Constructing and Validating a Nomogram Model for Short-Term Prognosis of Patients with AChR-Ab+ GMG

Myasthenia gravis (MG) is an autoimmune neuromuscular disorder characterized by fluctuating muscle weakness and fatigability [1, 2]. It occurs when antibodies attack and disrupt the communication between nerves and muscles at the neuromuscular junction. Despite its rarity with a prevalence ranging from 15 to 179 per million people [3], the impact of MG on patients’ quality of life and health care systems is substantial [4].

Treatment response in MG is variable and can range from complete symptom relief to minimal improvement or even disease progression [5]. Unfortunately, a significant proportion of patients have a suboptimal response to therapy. Studies report that up to 10–20% of patients do not achieve satisfactory control of their symptoms despite treatment [6], highlighting the need for better therapeutic strategies and prognostic tools.

The ability to predict which patients are more likely to have a poor response to treatment would have significant clinical implications. Early prediction of treatment response can guide personalized treatment plans, potentially enabling more aggressive or alternative therapies for patients identified as high-risk. This can improve the therapeutic success rate, mitigate disease progression, decrease hospitalization rates, and improve quality of life [7]. Furthermore, it can facilitate patient counseling regarding the prognosis and expected course of the disease, which is important in disease acceptance and treatment adherence [7].

Current prediction of treatment response primarily relies on clinical experience and certain indicators such as disease duration and severity [8‐10]. However, these measures are not always reliable or precise. Therefore, there is a pressing need to develop reliable, data-driven prediction tools to guide therapeutic decision-making in MG. Zhao et al. developed a model for the 12-month outcome of AChR-Ab+ (acetylcholine receptor antibody-positive) GMG (generalized myasthenia gravis) patients, which could predict short-term prognosis effectively [11]. However, the treatment responses of a shorter term are also of great importance. Physicians and patients are more interested in whether a balance between efficacy and side effects can be achieved within a short period.

In the realm of clinical decision-making, particularly in neurology, the nomogram stands as a pivotal tool, bridging the gap between statistical predictions and practical patient care. A nomogram is a graphical representation tool that simplifies the statistical prediction model into a user-friendly interface, allowing for the visualization and calculation of the probability of a clinical event based on multiple predictors. In the context of AChR-Ab+ GMG, the complexity of the disease necessitates an approach that can encompass multiple variables to predict patient outcomes accurately. Here, the nomogram’s ability to integrate various clinical parameters into a single predictive model is invaluable. It aids neurologists in making informed, personalized treatment decisions, aligning with the evolving paradigm of precision medicine in neurology. The present study aims to develop and validate a predictive nomogram for poor shorter-term (6 months) outcomes in AChR-Ab+ GMG patients.

Patient Characteristics:

Patient recruitment for the study is presented in the flow diagram (Fig. 1). Ninety of 248 patients were enrolled. The median age at disease onset was 55 (IQR: 40.500–64) years, and 46.700% were women. A total of 30 patients were ‘unchanged’ or ‘worse’ after 6-month standard treatment (Table 1).

In our study, we have constructed a predictive nomogram to estimate the short-term response to treatment in AChR-Ab+ GMG patients based on six significant factors, namely gender, BMI, QMG score, disease duration, Hb, and WBC count. This nomogram represents a step towards a more personalized, data-driven approach to the management of MG.

The finding that female gender is predictive of poorer response aligns with the recognized gender disparities in autoimmune diseases, including MG. Females have been shown to have a higher susceptibility and more severe disease course [16]. Hormonal and genetic factors have been proposed to underlie these disparities, and a deeper exploration of these aspects might refine the predictive power of gender in our model and contribute to a better understanding of sex-related differences in MG outcomes.

Prior literature has consistently indicated a linkage between elevated BMI and heightened clinical symptomatology in MG patients [16, 17]. However, there is a conspicuous dearth of research elucidating its implications on short-term prognostic outcomes. Within our investigational purview, an augmented BMI emerged as a potential harbinger of improved short-term outcomes. This correlation may stem from the perturbation in the disease’s natural trajectory, likely attributable to the initiation of a more rigorous therapeutic regimen for these particular patients during the disease’s early stages. For instance, within our dataset, 33.33% of patients with a BMI > 25.610 were administered a combined regimen of steroids and tacrolimus compared to 19.610% of those with a BMI < 25.61 who received the dual therapy. Nevertheless, it remains imperative to underscore the significance of adept weight management in the holistic care paradigm for MG.

Our model’s emphasis on high QMG scores and longer disease duration as indicators of poorer response is congruent with the understanding that these factors are generally linked to disease severity and chronicity. This stresses the need for early, aggressive intervention in patients presenting with severe symptoms and those with long-standing disease to improve treatment outcomes. This result is consistent with the results of some previous studies [14], but some studies have shown that patients with higher QMG scores at initial diagnosis have better prognosis [11, 18]. These differences may be explained by the longer observation periods of these studies, during which more drug doses may be administered, and by greater attention from physicians and patients with more severe symptoms. Second, changes in the definition of outcomes according to QMG scores in these studies may have influenced the results of the statistical analyses.

The inclusion of Hb and WBC levels in our model underscores the importance of considering the overall health status of MG patients. Suboptimal Hb levels coupled with elevated WBC counts potentially indicate systemic health anomalies—encompassing inflammatory responses, infectious processes, or persistent pathological conditions—that could influence the therapeutic responsiveness in MG. The amplification in WBC count, along with derivative markers like the neutrophil-to-lymphocyte ratio (NLR), predominantly indicates an underlying systemic inflammatory condition. Notably, NLR has emerged as a prognostic marker for a plethora of pathologies, as evidenced by several research publications [19‐21]. In a retrospective investigation by Hsu et al., leukocytosis was identified as a robust predictor for mortality in MG [22]. Complementing this, numerous studies have elucidated the association of an elevated NLR with an adverse prognosis in MG [23‐25]. Anemia’s role as a harbinger of adverse outcomes spans various diseases including but not limited to ischemic stroke and chronic obstructive pulmonary disease [26, 27]. Notably, a plethora of clinical studies have documented the occurrence of anemia in MG cohorts [28, 29]. Certain academic contributions suggest a heightened severity of MG in anemic individuals, although the underlying mechanisms remain an area of ongoing inquiry [22, 30]. Extrapolating from ischemic stroke analyses, some scholars propose that anemia might inflict localized skeletal muscle hypoxia by compromising tissue oxygenation, subsequently diminishing muscular prowess [31]. Such a hypothesis might elucidate the observation where a decrement in Hb levels correlates with exacerbated MG presentations.

The direction our study has taken towards developing predictive models for patient outcomes mirrors broader trends in MG research and healthcare. Machine learning and other advanced analytics tools are being employed to construct robust predictive models for disease outcomes, as illustrated by the study of Huahua Zhong et al. [15]. This move towards personalized medicine allows clinicians to tailor treatment strategies based on a patient’s individual disease characteristics and potential response to therapy.

Our study is an addition to the suite of tools being developed to predict disease outcomes in MG. However, it is important to note that further validation of our model in larger, multicenter cohorts is needed to ensure its broad applicability and robustness. The aim is to develop a model that can be used across various patient populations, thereby contributing significantly to improving the quality of life for individuals living with MG.

Historical literature has posited that factors such as dysphagia, hypoproteinemia, presence of anti-titin antibodies, and an advanced age of onset might be indicative of a worsening patient prognosis [13, 32, 33]. In the current investigation, variables including bulbar symptoms, thymoma, serum albumin levels, and the presence of anti-titin antibodies were precluded because their P-values exceeded 0.2 in the univariate logistic regression analysis. It is pertinent to note that the limited sample size necessitated stringent inclusion criteria. Consequently, the ‘age of onset’ variable, failing to meet these stipulated criteria, was not incorporated into the analysis.

In our endeavor to establish a nomogram for predicting short-term efficacy in AChR-Ab+ GMG patients, several limitations merit consideration. First, the study was conducted at a single institution and, with a cohort of only 90 patients, might reflect regional-specific biases and limitation of small sample size, thereby limiting its generalizability. Additionally, the retrospective design inherently introduces potential biases, such as selection bias. Furthermore, while our statistical models are conventional, they operate under certain assumptions, and these might not always mirror the nuances of complex clinical scenarios. Lastly, the model’s external validation across diverse settings remains imperative to affirm its universal applicability. In light of these considerations, we strongly advocate for future multicentric research with larger sample sizes to enhance the model’s robustness.

In conclusion, the predictive nomogram developed in our study provides an additional tool for the personalized management of MG. Incorporating key demographic, clinical, and laboratory parameters, our model reflects the multifaceted nature of MG prognosis. With further validation and potential refinement, this model could play a vital role in guiding therapeutic decisions, ultimately leading to improved patient outcomes.