Medical checkup data analysis method based on LiNGAM and its application to nonalcoholic fatty liver disease.

Although medical checkup data would be useful for identifying unknown factors of disease progression, a causal relationship between checkup items should be taken into account for precise analysis. Missing values in medical checkup data must be appropriately imputed because checkup items vary from person to person, and items that have not been tested include missing values. In addition, the patients with target diseases or disorders are small in comparison with the total number of persons recorded in the data, which means medical checkup data is an imbalanced data analysis. We propose a new method for analyzing the causal relationship in medical checkup data to discover disease progression factors based on a linear non-Gaussian acyclic model (LiNGAM), a machine learning technique for causal inference. In the proposed method, specific regression coefficients calculated through LiNGAM were compared to estimate the causal strength of the checkup items on disease progression, which is referred to as LiNGAM-beta. We also propose an analysis framework consisting of LiNGAM-beta, collaborative filtering (CF), and a sampling approach for causal inference of medical checkup data. CF and the sampling approach are useful for missing value imputation and balancing of the data distribution. We applied the proposed analysis framework to medical checkup data for identifying factors of Nonalcoholic fatty liver disease (NAFLD) development. The checkup items related to metabolic syndrome and age showed high causal effects on NAFLD severity. The level of blood urea nitrogen (BUN) would have a negative effect on NAFLD severity. Snoring frequency, which is associated with obstructive sleep apnea, affected NAFLD severity, particularly in the male group. Sleep duration also affected NAFLD severity in persons over fifty years old. These analysis results are consistent with previous reports about the causes of NAFLD; for example, NAFLD and metabolic syndrome are mutual and bi-directionally related, and BUN has a negative effect on NAFLD progression. Thus, our analysis result is plausible. The proposed analysis framework including LiNGAM-beta can be applied to various medical checkup data and will contribute to discovering unknown disease factors.

Risk for Colorectal Cancer in Patients with Serially Positive Fecal Immunochemistry Test in an Annual Screening Program.

OBJECTIVES: There are patients who do not undergo colonoscopy even if the fecal immunochemistry test (FIT) results are positive and even with repeated positive test results the following year. We aimed to investigate colorectal cancer (CRC) risk in examinees with positive FIT results in our annual screening program. METHODS: We analyzed patients who underwent initial colonoscopy from April 2010 to March 2017 because of positive FIT results using an endoscopy database in our hospital. We investigated the difference in the risk of advanced colorectal neoplasia as a surrogate marker of CRC between those who had an initial positive test and those who had repeated positive tests. RESULTS: A total of 748 patients were included in this analysis. The advanced neoplasia detection rates were 7.6% (50/656) and 18.5% (17/92) for the initial and repeated positive test groups, respectively. Subgroup analysis of those with repeated positive tests revealed that the detection rates in examinees with positive tests 1-2 and >2 years ago were 16.7% (6/36) and 19.6% (11/56), respectively. The odds ratios for advanced neoplasia detection in patients with positive tests 1-2 and >2 years ago compared with those in the initial positive test group were 2.72 (95% confidence interval [CI], 1.04-7.10) and 3.09 (95% CI, 1.47-6.48), respectively. CONCLUSIONS: The risk of CRC appears more than doubled in patients with a repeated positive FIT result. Prompt colonoscopy is recommended for FIT-positive cases.

Significance of fecal hemoglobin concentration for predicting risk of colorectal cancer after colonoscopy.

BACKGROUND AND AIM: As the significance of the quantitative fecal immunochemical test (FIT) in patients who previously underwent a colonoscopy is unknown, this study aimed at investigating the association between fecal hemoglobin concentration and the risk of colorectal cancer (CRC). METHODS AND RESULTS: We retrospectively analyzed FIT-positive patients who underwent a colonoscopy through our opportunistic annual screening program from April 2010 to March 2017 at the Kyoto Second Red Cross Hospital. We stratified them into no colonoscopy and past colonoscopy (>5 years or ≤5 years) groups based on whether they had a history of undergoing a colonoscopy and analyzed the correlation between fecal hemoglobin concentration and advanced neoplasia or invasive cancer detection in each group. We analyzed 1248 patients with positive FIT results. There were 748 (59.9%), 198 (15.9%), and 302 (24.2%) patients in the no colonoscopy, past colonoscopy (>5 years), and past colonoscopy (≤5 years) groups, respectively. In the no colonoscopy group, the advanced neoplasia detection rate significantly increased with the fecal hemoglobin concentration (P < 0.001). However, no significant trend was observed in the past colonoscopy (both >5 years and ≤5 years) group (P = 0.982). No invasive cancer was detected in the past colonoscopy (≤5 years) group. CONCLUSION: The risk of CRC might be low even if fecal hemoglobin concentration was high, especially in those who underwent colonoscopy within 5 years.

Over- and Under-sampling Approach for Extremely Imbalanced and Small Minority Data Problem in Health Record Analysis.

A considerable amount of health record (HR) data has been stored due to recent advances in the digitalization of medical systems. However, it is not always easy to analyze HR data, particularly when the number of persons with a target disease is too small in comparison with the population. This situation is called the imbalanced data problem. Over-sampling and under-sampling are two approaches for redressing an imbalance between minority and majority examples, which can be combined into ensemble algorithms. However, these approaches do not function when the absolute number of minority examples is small, which is called the extremely imbalanced and small minority (EISM) data problem. The present work proposes a new algorithm called boosting combined with heuristic under-sampling and distribution-based sampling (HUSDOS-Boost) to solve the EISM data problem. To make an artificially balanced dataset from the original imbalanced datasets, HUSDOS-Boost uses both under-sampling and over-sampling to eliminate redundant majority examples based on prior boosting results and to generate artificial minority examples by following the minority class distribution. The performance and characteristics of HUSDOS-Boost were evaluated through application to eight imbalanced datasets. In addition, the algorithm was applied to original clinical HR data to detect patients with stomach cancer. These results showed that HUSDOS-Boost outperformed current imbalanced data handling methods, particularly when the data are EISM. Thus, the proposed HUSDOS-Boost is a useful methodology of HR data analysis.

Combination of PNPLA3 and TLL1 polymorphism can predict advanced fibrosis in Japanese patients with nonalcoholic fatty liver disease.

BACKGROUND: Hepatic fibrosis is an independent risk factor for mortality and liver-related events in patients with nonalcoholic fatty liver disease (NAFLD). PNPLA3 rs738409 has been associated with fibrosis in viral and non-viral hepatitis. TLL1 rs17047200 also has been associated with developing hepatocellular carcinoma probably via hepatic fibrogenesis. We estimated the impact of these genetic polymorphisms on hepatic fibrosis in Japanese patients with NAFLD. METHODS: We analyzed the association between these genetic variants and the backgrounds of 817 individuals who received health checkups (health check cohort) from 2012 to 2014. Then, we investigated the relationship between genetic variants and liver histology in 258 consecutive patients with biopsy-proven NAFLD in Japan (NAFLD cohort) from 2012 to 2017 (UMIN000027399). RESULTS: The prevalence of PNPLA3 CG/GG in the NAFLD cohort was higher than that in the health check cohort (p < 0.001). The prevalence of patients with advanced fibrosis (stages 3-4) was higher for PNPLA3 genotype CG/GG than CC (p = 0.048) and for TLL1 genotype AT/TT than AA (p = 0.044). The high-risk group which had at least two risk alleles of these variants was more likely to have advanced fibrosis (p = 0.004). Multivariate analysis identified body mass index [odds ratio (OR) 1.123, serum AST (OR 1.037, p = 0.004], serum albumin (OR 0.247, p = 0.032), and genetic high risk (OR 2.632, p = 0.026) as predictors of advanced fibrosis. CONCLUSIONS: In Japanese patients with NAFLD, individuals with risk alleles of PNPLA3 and TLL1 may have a risk of advanced fibrosis.

Examination time as a quality indicator of screening upper gastrointestinal endoscopy for asymptomatic examinees.

BACKGROUND AND AIM: The significance of examination time of esophagogastroduodenoscopy (EGD) for asymptomatic examinees is yet to be established. We aimed to clarify whether endoscopists who allot more examination time can detect higher numbers of neoplastic lesions among asymptomatic examinees. METHODS: We reviewed a database of consecutive examinees who underwent EGD in our hospital from April 2010 to September 2015. Staff endoscopists were classified into fast, moderate, and slow groups based on the mean examination time of EGD without a biopsy. Neoplastic lesion detection rate among these groups was compared using multiple logistic regression. RESULTS: Of the 55 786 consecutive examinees who underwent EGD, 15 763 asymptomatic examinees who were screened by staff doctors were analyzed. Mean examination time of 13 661 EGD without biopsy was 6.2 min (range, 2-18 min). When cut-off times of 5 and 7 min were used, four endoscopists were classified into the fast (mean duration, 4.4 ± 1.0 min), 12 into the moderate (6.1 ± 1.4 min), and four into the slow (7.8 ± 1.9 min) groups. Neoplastic lesion detection rates in the fast, moderate, and slow groups were 0.57% (13/2288), 0.97% (99/10 180), and 0.94% (31/3295), respectively. Compared with that in the fast group, odds ratios for the neoplastic lesion detection rate in the moderate and slow groups were 1.90 (95% confidence interval [CI], 1.06-3.40) and 1.89 (95% CI, 0.98-3.64), respectively. CONCLUSION: Endoscopists who do not allot adequate examination time may overlook neoplastic lesions in the upper gastrointestinal tract.

Comparison of peg-interferon, ribavirin plus telaprevir vs simeprevir by propensity score matching.

AIM: To compare efficacy of telaprevir (TVR) and simeprevir (SMV) combined with pegylated interferon (PEG-IFN) and ribavirin (RBV) while treating chronic hepatitis C (CHC). METHODS: In all, 306 CHC patients were included in this study. There were 159 patients in the TVR combination therapy group and 147 patients in the SMV combination therapy group. To evaluate pretreatment factors contributing to sustained virological response at 12 wk (SVR12), univariate and multivariate analyses were performed in TVR and SMV groups. To adjust for patient background between TVR and SMV groups, propensity score matching was performed. Virological response during treatment and SVR12 were evaluated. RESULTS: Overall rates of SVR12 [undetectable serum hepatitis C virus (HCV) RNA levels] were 79.2% and 69.4% in TVR and SMV groups, respectively. Patients in the SMV group were older, had higher serum HCV RNA levels, lower hemoglobin, higher prevalence of unfavorable interleukin-28B (IL28B) genotype (rs8099917), and poorer response to previous PEG-IFN and RBV treatment. Propensity score matching was performed to adjust for backgrounds (n = 104) and demonstrated SVR12 rates of 74.0% and 73.1% in the TVR and SMV groups, respectively. In the TVR group, discontinuation rates were higher because of adverse events; however, breakthrough and nonresponse was more frequent in the in SMV group. Multivariate analysis revealed IL28B genotype (rs8099917) as the only independent predictive factor of SVR12 in both groups. CONCLUSION: SVR12 rates were almost identical following propensity score matching.

The Impact of PNPLA3 rs738409 Genetic Polymorphism and Weight Gain ≥10 kg after Age 20 on Non-Alcoholic Fatty Liver Disease in Non-Obese Japanese Individuals.

Non-alcoholic fatty liver disease (NAFLD) in non-obese individuals is inadequately elucidated. We aim to investigate the impact of known genetic polymorphisms on NAFLD and the interaction between genetic risks and weight gain on NAFLD in obese and non-obese Japanese individuals. A total of 1164 participants who received health checkups were included. Participants with excessive alcohol consumption, with viral hepatitis or other inappropriate cases were excluded. Fatty liver was diagnosed by ultrasonography. Participants with a body mass index (BMI) of <18.5 kg/m2, 18.5-22.9 kg/m2, 23.0-24.9 kg/m2 and ≥25 kg/m2 were classified underweight, normal weight, overweight and obese, respectively. Self-administered questionnaire for lifestyle was assessed and a total of 8 previously reported genetic polymorphisms were chosen and examined. In all, 824 subjects were enrolled. The overall prevalence of NAFLD was 33.0%: 0% in underweight, 15.3% in normal weight, 41.1% in overweight and 71.7% in obese individuals. The prevalence of NAFLD is more affected by the G allele of patatin-like phospholipase domain-containing protein 3 (PNPLA3) rs738409 in normal weight (odds ratio (OR) 3.52; 95%-CI: 1.42-8.71; P = 0.0063) and in overweight individuals (OR 2.60; 95%-CI: 1.14-5.91; P = 0.0225) than in obese individuals (not significant). Moreover, the G allele of PNPLA3 rs738409 and weight gain ≥10 kg after age 20 had a joint effect on the risk of NAFLD in the normal weight (OR 12.00; 95% CI: 3.71-38.79; P = 3.3×10-5) and the overweight individuals (OR 13.40; 95% CI: 2.92-61.36; P = 0.0008). The G allele of PNPLA3 rs738409 is a prominent risk factor for NAFLD and the interaction between the PNPLA3 rs738409 and weight gain ≥10 kg after age 20 plays a crucial role in the pathogenesis of NAFLD, especially in non-obese Japanese individuals.

Prevalence of and risk factors for non-alcoholic fatty liver disease in a non-obese Japanese population, 2011-2012.

BACKGROUND: Non-alcoholic fatty liver disease (NAFLD) in non-obese subjects is not rare in Japan, but it has not been clearly described. To clarify its prevalence and risk factors, we investigated the clinical characteristics of NAFLD in non-obese subjects in comparison with NAFLD in obese subjects in the Japanese general population. METHODS: A cross-sectional study was performed with 5433 subjects who received health checkups from 2011 to 2012. Subjects consuming more than 20 g of alcohol per day and those with autoimmune liver disease, viral hepatitis, uncontrolled biliary disease and insufficient data were excluded. Subjects with a body mass index (BMI) ≥25 kg/m(2) were considered obese, and subjects with a BMI <25 kg/m(2) were considered non-obese. RESULTS: A total of 3271 subjects were enrolled. The overall prevalence of NAFLD was 24.6%: 68.5% in obese subjects and 15.2% in non-obese subjects. The multivariate logistic regression analysis revealed that ≥10 kg of weight gain since the age of 20 was significantly associated with NAFLD in non-obese subjects of both genders, and eating an evening meal within 2 h before going to bed 3 days or more per week and drinking <20 g of alcohol per day were negatively associated in non-obese females. Metabolic factors such as waist circumference and triglycerides were predictors of NAFLD in non-obese subjects, and body fat percentage was a predictor in non-obese males. CONCLUSIONS: Lifestyle as well as metabolic factors may play crucial roles in the pathogenesis of NAFLD, even in the non-obese Japanese population.

[Severe advanced lower gingival carcinoma effectively reduced by chemoradiation with S-1--report of a case].

Severe advanced head and neck carcinoma which can not be removed via surgical procedure combined with a large lymph node metastasis has a poor prognosis. We administered concurrent chemoradiotherapy with S-1 for a lower gingival carcinoma. As a direct result, we discovered that the treatment greatly reduced the size of tumor, and we consider that this treatment prolonged the patient.s life. The treatment results suggest that the so-called dormancy state of the tumor was continued. In this case study, radiotherapy with S-1 showed a highly effective response from the viewpoint of QOL improvement.

The predictive value of liver fibrosis in determining the effectiveness of interferon and lamivudine therapies for chronic hepatitis B.

BACKGROUND: To determine the best indicator of the effective use of interferon and lamivudine for the treatment of hepatitis B e antigen-positive chronic hepatitis, we retrospectively analyzed histologic and virologic status in 200 patients who were treated with interferon and 45 patients who were treated with lamivudine. METHODS: Histological grading and staging scores were determined by international criteria and the METAVIR scoring system. The YMDD motif associated with lamivudine resistance was analyzed by the sequencing of hepatitis B virus (HBV) DNA. RESULTS: Of 200 interferon-treated patients, 62 (31%) seroconverted to anti-hepatitis B e (anti-HBe). Multivariate analysis showed that the significantly important predictors of response were a higher grading score (P = 0.0056) and lower staging score (P = 0.0010). Twenty (44%) of the 45 lamivudine-treated patients seroconverted to anti-HBe, and multivariate analysis showed that the significantly important predictors of response were a higher alanine aminotransferase (ALT) level (P = 0.0034) and lower hepatitis B e antigen levels ( P = 0.0128). YMDD mutations occurred during therapy in 12 patients (27%). The significantly important predictor of the development of mutation was a higher staging score (P = 0.0226). CONCLUSIONS: Both interferon and lamivudine were effective for patients with high ALT levels, but interferon's efficacy appeared to be limited by the degree of fibrosis. Lamivudine appeared to be effective irrespective of the degree of fibrosis, but YMDD mutations seemed to develop sooner in patients with advanced liver fibrosis.

Activated Kupffer cells play an important role in intra-hepatic Th1-associated necro-inflammation in Concanavalin A-induced hepatic injury in mice.

BACKGROUND/AIMS: To examine whether or not activated Kupffer cells play an important role in intra-hepatic Th1-associated necro-inflammation in Concanavalin A (Con A)-induced hepatic injury in mice. METHODS: Con A was administered to Balb/c mice pretreated with or without gadolinium chloride (GdCl(3)). Kupffer cell activation was evaluated by their ability to produce superoxide anions in situ under liver perfusion with nitro blue tetrazolium (NBT). Hepatic concentration of cytokines was measured by ELISA and the mRNA expression of CXC chemokine receptor 3 (CXCR3) was evaluated by RT-PCR. Immunohistochemical detection of CD4 positive lymphocytes in the liver was also performed. RESULTS: GdCl(3)-pretreatment significantly (P<0.01) reduced the serum levels of alanine aminotransferase (ALT) in Con A-treated mice. Formazan deposition in Kupffer cells, the hepatic concentration of tumor necrosis factor-alpha and interferon-gamma, the mRNA expression of CXCR3 and the CD4 positive lymphocytes in the liver were decreased in GdCl(3)-pretreated mice as compared with those without GdCl(3)-pretreatment (P<0.05, respectively). CONCLUSIONS: Activated Kupffer cells, which produce superoxide anions, are involved in Con A-induced hepatic necro-inflammation in mice possibly through the activation of Th1-associated immune response mediated by CD4 and/or CXCR3 positive cells recruited into the liver.