure rather than conventional cancer-related inflammation. This study aimed to evaluate the association between pretreatment SII and overall survival (OS) in patients with leukemia and to assess its prognostic relevance in a multicenter cohort.MethodsWe retrospectively analyzed 609 leukemia patients from a multicenter registry, split into training (n=427) and validation (n=182) cohorts. Baseline SII was calculated before therapy. A data-derived SII threshold was identified in the training cohort using maximally selected rank statistics. OS was assessed using Kaplan-Meier curves, Cox regression, and restricted cubic splines (RCS). A nomogram was constructed based on the multivariable Cox model and internally evaluated using bootstrap calibration for 1- and 3-year survival.ResultsAn SII cutoff of 173.56 significantly separated survival (log-rank P<0.001). After adjustment for the demographic, comorbidity, and treatment variables available in the registry, higher SII (≥173.56) remained associated with a lower hazard of death [hazard ratio (HR) =0.41; 95% confidence interval (CI): 0.31–0.55; P<0.001]. Spline analyses suggested a nonlinear inverse association between SII and mortality. The nomogram showed acceptable internal calibration in both cohorts.ConclusionsHigher baseline SII was consistently associated with better OS in this multicenter leukemia cohort. This association may reflect differences in underlying disease burden and preserved hematopoietic function, rather than an independent biological effect of SII. Because leukemia subtype, disease status, and cytogenetic and molecular risk were unavailable, whether SII adds prognostic value beyond established leukemia risk factors remains uncertain. These findings support further evaluation of SII in well-characterized, subtype-specific cohorts.Keywords: Systemic immune-inflammation index (SII), leukemia, overall survival (OS), multicenter cohort Highlight box Key findings • In this multicenter cohort of 609 patients with leukemia, higher baseline systemic immune-inflammation index (SII) was associated with better overall survival after adjustment for registry-available covariates. The association was nonlinear, with the greatest increase in mortality risk observed at very low SII levels, and was consistent across the training and internal validation cohorts. What is known and what is new? • SII is commonly interpreted as a systemic inflammatory marker in solid tumors. In leukemia, however, its component cell counts are also directly influenced by marrow involvement, disease-related cytopenias, and hematopoietic reserve. • This study suggests that SII captures clinically relevant variation in survival in a real-world leukemia cohort. The inverse association should therefore be interpreted as a composite hematologic signal that may reflect both disease severity and host physiological reserve, rather than as evidence of an independent protective inflammatory mechanism. What is the implication, and what should change now? • SII may serve as a simple, routinely available marker associated with patient vulnerability and survival risk in leukemia. Its value lies in complementing clinical assessment rather than replacing established disease-specific prognostic systems. • Future studies incorporating leukemia subtype, disease status, cytogenetic and molecular risk, and treatment intensity are needed to determine whether SII adds prognostic information beyond established measures of disease severity. Introduction Leukemia is a heterogeneous group of hematologic malignancies and remains a major contributor to cancer-related morbidity and mortality worldwide (1,2). Despite notable advances in chemotherapy, targeted agents, and hematopoietic stem cell transplantation, survival outcomes among patients vary considerably (3-6). This underscores the importance of accurate prognostic assessment. Conventional prognostic systems largely rely on clinical features and cytogenetic or molecular abnormalities; however, marked heterogeneity in survival often persists within the same risk category. Existing prognostic models in leukemia, including the European LeukemiaNet (ELN) risk classification for acute myeloid leukemia (AML) and other cytogenetic- and molecular-based risk stratification systems, have improved prognostic assessment. However, substantial heterogeneity in outcomes remains within the same risk categories, highlighting the need for additional readily available biomarkers that may complement existing prediction models. Growing evidence has highlighted the close interplay between systemic inflammation, immune response, and cancer progression. In solid tumors, several inflammation-based indices—such as the neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and systemic immune-inflammation index (SII)—have been shown to predict survival outcomes (7-9). However, the biological meaning of blood count-derived indices in leukemia may not be the same as that in solid tumors. In leukemia, neutrophil, lymphocyte, and platelet counts are often affected directly by leukemic bone marrow infiltration, suppression of normal hematopoiesis, and treatment-related myelosuppression, rather than by systemic inflammation alone (10). Cytopenias also reflect reduced hematopoietic reserve and are associated with a higher risk of infection and bleeding, both of which can independently influence clinical outcomes. Thus, a lower inflammation-based index in leukemia should not be interpreted simply as evidence of a favorable low-inflammatory status; it may instead indicate more profound marrow dysfunction and greater clinical vulnerability (11,12). These indices remain attractive because they are inexpensive and routinely available in clinical practice; in leukemia, however, they may integrate systemic inflammatory activity with disease-related cytopenias and hematopoietic reserve rather than represent inflammation alone. Compared with NLR and PLR, SII combines neutrophil, lymphocyte, and platelet counts into a single measure and may therefore capture broader variation in the peripheral blood profile. Nevertheless, the prognostic relevance of SII in hematologic malignancies, particularly leukemia, has rarely been investigated. To date, most available studies have focused on solid tumors, and whether SII retains the same biological and prognostic meaning in leukemia remains uncertain. Accordingly, using the multicenter INSCOC cohort, we sought to characterize the association between pretreatment SII and overall survival (OS) in patients with leukemia and to determine whether this routinely available composite index identifies clinically meaningful variation in survival after adjustment for registry-available covariates. We examined SII as both a continuous and a categorized measure, assessed the shape and consistency of the association in the training and internal validation cohorts, and developed an internally calibrated prognostic model for 1- and 3-year survival. Because the INSCOC database lacks leukemia subtype, disease status, cytogenetic, and molecular-risk information, our objective was not to establish SII as a disease-independent inflammatory biomarker or a substitute for established prognostic systems. Rather, this study was designed to define the prognostic information carried by SII in a real-world leukemia cohort and to provide a disease-contextualized basis for subsequent subtype-specific and externally validated evaluation. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2026-1196/rc). Methods Study design and population This study was designed as a retrospective analysis based on the INSCOC cohort, a nationwide registry that prospectively recruited cancer patients from more than 40 tertiary hospitals in China (Chinese Clinical Trial Registry, ChiCTR1800020329). For the present study, we identified 609 individuals with a confirmed diagnosis of leukemia who were enrolled between June 2012 and June 2021. Patients were considered eligible if they had (I) a clear pathological or clinical diagnosis of leukemia; (II) available baseline demographic, clinical, and laboratory information; and (III) complete follow-up data for survival outcomes. Those with major data missing, concurrent severe infections, or autoimmune disorders were excluded. The study sample was randomly split into a training set (n=427) and a validation set (n=182) using a 7:3 ratio (Figure S1). Because detailed leukemia subtype information was not available in the INSCOC registry for the present analysis, all eligible leukemia cases were analyzed as a unified cohort. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of the First Affiliated Hospital of Sun Yat-sen University {Medical Research Ethics Review [2013] No. 82; registration number: ChiCTR1800020329}. Informed consent was obtained from all participants prior to enrollment. Data collection Clinical and demographic variables collected at baseline included age, sex, smoking and alcohol history, comorbidities such as hypertension, diabetes, and coronary artery disease, family history of malignancy, and registry-available treatment indicators, including chemotherapy, radiotherapy, and surgery. Laboratory measurements obtained prior to anti-leukemia therapy comprised neutrophil, lymphocyte, and platelet counts, from which the SII was calculated as platelet × neutrophil/lymphocyte. Age was analyzed as a continuous variable and further dichotomized at 65 years in sensitivity analyses. Outcome and follow-up The primary study outcome was OS, defined as the time from diagnosis to death from any cause or last known follow-up. Survival information was obtained through electronic medical records, hospitalization data, and routine telephone interviews conducted by trained research staff. Patients alive at last contact were censored at that date. Statistical analysis Continuous data were summarized as mean with standard deviation or median with interquartile range (IQR), depending on distribution, and compared using Student’s t-test or Mann-Whitney U test as appropriate. Categorical variables were presented as counts and percentages, with group differences assessed by Chi-squared test or Fisher’s exact test. The optimal cutoff value for SII was determined in the training set using maximally selected rank statistics. Survival curves were plotted by the Kaplan-Meier method and compared with the log-rank test. Univariate and multivariate Cox proportional hazards regression models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). Multivariable models adjusted for age, sex, smoking, alcohol use, family history, comorbidities (diabetes, hypertension, and coronary artery disease), and treatment-related variables available in the registry. Restricted cubic spline (RCS) functions were applied to assess potential nonlinear associations between SII and mortality risk. A nomogram was constructed incorporating the SII group together with clinically relevant baseline characteristics (smoking, drinking, hypertension, and diabetes) that are routinely available in the INSCOC registry. These variables were selected to reflect standard baseline comorbidity and lifestyle profiles, rather than solely on the basis of statistical significance in the multivariable Cox model. The predictive performance was evaluated by calibration curves using 1,000 bootstrap resamples in both training and validation sets. All statistical analyses were performed using R software (version 4.2.1; R Foundation for Statistical Computing, Vienna, Austria). Kaplan-Meier estimation, log-rank testing, and Cox proportional hazards regression were performed using the survival package. Maximally selected rank statistics and survival-curve visualization were implemented using the maxstat and survminer packages, respectively. RCS modeling, nomogram construction, and bootstrap calibration were performed using the rms package. Additional graphical outputs were generated using the ggplot2 package. All statistical tests were two-sided, and P values <0.05 were considered statistically significant. Given the exploratory nature of this analysis and the lack of leukemia subtype data, the findings were interpreted primarily as association estimates rather than causal effects. Results Baseline characteristics Among the 609 patients, 427 were assigned to the training cohort and 182 to the validation cohort. Among the 609 included patients, 235 deaths occurred during follow-up. In the training cohort, 159 deaths occurred among 427 patients. In the validation cohort, 76 deaths occurred among 182 patients. The median follow-up time was 22.5 months (IQR, 11.9–37.9 months) in the overall cohort, 23.1 months (IQR, 12.2–37.7 months) in the training cohort, and 21.5 months (IQR, 10.9–38.9 months) in the validation cohort. As shown in Table 1, the two groups were generally comparable in age, sex distribution, comorbidities, and treatment history (all P>0.05). Chemotherapy was the predominant treatment (>85% in both cohorts), while radiotherapy and surgery were rare. In terms of blood indices, neutrophil and platelet counts were similar between groups. Lymphocyte counts were slightly higher in the validation cohort (median 1.23 vs. 1.05 ×109/L, P=0.03), and the median SII was modestly lower in the validation cohort than in the training cohort (104.92 vs. 166.15, P=0.046). Table 1Baseline characteristics of patients in the training and validation cohortsVariableTraining cohort (n=427)Validation cohort (n=182)P valueGender, male211 (49.4)102 (56.0)0.16Smoking, yes109 (25.5)59 (32.4)0.10Drinking, yes60 (14.1)26 (14.3)>0.99Chemotherapy, yes365 (85.5)158 (86.8)0.76Radiotherapy, yes4 (0.9)0 (0.0)0.45Diabetes, yes27 (6.3)16 (8.8)0.36Hypertension, yes44 (10.3)24 (13.2)0.37Coronary disease, yes9 (2.1)3 (1.6)0.96Family history, yes36 (8.4)16 (8.8)>0.99Surgery, yes3 (0.7)0 (0.0)0.62Age, years46.00 (33.00, 56.50)43.00 (30.00, 56.00)0.20Neutrophils, 109/L1.88 (0.56, 3.60)2.04 (0.66, 4.03)0.60Lymphocyte, 109/L1.05 (0.64, 1.69)1.23 (0.73, 1.93)0.03PLT, 109/L107.00 (39.00, 199.00)87.50 (26.25, 185.00)0.08SII166.15 (24.80, 446.58)104.92 (14.83, 359.01)0.046Data are presented as n (%) or median (interquartile range). PLT, platelet; SII, systemic immune-inflammation index. Baseline characteristics of patients in the training and validation cohorts Data are presented as n (%) or median (interquartile range). PLT, platelet; SII, systemic immune-inflammation index. Prognostic value of SII We next evaluated the prognostic role of SII in OS. Using maximally selected rank statistics, a cut-off value of 173.56 was identified (Figure S2). Patients were subsequently classified into low-SII (<173.56) and high-SII (≥173.56) subgroups. According to this cutoff, 324 patients were classified into the low-SII group and 285 into the high-SII group. Death occurred in 164 patients in the low-SII group and 71 patients in the high-SII group. Kaplan-Meier survival curves showed a significant separation between the two groups, with the high-SII group demonstrating better OS than the low-SII group (log-rank P<0.001; Figure 1). Survival curves of different cohorts based on SII index. SII, systemic immune-inflammation index. Cox regression analysis of prognostic factors The prognostic significance of SII and other clinical characteristics was further evaluated using Cox proportional hazards regression (Table 2). In the univariate analysis (Model a), high SII (≥173.56) was associated with a significantly lower hazard of death than low SII (HR =0.39, 95% CI: 0.29–0.51, P<0.001). After adjusting for age and sex (Model b), the association between high SII and a lower hazard of death remained statistically significant (HR =0.41, 95% CI: 0.31–0.54, P<0.001). In the fully adjusted model incorporating additional clinical covariates (Model c), high SII remained significantly associated with a lower hazard of death (HR =0.41, 95% CI: 0.31–0.55, P<0.001). No statistically significant associations with OS were observed for the other included baseline variables in the fully adjusted model (all P>0.05). Table 2Univariate and multivariate Cox regression analyses of overall survivalVariablesModel aModel bModel cHR (95% CI)P valueHR (95% CI)P valueHR (95% CI)P valueAge ≥65 years2.10 (1.49, 2.96)<0.0011.30 (0.82, 2.05)0.2561.34 (0.84, 2.14)0.21Gender (female)1.22 (0.95, 1.58)0.111.18 (0.91, 1.53)0.1891.27 (0.92, 1.76)0.14Family history (yes)0.87 (0.54, 1.39)0.570.78 (0.48, 1.25)0.3050.79 (0.49, 1.28)0.35Smoking (yes)1.08 (0.81, 1.44)0.571.16 (0.81, 1.66)0.3941.14 (0.78, 1.65)0.48Drinking (yes)1.04 (0.72, 1.50)0.801.06 (0.71, 1.57)0.7501.02 (0.68, 1.54)0.89Chemotherapy (yes)1.28 (0.86, 1.91)0.211.32 (0.89, 1.96)0.1651.29 (0.86, 1.94)0.21Radiotherapy (yes)1.16 (0.29, 4.70)0.830.94 (0.23, 3.81)0.9381.12 (0.26, 4.73)0.87Diabetes (yes)1.25 (0.78, 2.00)0.350.98 (0.61, 1.59)0.9601.00 (0.61, 1.65)0.99Hypertension (yes)1.32 (0.89, 1.93)0.160.93 (0.62, 1.39)0.7410.94 (0.61, 1.43)0.78Coronary disease (yes)1.72 (0.81, 3.67)0.151.10 (0.51, 2.38)0.7991.20 (0.54, 2.64)0.65SII <173.56Ref.Ref.Ref. ≥173.560.39 (0.29, 0.51)<0.0010.41 (0.31, 0.54)<0.0010.41 (0.31, 0.55)<0.001Model a: no adjusted; Model b: adjusted for age, sex; Model c: adjusted for age, sex, coronary heart disease, drinking, family history, smoking, chemotherapy, radiotherapy, diabetes, hypertension. CI, confidence interval; HR, hazard ratio; SII, systemic immune-inflammation index. Univariate and multivariate Cox regression analyses of overall survival Model a: no adjusted; Model b: adjusted for age, sex; Model c: adjusted for age, sex, coronary heart disease, drinking, family history, smoking, chemotherapy, radiotherapy, diabetes, hypertension. CI, confidence interval; HR, hazard ratio; SII, systemic immune-inflammation index. To evaluate whether the SII-survival association was consistent across the measured clinical strata, subgroup analyses were conducted according to demographic characteristics, comorbidities, and registry-available treatment indicators. As shown in Figure S3, higher SII was associated with a lower hazard of death across most of the examined subgroups. Estimates were less precise in several subgroups with small sample sizes, including patients with coronary artery disease or a family history of malignancy, but no clear reversal in the direction of the association was observed. These findings support the consistency of the SII-survival association across the measured clinical characteristics, although the subgroup analyses could not account for unmeasured differences in leukemia subtype or disease severity. The results remained consistent across multiple sensitivity analyses in which SII was modeled as a continuous variable, a dichotomous variable, or in quartiles. These analyses reduced the likelihood that the observed findings were solely dependent on the data-driven dichotomization threshold. Higher SII was persistently associated with reduced mortality, and a significant dose-response trend was observed (all P for trend <0.001; Table S1). RCS analysis To further characterize the shape of the association between SII and mortality risk, RCS analysis was performed. As shown in Figure S4, SII demonstrated a significant nonlinear inverse association with mortality risk (nonlinear P=0.002). Specifically, the hazard of death declined steeply at lower SII levels and then gradually plateaued at higher values, suggesting that the adverse prognostic impact was concentrated primarily among patients with very low SII. Nomogram construction and calibration To provide a comprehensive prognostic assessment grounded in real-world baseline clinical profiles, we developed a nomogram incorporating SII group along with smoking, drinking, hypertension, and diabetes (Figure 2). Each variable contributed a weighted point score, and the total score was mapped to predict individual probabilities of 1- and 3-year OS. Although these clinical covariates did not reach independent statistical significance in the fully adjusted Cox model (Table 2), they were retained to reflect standard, routinely collected patient characteristics that inform baseline risk stratification and lifestyle profiling. Within the nomogram, SII contributed the dominant prognostic weight. Calibration curves were subsequently plotted to evaluate the agreement between predicted and observed survival. As illustrated in Figure 3, the nomogram performed well across different time horizons. Taken together, these findings suggest that the SII-based nomogram may serve as a practical tool for individualized survival estimation in leukemia, although further external validation is required before broader clinical application. Nomogram for predicting 1- and 3-year overall survival in patients with leukemia. SII, systemic immune-inflammation index; prob., probability. Calibration plots for the SII-based nomogram at 1 and 3 years in the training and validation cohorts. OS, overall survival; SII, systemic immune-inflammation index. Discussion In this multicenter retrospective cohort of 609 patients with leukemia, higher baseline SII was associated with better OS in both the training and validation cohorts. This association remained statistically significant after adjustment for the demographic, comorbidity, and treatment variables available in the registry, and RCS analyses further indicated a nonlinear relationship between SII and mortality risk. The SII-based nomogram also showed acceptable internal calibration for estimating 1- and 3-year survival. These findings identify a consistent association between baseline SII and survival within the INSCOC cohort. In solid tumors, SII has generally been associated with adverse prognosis, including in hepatocellular carcinoma, colorectal cancer, and lung cancer (13-15). Elevated neutrophil counts and thrombocytosis are thought to promote tumor progression by fostering an immunosuppressive microenvironment, facilitating angiogenesis, and impairing cytotoxic lymphocyte activity (16-20). Consequently, higher SII usually reflects heightened systemic inflammation and immune escape, correlating with worse survival outcomes. However, the literature in hematologic malignancies is more heterogeneous and appears to be disease-context dependent. In a retrospective cohort of 300 patients with mixed hematologic malignancies, Ucar et al. identified higher SII and pan-immune-inflammation value as independent predictors of inferior OS, with particularly strong associations in AML and multiple myeloma (21). Likewise, in de novo non-M3 AML, Nan et al. reported that an SII ≥43.3×109/L was associated with significantly shorter overall and event-free survival and incorporated SII into a prognostic nomogram (22). Earlier work by Zhang et al. also showed that a higher pretreatment NLR predicted worse overall and disease-free survival in 181 patients with newly diagnosed non-M3 AML (23). Similar findings have been described in aggressive lymphoid neoplasms: in newly diagnosed diffuse large B-cell lymphoma, an NLR ≥6.2 identified patients with inferior 24-month progression-free survival, while in relapsed/refractory adult T-cell leukemia/lymphoma treated with mogamulizumab, low lymphocyte-to-monocyte ratio, especially in combination with elevated lactate dehydrogenase, marked poorer outcomes (24,25). Collectively, these data indicate that, in many hematologic settings, elevated blood count-derived inflammatory indices continue to track with biologically aggressive disease and worse prognosis. The inverse association observed in our cohort should be interpreted in light of the leukemia-specific determinants of peripheral blood counts. Unlike in most solid tumors, the neutrophil, lymphocyte, and platelet counts that constitute SII may be directly altered by marrow infiltration, ineffective hematopoiesis, infection, disease status, and the timing of blood sampling. Moreover, the same SII value may arise from biologically distinct combinations of thrombocytopenia, neutropenia, or relative lymphocytosis. Very low SII may therefore identify a heterogeneous group of patients with impaired marrow function or greater baseline clinical vulnerability rather than a favorable low-inflammatory state. In this context, SII may function less as a conventional marker of tumor-promoting systemic inflammation and more as a composite signal integrating inflammation, marrow function, and host reserve. The steep decline in mortality risk across the lowest SII range in the spline analysis is compatible with this interpretation, but it does not establish that cytopenias, infection, bleeding, or disease burden caused the observed association. Accordingly, the better survival observed at higher SII levels should not be interpreted as evidence of a direct protective effect of SII. Available evidence suggests that blood count-derived inflammatory indices do not carry a fixed biological meaning across all hematologic malignancies. In disorders characterized by marrow failure, treatment-related cytopenias, or profound susceptibility to infection, lower count-derived indices may reflect impaired hematopoietic reserve rather than a favorable inflammatory milieu. In myelofibrosis, Ersal et al. observed lower PLR and SII values among patients with fatal outcomes (26), and in hematologic malignancy patients with chemotherapy-induced neutropenia, reduced monocyte-to-lymphocyte ratio (MLR), PLR, and SII were associated with bloodstream infection risk (27). In myelodysplastic syndrome, inflammatory indices were linked to disease burden, immunophenotypic features, and survival, but their clinical interpretation varied across disease subsets (28). Together, these studies support the view that blood count-derived indices do not carry a uniform biological meaning across hematologic diseases and must be interpreted in relation to marrow function and disease context. Several limitations should be acknowledged. First, the registry did not capture detailed leukemia subtype or key measures of disease severity, including disease status, marrow or peripheral blood blast burden, cytogenetic and molecular risk, and established subtype-specific prognostic classifications. Residual confounding by leukemia burden therefore remains possible, and the present analysis cannot determine whether SII provides prognostic information independent of, or incremental to, established leukemia-specific risk factors and prognostic tools. The observed association should accordingly be interpreted within the set of covariates available in the registry. Second, the retrospective design introduces the possibility of selection bias, unmeasured confounding, and heterogeneity in clinical management across participating centers. Third, several factors that may influence both peripheral blood counts and survival—including infection, transfusion support, growth-factor use, treatment regimen and intensity, treatment response, transplantation, and subsequent therapy—were incompletely captured. Although complete blood counts were classified as pretreatment measurements, their precise timing relative to diagnosis, supportive interventions, and transfusion was unavailable, and baseline SII may therefore not have represented a uniform clinical state across patients. Fourth, SII was measured at a single time point, precluding assessment of temporal variation and treatment-related changes. Finally, the validation cohort was generated by random division of the same registry population and thus provided internal rather than independent external validation. The generalizability and predictive performance of the model should be evaluated in external, clinically well-characterized, subtype-specific cohorts. Taken together, these limitations preclude the use of SII as a stand-alone prognostic measure but support its further evaluation in more rigorously characterized leukemia populations. Conclusions In this multicenter cohort, higher baseline SII was consistently associated with better OS in patients with leukemia. This finding highlights the prognostic relevance of a routinely available blood count-derived index while suggesting that SII may reflect underlying disease burden, hematopoietic function, and host physiological reserve rather than an independent biological effect. Given the lack of detailed leukemia subtype, disease status, and cytogenetic and molecular risk information, these findings should be interpreted cautiously. Further validation in well-characterized, subtype-specific cohorts is warranted to determine whether SII provides prognostic information beyond established leukemia-specific risk factors. Supplementary The article’s supplementary files as