Artificial intelligence-driven strategies for managing renal and urinary complications in inflammatory bowel disease

Abstract

In this editorial, we discuss the article by Singh et al published in World Journal of Nephrology, stating the need for timely adjustments in inflammatory bowel disease (IBD) patients' long-term management plans. IBD is chronic and lifelong, with recurrence and remission cycles, including ulcerative colitis and Crohn's disease. It's exact etiology is unknown but likely multifactorial. Related to gut flora and immune issues. Besides intestinal symptoms, IBD can also affect various extraintestinal manifestations such as those involving the skin, joints, eyes and urinary system. The anatomical proximity of urinary system waste disposal to that of the alimentary canal makes early detection and the differentiation of such symptoms very difficult. Various studies show that IBD and it's first-line drugs have nephrotoxicity, impacting the patients' life quality. Existing guidelines give very few references for kidney lesion monitoring. Singh et al's plan aims to improve treatment management for IBD patients with glomerular filtration rate decline, specifically those at risk. Most of IBD patients are young and they need lifelong therapy. So early therapy cessation, taking into account drug side effects, can be helpful. Artificial intelligence-driven diagnosis and treatment has a big potential for management improvements in IBD and other chronic diseases.

Key Words: Inflammatory bowel disease; Renal complications; Artificial intelligence; Long-term management; Nephrotoxicity

Core Tip: Inflammatory bowel disease (IBD) requires long-term continuous drug therapy. However, many current first-line drugs exhibit nephrotoxicity or potential nephrotoxic induction with prolonged use. Artificial intelligence can assist in analyzing and adjusting treatment plans, offering theoretical prospects for long-term IBD management. Establishing a comprehensive and accurate database of clinical predictive indicators is crucial.

INTRODUCTION

Inflammatory bowel disease (IBD) is a group of chronic, lifelong diseases characterized by repeated alternations between flare-ups and remissions. The clinical classification includes ulcerative colitis (UC) and Crohn's disease (CD). Clinical manifestations of IBD feature intestinal symptoms including diarrhea, abdominal pain, bloody stools, and intestinal obstruction, as well as extensive extraintestinal multisystem lesions such as the skin, joints, eyes, and urinary system. These lesions, observed outside the digestive system, are collectively referred to as extraintestinal manifestations of IBD[1,2]. Although the current etiology is still unclear, it is generally believed that the joint action of genetic and environmental factors forms it. In recent years, studies have also found its relationship with the abnormal activation of the immune system and the microecology of the intestinal flora[3,4]. IBD patients can develop the disease at any age, but it mainly concentrates in the young. This results in a projected up to 25% of IBD cases developing before an individual attains the age of 20 years. In fact, about 20% of the IBD patients contracted the disease at an age below 10 years, and even 5% at an age below 5 years. The chronic nature of the disease implies that the patients will need long-term treatment, which not only translates into high treatment costs but also means that the long duration of administering the medication serves as a test to ascertain the long-term effectiveness and side effects of the treatment drugs. It is increasingly agreed by researchers that more clinical detection indicators need to be explored in order to adjust or even stop the treatment plan in time. The new treatment goal should put more weight on improving quality of life, and it has to be admitted that IBD itself and treatment drugs have a great impact on the physical and mental health of patients[5-7].

TREATMENT CHALLENGES AND AI POTENTIAL

Although some treatments for IBD have been developed and improved, the unknown etiology of IBD presents challenges. The goal of traditional treatment is mainly to induce and maintain remission, but there are clear limitations in the various treatment methods. Nephrotoxicity from aspirin-sensitive asthma (ASAs) and cyclosporine A (CsA) has been recognized, and tumor necrosis factor-alpha inhibitors carry a significant potential risk for triggering impaired renal function. While biological agents and small molecule drugs have not shown obvious renal toxicity, some case reports indicate they may potentially lead to adverse reactions. This is especially true for dual biological therapy in refractory patients, which has become increasingly common in recent years. Although it has shown good clinical treatment effects, potential side effects should not be ignored[8-10].

Meštrović et al[6] emphasizes the importance of timely adjustments to the treatment plan and drug discontinuation when needed. While randomized clinical trials on drug discontinuation are lacking, it is crucial to identify potential predictors of disease recurrence before deciding to stop treatment or alter the treatment plan. The fecal calprotectin level has been identified as the strongest predictor of recurrence. Other recognized predictors include a high CD activity index, young age of onset (under 40 years), long disease duration (over 40 years), smoking, and the use of steroids within six months to a year before drug discontinuation. An important factor in deciding to stop treatment is the success of retreatment with the same or different drugs. Treatment discontinuation must be approached individually, considering the severity, extent, and duration of the disease, possible side effects, risk of recurrence, and patient preferences[11].

It is therefore evident that the pivotal factor in the rational modification of the treatment plan is the ability to make a thorough and precise evaluation of the patient's specific circumstances in accordance with an established criterion. In addition to the aforementioned factors, such as fecal calprotectin level, high CD activity index, age of onset, duration of disease, and smoking history, the personal experience of the attending physician should also be taken into consideration. It is also necessary to consider the patient's actual medication history. IBD patients require regular follow-up examinations, and the academic community must continue to explore and study IBD in order to ensure that the aforementioned content is adjusted dynamically and that patients derive the greatest possible benefit. Consequently, this work has become more challenging. At this time, the significance of artificial intelligence (AI) is becoming increasingly evident. With its extensive database capacity and rapid processing capabilities, AI can not only accurately assess the individual condition of IBD patients in the context of evidence-based medicine, thereby providing enhanced patient benefits, but also reduce the workload of attending physicians with its assistance, thereby promoting the rational allocation of medical resources.

KIDNEY AND URINARY SYSTEM INVOLVEMENT

Both kidneys and the urinary system are located in an anatomical position on the body surface adjacent to the gastrointestinal tract, making early detection and differentiation difficult. It is reported that IBD itself and most first-line treatment drugs present varying degrees of direct or potential kidney damage, which significantly impacts patients' quality of life and is a clinical symptom that deserves attention. However, current IBD diagnosis and treatment guidelines still lack monitoring and guidance methods for patients with secondary renal lesions. 4% to 23% of adult IBD patients have renal manifestations, but only 1% to 2% of child patients experience kidney involvement. The specific mechanism is still unknown. Disorders in immune responses, susceptible sites, and shared signaling pathways may all play a role, but exploratory research in this area continues[12,13]. The range of secondary renal lesions in IBD patients is extensive, including various types of kidney stones, chronic kidney disease (CKD), acute kidney injury (AKI), glomerulonephritis (GN), interstitial nephritis, nephrotic syndrome, kidney and urinary tract tumors, and the nephrotoxicity of treatment drugs.

THE RENAL EFFECTS OF IBD AND FIRST-LINE TREATMENT DRUGS

In a controlled study conducted in Saudi Arabia with 75236 IBD patients, it was found that the likelihood of being diagnosed with urolithiasis was more than twice as high for IBD patients compared to the general population, with the risk being notably higher for those with CD[14]. Additionally, research has shown that male patients have a higher susceptibility to urinary stones, and surgical procedures can be a risk factor for secondary kidney stones in IBD patients[15]. Onime et al[16] highlighted that terminal ileum dysfunction can cause malabsorption of bile salts and fatty acids. As a result, instead of forming a calcium oxalate complex, calcium binds to fatty acids, leading to increased intestinal absorption of oxalic acid and the development of "enteric hyperoxaluria" (urine oxalic acid concentration > 45 mg/dL). Furthermore, Parmar et al[17] suggested that changes in the permeability of diseased colonic mucosa to oxalate and the decolonization of oxalobacter in the intestine reduce colonic oxalate catabolism. The use of total parenteral nutrition and impaired liver function can also lead to endogenous oxalate synthesis and oxalate stone formation. Dehydration from diarrhea can contribute to uric acid stone formation due to urine concentration[18].

Ambruzs et al[19] reported a few cases of calcium phosphate stones in children with IBD, with the mechanism possibly related to bone mobilization from calcium deficiency. Abdulrhman et al[20] found that vedolizumab could increase kidney stone incidence, although the exact mechanism is unclear; however, the incidence can decrease after drug withdrawal.

Regarding active IBD and CKD, no definitive evidence of an association currently exists. Yang et al[21] conducted a retrospective study with 10117 participants and found that the risk ratio for CKD development in IBD patients was 1.24 (95%CI: 1.10-1.40). Nevertheless, further analyses exploring links between IBD, microalbuminuria, and glomerular filtration rate (GFR) yielded inconclusive results. Wu et al[22] established a causal relationship between UC and elevated urine albumin levels in a genome-wide association study model, but in another study with 86 samples, no connection between active IBD and proteinuria was identified. The clinical evidence regarding IBD and GFR is conflicting, with Jang et al[23] suggesting that cystatin C could be more sensitive than creatinine levels as a monitoring indicator[24].

In patients receiving 5-ASA treatment, up to 1 in 100 experience renal impairment. Patel et al[25] found that creatinine clearance decreased in a dose- and duration-dependent manner in IBD patients. The occurrence of tubulointerstitial nephritis, GN, and other conditions may be related to delayed hypersensitivity reactions induced by the treatment. IgA nephropathy is the most common form of IBD-related nephrotic syndrome, with intestinal mucosal immune responses causing elevated antibody levels, and increased intestinal permeability allowing more antigens to enter the bloodstream. These factors contribute to IgA-containing immune complexes depositing in the glomeruli, leading to IgA nephropathy[26]. Rehnberg et al’s retrospective analysis found that the risk of IBD and nephropathy co-occurrence was higher in patients with IgA nephropathy, and that IBD patients with IgA nephropathy were at higher risk of progressing to end-stage kidney disease, indicating a vicious cycle between the two conditions[27]. The risk of IBD patients developing renal and urothelial tumors is five times that of the general population, potentially due to abnormal immune responses and prolonged inflammatory states in these patients, though consensus on this is lacking[28,29]. Sandborn et al[30] reported that approximately 6% of IBD patients treated with high-dose CsA experienced kidney damage. Long-term use of immunomodulators and immunosuppressants carries risks, including infections and malignancies.

Although biologics play an important role in managing IBD, there is limited research on their long-term safety. Both vedolizumab and infliximab have been linked to tubulointerstitial nephritis and even permanent renal damage[31,32].

DISCUSSION

At present, there is no uniform standard for the assessment of renal function in IBD patients internationally. The predictive value for secondary AKI under Singh et al's early urine and computed tomography (CT) imaging screening urgently required an approach that could guide and improve the management of IBD patients[33]. This appraisal scheme initiates from the deterioration of GFR. The IBD disease itself, however, and clinical indicators including microalbumin and creatinine were not found to have a causal relationship in the retrospective statistical data analysis of relevant cases at the present time. If one does not consider deviation caused by sample size, then novel and more accurate cystatin C might be an indicator of GFR that is worth investigating.

Both the disease course and management of the patient with IBD are complicated and long-term procedures. It will change along with the occurrence, development, and long-term treatment of this disease because of the complex mechanisms of mutual influence for various indicators; hence, the formulation and adjustment of a treatment plan in time has become a heavy burden. Singh et al[33] conducted a detailed analysis of a series of renal and urinary complications associated with IBD patients. The article presents an in-depth examination of various conditions, including amyloidosis, nephrolithiasis, tubulointerstitial nephritis, GN, obstructive pathologies, and CKD. The author posits that, on the one hand, the surface projections of the renal and urinary systems in anatomical positions are in close proximity to or overlap with those of the gastrointestinal tract. This, in turn, renders the early detection of this particular extrain-testinal manifestation more challenging. Conversely, the inflammatory damage caused by IBD itself and the related medications has been observed to result in varying degrees of renal harm or potential harm to the kidneys. Furthermore, there is currently no consensus within the academic community on the monitoring of renal function in these patients. Consequently, the author proposes a novel diagnostic examination plan. Initially, urine analysis should be conducted for IBD patients exhibiting a decline in GFR that is not attributable to dehydration. Subsequently, the results can be classified into one of three categories: (1) In cases of proteinuria, existing pharmacological treatments should be discontinued and steroids employed for the purpose of providing protective anti-inflammatory therapy. Subsequently, the proteinuria level should be reassessed. Should there be an improvement, an alternative treatment option may be more appropriate. In the absence of improvement, it is recommended that a nephrologist be consulted at the earliest opportunity, and that a kidney biopsy be considered when appropriate; (2) In cases where there is an increase in oxalate, phosphate, and uric acid in the urine, a reduction in urine pH, and a decrease in urine volume, a renal ultrasound examination is advised; and (3) In cases where there is a reduction in GFR but no overtly abnormal findings in the urine analysis, a CT scan is essential to exclude the possibility of intestinal obstruction, enterovesical fistula, and renal cell carcinoma.

Although the diagnostic examination scheme proposed by Singh et al[33] has of course become more sophisticated in recent years, the principle remains the same. It is useful to improve the management in IBD patients with secondary decline in GFR but has little utility for the management of all IBD patients[34].

While this innovative approach to diagnosis and treatment offers substantial advantages in the early detection and management of renal and urinary system complications in IBD patients, it is important to acknowledge the potential limitations in its practical application. In addition to the increased economic costs for patients undergoing follow-up examinations and the additional time costs incurred by attending physicians, ensuring uniformity across the entire diagnostic and treatment process represents a more significant challenge. The formulation and adjustment of IBD remedies are based on high-quality assessments of kidney biopsy images, cross-sectional imaging such as CT scans, and even endoscopic pictures, which are used as a routine diagnostic examination for IBD[35]. However, the quality of these assessments is dependent on the subjective judgement of clinicians or examiners. Individuals with varying levels of experience may encounter challenges related to inter-observer and intra-observer variations, as well as potential biases[36]. The incorporation of AI-based image analysis tools could significantly enhance the accuracy and consistency of these processes. The initial step is to convert digital CT, magnetic resonance imaging, endoscopic videos, histological images, and other types of images into discrete mathematical features. These features include brightness, texture, color balance, and relationships between pixel groups. The aforementioned features, which are employed as data inputs for AI models, are utilized to identify and match corresponding patterns, which in turn constitute the output. These outputs may include expert interpretations, whole disease phenotypes, and even the treatment contents and prognosis of patients. Following training with thousands of existing samples, it is possible to obtain accurate probabilities for the evaluation of new images. Moreover, image segmentation can automatically identify the most salient or meaningful features within an image, such as neutrophils on the intestinal wall boundary in CT images. The term "image classifier" is used to describe a system that is capable of classifying an entire image. They replicate and integrate the specific assessments of domain experts, while segmentation provides measurements or counts of symbolic features in an image to obtain more discrete disease evaluations[37,38]. Barash et al[39] and Klang et al[40] demonstrated that AI can not only alleviate the burden of video capsule endoscopy (VCE) review and interpretation but also automatically identify intestinal ulcers in VCE. The evaluation has been proven to be of commensurate or higher quality to that of human examiners (κ = 0.78, 95%CI: 0.716–0.844). Furthermore, large language model (LLM) represents another manifestation of AI. LLM are capable of automatically extracting textual information and replicating conversations or generating anticipated responses for related events through targeted training. ChatGPT is an automatic dialogue model trained with data extracted from the internet on the basis of being supported by LLM. Although the accuracy rate of GPT-4 in answering the 145 questions proposed in the American college of gastroenterology self-assessment from 2021 to 2022 was only 62.4% to 65.1%, this can be enhanced through more targeted training and technological renovation. In 80% of cases, GPT-4 assigned the same UC evaluation as gastroenterologists did. Nevertheless, only 50%–60% of patients concurred with GPT-4's proposed treatment, suggesting that the machine may not have accounted for additional factors. It is only a matter of time before LLM will be able to care for IBD patients independently, although this is not yet the case[41,42].

Duttagupta et al[43] also did a controlled study with 20 patients having UC and 20 healthy controls for analyzing the microRNA profile using support vector machine. The measurement results showed that the accuracy of the prediction score was 92.8%, specificity was 96.2%, and sensitivity was 89.5%. Boal Carvalho et al[44] reviewed the evidence of several UC drugs for their capacity to maintain mucosal healing (MH), and he concluded that MH has an important correlation with the clinical recurrence rate, surgical rate, and incidence of colorectal malignancies in UC patients. It is therefore recommended to use MH as the treatment endpoint for UC patients. MH includes endoscopic remission and histological improvement. Intestinal microvasculature manifestations are the main feature of endoscopic MH[45]. Since this is an indicator that can detect UC inflammation and give guidance on the prognosis, diagnosing intestinal microvascular manifestations usually requires experienced professional endoscopists; consequently, there are limited wide applications in clinics, and the AI assistant can make up for this defect. Study design: It was a one-year prospective cohort study by Maeda et al[46]. Participants: The study population included 135 patients. Results: The recurrence rate of patients within the active group diagnosed with AI was significantly higher than that of the healing group under AI diagnosis, with 95%CI at 28.4% (21/74): 18.5% to 40.1% and 4.9%(3/61): 1.0% to 13.7%; P < 0.001[36]. Subsequently, Turan and Durmus[47] developed the UC-normalizer-free networks classification method in order to do the detailed classification of the colonoscopy images of patients with UC and guide individualized treatment; the effect was evident and obviously much better than that of the classification model by normalizer-free networks, conditional vision transformer, ResNets, and Inception-v4.

CONCLUSION

AI will greatly promote the analysis, integration, and interpretation of big data sets of IBD. Not only can it update in time the latest diagnosis and evaluation indexes, but AI can also simulate the empirical judgment process of authoritative clinical experts with the maximum effect of IBD management. For treatment, AI is a tool to help outline a detailed personal condition, such as age, lesion degree, complications, and many others, which would lead to personal treatment. It can not only provide the best possible treatment effect but also reduce unnecessary use of drugs with side effects. It can also be used in monitoring the patient's condition. These day-to-day variations of the condition can be assessed in real time with any data retrieved from portably wearable devices developed through research and development. Combining the reported patient symptoms with data such as heart rate, body temperature, or exercise status enables assessment of any variations that occur in their condition. However, the utilization of AI-assisted treatment of IBD still encounters many challenges, while the quality and security of data is a significant one. Diagnostic and therapeutic biases of major consequences may ensue from even the smallest error or omission in the data. Furthermore, improvements are still to be realized regarding AI systems and related facilities. The accumulation of clinical doctors' practice and experience in the management of chronic diseases is a long process, but AI assistance just solves this problem. Almost all practitioners' experience, treatment evidence, and the latest research results could upload to the database in a short time. Thus, with the help of AI in treating IBD, optimal management of the disease can be achieved within a relatively short time. In general, AI-assisted IBD treatment indeed has enormous potential; however, it has to be continuously improved and regulated regarding technology, ethics, and law to ensure a safe and effective application in clinical practice.