Introduction
Allogeneic haematopoietic cell transplantation (allo-HSCT) can cure many patients with certain malignant and nonmalignant haematologic disorders [
1]. However, it is associated with many acute and chronic complications, of which graft-versus-disease (GVHD) is one of the most frequent and severe, affecting the patient’s lifespan and quality of life [
2]. Generally, according to the time of onset, GVHD is classified as acute (0–100 days after transplantation) or chronic (> 100 days after transplantation), but it is the characteristic clinical performance that determines whether the disease is chronic or acute,rather than the time of onset. And some clinical features could appear in both acute GVHD or chronic GVHD [
3,
4].
As a systemic immune-related disease, GVHD can involve several organs [
5]. Skin GVHD usually presents with erythematous or lichen planus-like manifestations. When it involves the gastrointestinal system, diarrhoea, nausea, and anorexia are common symptoms [
6]. Lung GVHD is often characterized by bronchiolitis obliterans syndrome (BOS), which can severely impact the quality of breathing [
7]. If the liver is involved, severe, potentially fatal liver function impairment can occur [
8]. All the lesions produced by GVHD, regardless of organ involvement, could cause irreversible damage, threatening human life and health [
5]. Chronic ocular graft-versus-host disease (coGVHD) usually involves the anterior segment of the eye, such as the cornea, conjunctiva, and meibomian gland, and typically manifests as dry eye disease (DED). CoGVHD usually occurs 2 years after transplantation with a morbidity of approximately 60% [
9,
10] and potential impacts on patient quality of life. About 80% patients used tear substitutes to relieve eye disorders. About 60% patients are unable to work because of ocular and systemic GVHD [
11]. Ocular GVHD (oGVHD) is closely associated with the presence of systemic GVHD [
12]. However, the relationship between ocular disorders and liver function and other organ lesions is unclear.
At present, coGVHD is mainly diagnosed based on the presence of ocular manifestations, which usually require ophthalmologists with equipment, such as slit lamps, for professional evaluation. Patients with coGVHD at the early or slight stages may have few obvious symptoms; however, at these stages, the tear glands and ocular surface have already been damaged [
13]. As the damage progresses, the patients can feel severe ocular discomfort and visual disorders, affecting quality of life. For patients with serious ocular abnormalities, effective treatment is poor and may place a huge financial burden on their families. Moreover, little research has been conducted on the use a model to predict chronic ocular graft-versus-host disease based on systemic risk factors. Therefore, the purpose of this study is to assess the probability of coGVHD according to routine variables (such as lung function tests, clinical biochemistry, routine blood tests, coagulation tests and others). We constructed and validated a nomogram model based on risk factors and vision-related quality of life scale scores to predict the probability of coGVHD. Nomogram models have been widely applied as highly accurate tools for predicting disease prognosis [
14]. If the prognostic factor can be identified, regular ocular examination is definitely recommended at proper time. By using this model, unexpected ocular complications could be prevented after allo-HSCT. Additionally, our nomogram could offer some protective treatment for high-risk groups, even those who feel little ocular discomfort or are in an early disease stage.
Discussion
Patients may experience eye injury in the early stage of GVHD without any symptoms [
18]. At this stage, ocular symptoms didn’t deteriorate. In the absence of conscious eye symptoms, it can be difficult for patients to take the initiative to visit the ophthalmology clinic, thus ignoring the early changes associated with the disease. Additionally, haematologists may focus more on diseases that seriously affect the lives of patients, such as those of the gastrointestinal tract, liver, and lungs, and neglect to remind their patients to visit the eye clinic. At present, only symptomatic treatment is available for patients whose severe eye diseases affects their survival and quality of life, but the effect is poor and results in inconveniences for the patient. By using this nomogram, the haematologist could calculate a score based on the relevant items in the patient's regular full-body review. If the patient has a high probability of eye disease according to the nomogram, a more professional evaluation is required even if the patient does not have any eye discomfort. For these patients or those at an early disease stage, relevant treatment or prevention can be given when necessary.
However, there is currently little research on the relationship between ocular and systemic GVHD, and the risk factors for ocular GVHD are still not fully understood. Dry eye disease is common in chronic ocular GVHD and has been seen in 60 to 90% of patients with systemic GVHD [
19,
20]. Additionally, systemic GVHD was described as a risk factor for ocular GVHD [
21]. However, few studies was able to determine the association between systemic and ocular GVHD. When the number of variables is large and much larger than the sample size, and there is serious multicollinearity between variables, LASSO regression can play its maximum utility. In our study, due to the large number of systemic conditions, so we chose this method to screen systemic data. We sought to construct a predictive model to determine the possibility of ocular disease according to the patient’s systemic condition.
The items we selected were chosen as predictors based on the strength of their univariate association with outcome through LASSO regression and logistic regression [
22]. At last, the risk factors we found included lymphocytes, PTA, CD3 + CD25 + cells, CD3 + HLA-DR + cells, and the OSDI, as reflected in the comparison between groups, laying the foundation for the next step of constructing a predictive model. This nomogram was established using these seven risk factors and demonstrated a great diagnostic ability with both the training and validation sets. We hope our predictive model can be used widely for post allo-HSCT patients to predict the probability of coGVHD when they visit their haematologist.
Nomograms have been used in a variety of ocular diseases, such as for predicting glaucoma progression in patients showing disc haemorrhage based on risk factors [
23] and in the prognosis of metastatic uveal melanoma according to the patient’s systemic condition, such as the percentage of liver involvement and lactate dehydrogenase (LDH) level [
24]. Therefore, the concept of predicting a diagnosis of oGVHD according to the patient’s systemic condition through a nomogram is very promising. The results could be used to have the haematologist remind patients with early-stage ocular disease or those at high risk to visit the ophthalmology clinic for further assistance. Those in the high-risk group would particularly benefit, as such a visit may be helpful for developing a treatment or intervention plan.
We also aimed to explore systemic risk factors associated with coGVHD. The guidelines indicate that lung injury with GVHD is related to a decrease in FVC, which may indicate that the patient has bronchopneumonia obliterans, a condition known to be caused by GVHD [
13]. However, there are few studies about the relationship between ocular injury and lung function. Here, we found that the actual and predicted FVC, FEV1 and TLC/SB% were related to ocular GVHD.
Liver damage is another main clinical manifestation of systemic GVHD [
25,
26]. Studies have confirmed that changes in bilirubin and ALT are the most important manifestations of liver GVHD. Additionally, changes in the levels of bilirubin and/or ALT and GGT prior to the manifestation of the liver GVHD can be a sign of future liver GVHD in the guideline [
27]. Interestingly, our research found that the liver index most closely related to ocular GVHD was GGT; this index has also been shown to be related to inflammatory disease and reactive oxygen species (ROS) production [
28], and thus we speculate that changes in GGT may be related to liver inflammation in early liver GVHD.
In addition to those organs talked above, previous studies had indicated that ocular complications may be related with skin disease [
29,
30]. Compared with patients without coGVHD, those with coGVHD had more skin lesions. The incidence of skin GVHD was higher in two groups, and the difference was not statistically significant. Therefore, we considered the reason why we don’t screen skin GVHD as risk factors might be associated with penalties.
Nomogram is widely used not only to predict ocular diseases, but also to predict other diseases, such as “Risk analysis of pulmonary metastasis of chondrosarcoma by establishing and validating a new clinical prediction model: a clinical study based on SEER database”, “Prediction of the risk of C5 palsy after posterior laminectomy and fusion with cervical myelopathy using a support vector machine: an analysis of 184 consecutive patients” and “Development and validation of a novel predictive model and web calculator for evaluating transfusion risk after spinal fusion for spinal tuberculosis: a retrospective cohort study” [
31‐
33]. These models have good diagnostic efficiency, therefore, we believe that we should continue to promote the application of such models in ocular diseases.
This study aims to propose the use of systematic data to predict ocular GVHD to prompt patients to get a more professional assessment, but there are certain limitations. Although we used rigorous internal validation, this study lack additional validation with external data because of sample size. However, small sample is inevitable. Our present model provides and confirm a trend for this eye disease ceased by systemic disease. And a larger-scale multi-center study may be required, and the validity of the nomogram needed to be verified, too. In the next following studies, the sample size will expanded, and our model still need external validation. The index we choose is commonly used, and it was more relevant to the disease. It was also an indicator that was very suitable to be extended to other hospitals. Therefore, although the model had not been externally verified, it still had good test performance.
We believe that this method will have great diagnostic capacity and extensive application prospects. We hope to cooperate with other research groups and enroll more patients to refine and validate this model in the future. More dynamic observations of the ocular and systemic conditions will be given.Thus, our model will show enlarged applicability domain.
In summary, this proposes the use of systematic data to predict ocular GVHD so that the patient can be referred to an ophthalmologist to obtain a more professional assessment. We established a statistical model and demonstrated that it effectively predicts the progression of ocular GVHD caused by systemic disease.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.