Ocular graft‐versus‐host disease: Risk factors of ocular graft‐versus‐host disease after allogeneic haematopoietic stem cell transplantation in Denmark

ENGLISH SUMMARY

Allogeneic haematopoietic stem cell transplantation (HSCT) is used to cure both malignant and non‐malignant haematological diseases. Despite HSCT has been available for more than 50 years, chronic graft‐versus‐host disease (cGVHD) remains a difficult immunologically mediated challenge, which increases morbidity and mortality after transplantation. When cGVHD targets the eyes, it causes reduced tears and inflammation which lead to red, irritated eyes, corneal damage and blindness in worst cases. Ocular cGVHD significantly reduces quality of life after HSCT. We need to gain further knowledge about this disease to help this patient group.

The overall aim of this PhD project was to investigate the incidence and risk factors for developing ocular cGVHD in both adults and children. Furthermore, the objective was to investigate possible associations between ocular cGVHD and cGVHD in other organs, and mortality after HSCT.

A conditioning regimen is given to the patient before transplantation, which can be either myeloablative (MA) or non‐myeloablative (NMA). Our studies showed that in adults, the 5‐year cumulative incidence of ocular cGVHD was 18% after MA and 35% after NMA regimen. Several factors were associated with a higher risk of ocular cGVHD after both conditioning regimens. In the MA group, malignant disease, Schirmer's test ≤10 mm/5 min before HSCT, the use of a matched unrelated donor or female donor, peripheral blood as stem cell source and acute GVHD (grade III–IV) increased the risk of ocular cGVHD. In the NMA group, Schirmer's test ≤10 mm/5 min before transplantation and higher recipient age increased the risk of ocular cGVHD. In children, the incidence of ocular cGVHD was 6% and therefore less common than in adults. Ocular cGVHD was more frequent in patients with extensive cGVHD, and when other ectodermal derived organs were involved (skin, mouth, genitals and nails). The frequency of ocular cGVHD was especially high in patients with skin sclerosis as a manifestation of cGVHD (70%). Our studies suggest that target antigens in ectodermal derived organs might be involved in the complex pathophysiology of ocular cGVHD, but more studies are needed to explore this. Ocular cGVHD was furthermore found to be associated with a higher non‐relapse mortality.

In conclusion, several risk factors for developing ocular cGVHD exists. This knowledge may be applied to guide clinical trials (i.e. power calculations), to inform patients of their risk of developing ocular cGVHD and to guide clinicians in scheduling patient follow‐up. Because of the many patients with signs of dry eyes before HSCT (which increase the risk of ocular cGVHD), we recommend performing a baseline ophthalmological examination before HSCT.

More studies are needed to elucidate the pathophysiology of ocular GVHD. In the future, this could lead to better treatment options and potentially prevention of the disease.

Abbreviations

aGVHD

acute graft‐versus‐host disease

BM

bone marrow

cGVHD

chronic graft‐versus‐host disease

DED

dry eye disease

HSCT

haematopoietic stem cell transplantation

ICCGVHD

International Consensus Criteria for Ocular cGVHD

MA

myeloablative

MRD

matched related donor

MUD

matched unrelated donor

NIH

National Institutes of Health

NMA

non‐myeloablative

oGVHD

ocular graft‐versus‐host disease

OSDI

Ocular Surface Disease Index

PB

peripheral blood

TBI

total body irradiation

UCB

umbilical cord blood

1. INTRODUCTION AND BACKGROUND

1.1. General introduction

Ocular graft‐versus‐host disease (oGVHD) is a complication to allogeneic haematopoietic stem cell transplantation (HSCT), which may reduce quality of life significantly after being treated for a life‐threatening disease (Ogawa, Okamoto et al., 2003). HSCT is used as a potentially curative therapy of both malignant and non‐malignant haematological diseases (i.e. leukaemia, myelodysplastic syndrome, aplastic anaemia and immunodeficiencies) (Anderson & Regillo, 2004; Kim, 2006; Ogawa, Okamoto et al., 2003). Although the outcome of HSCT has shown improvement over time, the treatment is still associated with increased morbidity, mortality and late health issues (Hamilton, 2021). Graft‐versus‐host disease (GVHD) is one of the greatest contributors to these challenges (Hamilton, 2021). The pathogenesis of GVHD is complex. Simplified, GVHD develops if the transplanted donor immune system recognizes healthy recipient cells as foreign, which lead to inflammation and tissue destruction (Anderson & Regillo, 2004; Hamilton, 2021; Kim, 2006; Ogawa, Okamoto et al., 2003). GVHD can affect many organs, including the eyes (Jagasia et al., 2015). Quality of life is often significantly reduced in patients with oGVHD because of severe ocular discomfort (Dietrich‐Ntoukas et al., 2012; Inamoto et al., 2019; Riemens et al., 2014). The patients with oGVHD need to apply lubricating eye drops up to numerous times per hour, which restricts common daily tasks such as reading, working in front of a computer, being outdoors and in worst cases, oGVHD may cause blindness (Dietrich‐Ntoukas et al., 2012). Furthermore, oGVHD is associated with poor sleep quality (Liao et al., 2024). During the last two decades, protocols with reduced intensity of HSCT have been introduced, which makes it possible for patients with comorbidities and elderly patients to receive the curative treatment (Bacigalupo et al., 2009). Albeit these protocols have been successful in limiting toxic side effects of conditioning, the incidence of GVHD is equivalent and thus still a major clinical problem (Malard & Mohty, 2023). Expansion of treatment indications causes increasing numbers of transplant recipients, which are at risk of developing ocular GVHD. This underscores the need for investigation of the disease. Despite HSCT has been around for over half a century, risk factors for developing oGVHD still need to be clarified. The studies in this thesis sought to map the risk factors of oGVHD in both adults and children. All the paediatric allogeneic HSCTs were performed nationwide at Rigshospitalet in Denmark in the entire follow‐up period. In adults, the transplantations were nationwide until 2009. In cases of cGVHD or other transplantation‐related complications in our cohort after completion of regular follow‐up, the patients were referred to Rigshospitalet for evaluation, which made it possible to have a unique large dataset to study risk factors for developing oGVHD and outcomes.

1.2. Haematopoietic stem cell transplantation—Procedure and types

Haematopoietic stem cell transplantations are performed with the intention to substitute a patient's malfunctioning haematopoietic system with healthy cells (Sureda et al., 2015). A stem cell transplantation can be either autologous with the use of own cells or allogeneic with the use of cells from a donor (Figure 1).

FIGURE 1.

Types of HSCT. *Syngeneic HSCT is not considered as classic allogeneic HSCT because the donor is HLA‐identical (an identical twin) and therefore do not cause issues with immunoincompatability (GVHD). Syngeneic HSCT is used as an alternative to autologous HSCT. BM, bone marrow; HSCT, haematopoietic stem cell transplantation; PB, peripheral blood; UCB, umbilical cord blood.

The treatment effect of autologous HSCT is achieved through the conditioning treatment with toxic therapies, which is given before each transplantation. The infusion of haematopoietic stem cells prevents prolonged myelosuppression which makes it possible to give high doses of the toxic treatment (consisting of chemotherapy with or without total body irradiation [TBI]) (Balassa et al., 2019).

In allogeneic HSCT, the donor immune cells contribute with the eradication of primary disease through the so‐called graft‐versus‐leukaemia effect (GVL) (Horowitz et al., 1990). Unfortunately, this mechanism can also lead to the unwanted GVHD, where the donor immune cells recognize the patient's own cells as foreign and cause an immunological response/attack towards the patient's healthy tissues (Horowitz et al., 1990). This thesis deals with patients, who have undergone allogeneic transplantation, and therefore are at risk of developing GVHD.

When a patient is eligible for an allogeneic HSCT, a donor needs to be selected. Human leukocyte antigen (HLA) matching is the most important factor, when choosing a donor and is the strongest determinant of the outcome after allogeneic HSCT (Malard et al., 2023). A donor can be either related (i.e. a sibling) or unrelated from a registry. The source of the donor allograft can be either from bone marrow (BM), peripheral blood (PB) or umbilical cord blood (UCB), and is determined based on the characteristics of the patient, the disease and the donor.

The procedure of allogeneic HSCT is as follows. When a patient has an indication for allogeneic HSCT, the patient is referred to a transplantation center, where it is decided whether a transplantation is the right choice based on a risk/benefit analysis. The conditioning regimen, source of allograft is then determined based on the patient's characteristics (i.e. age, sex and comorbidities), the disease and available donor. A transplant coordinator plans all the logistics regarding the transplantation. A comprehensive pre‐evaluation of the patient is important to reduce the risk of complications (Carreras & Rambaldi, 2019). This evaluation includes examinations by the ophthalmologist, gynaecologist and dentist, to optimize the patient's health prior to transplantation.

Before the day of stem cell infusion, a conditioning regimen consisting of chemotherapy, with or without TBI, is given. The purpose of the conditioning is to eliminate the patient's immune system, in order to avoid rejection of the graft and destroy malignant cells. The conditioning can be either myeloablative (MA) or non‐myeloablative (NMA). MA conditioning, which consists of high dose chemotherapy (with or without TBI), cause irreversible cytopenia where stem cell support is mandatory. MA ensure a high disease control and has a lower relapse rate, but not all patients can tolerate the higher toxicity (Bacigalupo et al., 2009). NMA has the lowest toxicity and only causes mild cytopenia but is sufficient enough to make successful engraftment (Bacigalupo et al., 2009). The original NMA conditioning consists of low dose chemotherapy and low dose TBI (the so‐called Seattle protocol) (Nagler & Shimoni, Nagler & Shimoni, 2019), which is the one used in Denmark. Engraftment is when the haematopoietic cells start to evolve and make healthy new blood cells. NMA has made it possible for older patients, and patients with comorbidities to receive allogeneic HSCT, and are now used for both malignant and non‐malignant indications (Slavin et al., 1998).

In Denmark, the intensity of conditioning used before HSCT was either MA or NMA in the study period. Even though immunosuppressants are given as GVHD prophylaxis (methotrexate, cyclosporine, mycophenolate, tacrolimus and anti‐thymocyte globulin are used in DK), GVHD is a major contributor to mortality and morbidity on short‐term and long‐term basis as mentioned previously.

1.3. Systemic GVHD

Graft‐versus‐host disease can be either acute or chronic. Acute GVHD (aGVHD) usually occurs within the first 100 days after transplantation but may occur later (Jacobsohn & Vogelsang, 2007). Previously, the definitions of aGVHD and chronic GVHD (cGVHD) was only based on timing (Ferrara et al., 2009). This definition was not satisfactory which is why they are now based on clinical manifestations instead of timing according to National Institutes of Health (NHH) (Jagasia et al., 2015). Whereas aGVHD is mediated by donor T cells in three acknowledged phases (Ferrara et al., 2009), the pathophysiology of cGVHD is more complex and still poorly understood (Cooke et al., 2017). The pathophysiology of cGVHD involves dysregulated B‐cells and has many similarities with autoimmune disease (Cooke et al., 2017).

1.3.1. Acute GVHD—Clinical manifestations, diagnosis and pathophysiology

Acute GVHD occurs primarily in the skin, gastrointestinal tract and liver. The skin is the most commonly affected organ with a characteristic pruritic maculopapular rash, and it is usually the first organ involved, often simultaneously with engraftment of donor cells (Ferrara et al., 2009). Acute GVHD is a clinical diagnosis, that may be supported by appropriate biopsies (Jacobsohn & Vogelsang, 2007). In our studies, aGVHD was diagnosed and graded (I–IV) according to the modified Glucksberg criteria, where each of the three major targets (skin, liver and gut) is graded according to the degree of skin involvement, bilirubin levels and amount of diarrhoea with or without ileus (Przepiorka et al., 1995). Classical aGVHD occurs within the first 100 days of transplantation, but it can also occur after day 100 (it can be persistent, recurrent or with late onset) (Malard et al., 2023).

The incidence of aGVHD ranges from 35% to 45%, when a fully matched sibling donor is used, to 60–80% in recipients where an unrelated donor with one antigen HLA mismatch is used (Ferrara et al., 2009).

The pathophysiology of aGVHD is mediated by activation of donor T cells and the release of pro‐inflammatory cytokines (Cooke et al., 2017). Acute GVHD develops in three phases (Figure 2): (1) initiation phase (conditioning‐mediated tissue damage leads to activation of recipients antigen presenting cells (APCs); (2) donor T‐cell activation; and (3) effector phase (target tissue destruction) (Ferrara et al., 2009; Malard et al., 2023).

FIGURE 2.

Pathophysiology of acute graft‐versus‐host disease (GVHD). The main organs affected in aGVHD are the skin, liver and gastrointestinal (GI) tract. (1) In the initiation phase, the conditioning regimen causes damage to the recipient tissue which leads to the release of pro‐inflammatory cytokines (TNF‐α, Il‐1, Il‐6 and lipopolysaccharide [LPS] among others). This signal leads to activation of recipient antigen presenting cells (APCs). Loss of microbiota in the gut contributes to loss of the normal barrier function of the epithelium and further inflammatory stimuli. (2) In the donor T‐cell activation phase, the recipient APCs activate alloreactive T cells (the T cells proliferate and differentiate). (3) In the effector phase, cytotoxic T‐lymphocytes (CTL), natural killer cells (NK) and pro‐inflammatory cytokines (including TNF‐α and IL‐1) damage the recipient's cells, leading to apoptosis and necrosis. The T cells produce even more pro‐inflammatory cytokines in cooperation with the APCs and other innate immune cells, leading to a vicious cycle of immune activation and inflammation (Ferrara et al., 2009; Malard et al., 2023). The development and severity of aGVHD is modulated by several types of immune cells that produce anti‐inflammatory cytokines, such as regulatory T cells (Tregs) (Lin et al., 2021; Malard et al., 2023). Illustration redrawn and modified from Malard et al. (2023)) (DOI: 10.1038/s41572‐023‐00438‐1). Permission from Springer Nature, CCC RightsLink licence no.: 5852431336244.

1.3.2. Chronic GVHD—Clinical manifestations, diagnosis and pathophysiology

Chronic graft‐versus‐host disease (cGVHD) is a huge and difficult challenge that needs to be overcome to improve non‐relapse mortality and morbidity after HSCT (Cooke et al., 2017).

Approximately 30–70% of patients receiving HSCT develop systemic cGVHD, where characteristic features may include chronic inflammatory changes with progression to clinically significant fibrosis (Cooke et al., 2017; Jagasia et al., 2015). Chronic GVHD may be limited to one organ or be systemic with multiple organ involvement, and primarily affects the skin, mouth, liver and eyes (Jagasia et al., 2015; Lee & Flowers, 2008; Riemens et al., 2010; Figure 3). Time of onset is most often within the first year after transplantation but may present several years after HSCT (Jagasia et al., 2015).

FIGURE 3.

Examples of chronic graft‐versus‐host disease (GVHD) in (1) skin, (2) eye and (3) mouth. (1a, b) Hand with ichthyosis in a patient with chronic GVHD (palmar and dorsal view). (1c) Both hands with sclerosis of the skin caused by chronic GVHD. (2a) Chronic ocular GVHD with conjunctival inflammation and hyperaemia. (3a–c) Chronic GVHD with lichen planus‐like changes of the oral mucosa.

National Institutes of Health (NIH) Consensus has made a set of criteria for diagnosing and scoring the severity of cGVHD (Jagasia et al., 2015). NIH recommends that diagnosis is be established from clinical features, whether they include diagnostic signs (enough to establish the diagnosis) or distinctive signs (not normally found in aGVHD, but can establish the diagnosis if another test, evaluation by a specialist or radiographic imaging shows cGVHD in the same or another organ) (Jagasia et al., 2015).

The pathophysiology of cGVHD is still enigmatic, and not as well described as aGVHD (Cooke et al., 2017). The pathways resembles autoimmune disease, and has numerous interactions between dysregulated T‐ and B cells as well as cells from the innate immune system (macrophages, dendritic cells and neutrophils) resulting in fibrosis (Cooke et al., 2017).

1.4. Ocular GVHD

1.4.1. Anatomy, normal function and embryology of the ocular surface and lacrimal system

The ocular surface consists of the cornea, conjunctiva with goblet cells, eyelids, lacrimal gland and meibomian glands, which all serve as a unique function for the maintenance of normal vision and ocular comfort (Kao et al., 2008). A tear film spreads over the corneal surface each time we blink, like a protecting coating, keeping the corneal epithelium intact and clear, providing nutrients to the cornea and washing away foreign particles. It serves as the first line of defence against both infection and physical or chemical damage (Kao et al., 2008). The tear film consists of three layers: the outer lipid layer (produced mainly by the meibomian glands and to some extent the glands of Zeiss), the middle aqueous layer (produced mainly by the lacrimal gland, and to some extend the accessory lacrimal glands of Krause and Wolfgang) and the inner mucin layer (produced by the conjunctival goblet cells and conjunctival epithelial cells) (Figure 4). The main purposes of the three layers are as follows: the outer lipid layer seal the tear film to reduce evaporation; the middle aqueous layer lubricates the eye, washes away particles and prevents infections with bactericidal lysozymes and other proteins; and the inner mucin layer helps the aqueous layer to spread evenly over the surface and provide nourishment to the corneal epithelium (Heegaard et al., 2022; Kao et al., 2008). To function normally, this sensitive ecosystem must be in balance. When something disrupts this homeostasis, like low tear volume, altered proteins/electrolytes/cytokines in the tears or reduced production of mucin/lipids, dry eye disease (DED) occurs (Craig et al., 2017).

FIGURE 4.

Structures involved in tear production. Credit: National Eye Institute, National Institutes of health. Permission to use image with credit from NIH Image Gallery: https://www.flickr.com/people/nihgov/. Figure No.: https://www.flickr.com/photos/nihgov/28759950030.

Numerous tightly regulated cellular events contribute to the embryonic development of the tissues involved in tear production (including cell migration, proliferation and differentiation) (Kao et al., 2008). Most of the tissues constituting the ocular surface and lacrimal system share the same embryological origin; the epithelium of the cornea and conjunctiva, the meibomian glands and the lacrimal gland are all derived from the ectoderm (Kao et al., 2008).

1.4.2. Ocular GVHD—Pathophysiology and histopathology

National Institutes of Health have defined histopathological diagnostic criteria for both aGVHD and cGVHD in the major organ systems (Shulman et al., 2006, 2015). Histopathological findings of ocular GVHD have been described in the lacrimal gland and conjunctiva, but there are no histopathological diagnostic features of ocular GVHD. In the skin, NIH have described minimal histological diagnostic cGVHD criteria where epidermal apoptosis located in basal cell layer or lower stratum spinosum is mandatory (Shulman et al., 2015). Typical features of both aGVHD and cGVHD are superficial interface dermatitis (involving the dermo‐epidermal junction) with formation of vacuoles in the basal cell layer or lymphoid inflammation forming a lichenoid pattern, and furthermore, necrotic keratocytes might be surrounded by lymphocytes (lymphocyte satellitosis) (Shulman et al., 2015). In the conjunctiva, some histological features are similar to the findings in the skin (lymphocyte satellitosis, vacuoles in the basal cell layer and epithelial cell necrosis), and other non‐specific features includes epithelial attenuation and depletion of goblet cells which are not pathognomonic for ocular GVHD (Shulman et al., 2015). Histological findings of pseudomembranes may be found in patient with acute oGVHD (Shulman et al., 2015). In the lacrimal gland, the histological changes are similar to the changes in the minor salivary glands where clusters of mononuclear cells are found around the ducts, and obliterated acinar lobules are replaced by fibrosis (Ogawa, Kuwana et al., 2003; Shulman et al., 2015). Ogawa et al. have examined lacrimal gland biopsies from patients with cGVHD. They found that fibroblasts located in the lacrimal gland had chimerism in patients with cGVHD, indicating that donor‐derived fibroblasts may contribute to the formation of excessive fibrosis in these patients (Ogawa et al., 2005). They also found that CD4+ and CD8+ T cells were detected mainly in the periductal areas, and the amount of CD8+ T cells in the glandular epithelia were higher in patients with cGVHD than in those with Sjögren's syndrome (Ogawa, Kuwana et al., 2003). Histology of ocular GVHD is not used in the clinical assessment of the disease yet.

1.4.3. Acute ocular GVHD—Clinical manifestations and diagnosis

Ocular complications may be caused by both acute and chronic GVHD, although they tend to be less often and less severe in patients with acute GVHD than in the chronic form (Anderson & Regillo, 2004).

The manifestations of acute GVHD in the eye include conjunctival inflammation, pseudomembranes and corneal epithelial sloughing (Jabs et al., 1989). There are no clinical diagnostic criteria of acute oGVHD, but Jabs et al. proposed the following grading system in 1989 with four stages: (1) conjunctival hyperaemia, (2) conjunctival hyperaemia with chemotic response or serosanguineous exudate, (3) pseudomembranous conjunctivitis and (4) pseudomembranous conjunctivitis with corneal epithelial slough (Jabs et al., 1989). Further investigations are needed to fully describe acute oGVHD (Kantor et al., 2024).

1.4.4. Chronic ocular GVHD—Clinical manifestations and diagnosis

Keratoconjunctivitis sicca (KCS), with both reduced tear quality and production, is the most frequent ocular complication in patients with cGVHD (Anderson & Regillo, 2004). Meibomian gland dysfunction is the second most frequent complication of ocular cGVHD (Cheng et al., 2023). In the slit lamp, punctate fluorescein staining is seen showing corneal epithelial disruption (Figure 5). The patients have symptoms of dry eyes with irritation, itchiness, photophobia, grittiness, burning, foreign body sensation, excessive tearing, blurred vision and pain (Dietrich‐Ntoukas et al., 2012). In secondary corneal infections, ulceration may occur, and perforation of the cornea in the worst cases, resulting in blindness (Figure 6). Cicatricial conjunctivitis may cause ectropion/entropion or trichiasis, which can cause additional unpleasant symptoms (Dietrich‐Ntoukas et al., 2012; Riemens et al., 2010). The ocular symptoms of GVHD may cause significantly reduced quality of life in this patient group (Dietrich‐Ntoukas et al., 2012).

FIGURE 5.

Corneal punctate fluorescein staining in a patient with ocular chronic graft‐versus‐host disease (cGVHD).

FIGURE 6.

Advanced ocular chronic graft‐versus‐host disease (cGVHD) with complications. (a) perforation of the cornea, pannus (corneal neovascularization), conjunctival hyperaemia and scarring of the eyelids. (b) Same eye after corneal graft complicated with a central ulcer.

Currently, there are two widely recognized international diagnostic criteria for ocular cGVHD (Cheng et al., 2023). The NIH Consensus diagnostic criteria: ocular cGVHD can be diagnosed in patients with low Schirmer's test (mean value of ≤5 mm at 5 min), or a new onset KCS determined by a slit lamp examination (with a mean Schirmer's test value of 6–10 mm) (Jagasia et al., 2015). If possible, the patients ought to have confirmed normal values at baseline and other causes of KCS should be excluded. These criteria apply to clinical trials investigating ocular GVHD (in general cGVHD studies, an additional distinctive feature of cGVHD is required to establish the diagnosis of ocular cGVHD) (Jagasia et al., 2015) In 2013, the International Consensus Criteria for Ocular cGVHD (ICCGVHD) was proposed. These criteria are more comprehensive, and the diagnosis is based on both subjective symptoms (the ocular surface disease index [OSDI]), objective ophthalmological evaluation (Schirmer's test, the degree of corneal fluorescein staining and conjunctival injection), and if non‐ocular cGVHD is present (Ogawa et al., 2013).

1.4.5. Ocular GVHD—Treatment

Artificial tears are the first choice of treatment in all patients with DED symptoms because of lacrimal gland hypofunction. Artificial tears also dilute inflammatory cytokines in tear fluid and may in this way maintain epithelial integrity (Kantor et al., 2024). Another way to optimize the tear volume in the eye for a longer time is to use punctal occlusion (Cheng et al., 2023).

To treat the acute inflammation or cicatricial conjunctivitis, corticosteroid drops are frequently used, but because of the potential development of cataract and glaucoma long‐term use is not recommended. Immune modulating eye drops, like cyclosporine and tacrolimus, are more suitable for chronic treatment in patients with oGVHD (Kantor et al., 2024).

To target MGD caused by oGVHD, antibiotics are also frequently used when warm compresses and meibomian hygiene is not enough. Serum eye drops have been shown to relieve symptoms and improve corneal surface health in patient with oGVHD. Scleral contact lenses and soft bandage contact lenses are used to manage pain (i.e. when the patients have developed filamentary keratitis or recurrent trichiasis) and help corneal healing (Kantor et al., 2024). Comprehensive treatment, like tarsorrhaphy can be necessary, and amniotic membrane graft or corneal graft can be used, when the cornea is perforated (Figure 6; Cheng et al., 2023; Kantor et al., 2024).

1.5. Historical review of stem cell transplantation and GVHD

Over 50 years, HSCT has evolved from a procedure with unmanageable complications to a standard treatment of many haematological diseases. It all started with the question of how to cure irradiation damage in survivors after the atomic bomb explosions in Japan 1945 (Carreras & Rambaldi, 2019; Little & Storb, 2002). In 1949, Jacobson et al. (1949) discovered that mice were protected from irradiation in lethal doses, when their spleens were shielded by lead. Two years later, Lorenz et al. discovered that protection against irradiation could also be established by intravenous infusions of BM (Lorenz et al., 1951).

E.D. Thomas and his team from Seattle pioneered the transplantation of BM in the late 1950s and 1960s although the attempts of BM transplantations in this period failed (Little & Storb, 2002). In 1990, E.D. Thomas was awarded the Nobel prize for his contribution to stem cell transplantation research (NobelPrize.org Accessed 13 May 2024).

In Denmark, the first allogeneic transplantation was performed in 1971 with nine transplants completing this decade. In the 1980s, almost 200 transplants were performed, and approximately 400 in the 1990s, where PB was also introduced as a stem cell source.

NMA was introduced as conditioning in 2000, making it possible to treat elderly, and patients with co‐morbidity. This increased the number of eligible patients. In the 2000's the number of transplantations increased to 100 per year, where today approximately 120 transplantations are performed each year.

Now, over 35.000 allogeneic HSCTs are performed globally each year (Niederwieser et al., 2022).

2. HYPOTHESIS AND AIMS OF THE STUDY

The following hypotheses were formulated and formed the basis of this thesis:

1. Ocular chronic GVHD is less common in children than in adults.

2. Certain factors increase the risk for developing ocular chronic GVHD, and they are similar in children and adults.

3. Ocular chronic GVHD occurs at the same time after HSCT in children and adults.

4. The patients, who develop ocular cGVHD have a lower relapse rate and better survival, because of the ‘graft‐versus‐tumour effect’.

2.1. Hypotheses of possible risk factors in details

2.1.1. Conditioning regimen

We hypothesize that ocular chronic GVHD is more common after MA conditioning (high dose regimen) because of more tissue destruction at higher chemotherapy doses, and thus, the release of more cytokines may create a greater immune response.

2.1.2. Stem cell source

T cells have an important role in the pathogenesis of the development of GVHD. As grafts obtained from PB contain a 10‐fold higher number of T cells compared with BM graft, we believe that the patients who receive PB develop oGVHD more often than when BM is used as stem cells.

2.1.3. Recipient/donor age

We hypothesize that recipient age is associated with higher risk of oGVHD as DED increases with age, and higher donor age is also associated with increased risk of oGVHD.

2.1.4. Donor/recipient relation

In patients receiving graft from an unrelated doner, there is a greater mismatch and thus an expected greater likelihood of oGVHD.

2.1.5. Donor sex/recipient sex

We expect a higher risk in women, as they generally have a greater tendency to develop DED and autoimmune diseases. Furthermore, we expect more oGVHD if there is a female donor to a male recipient.

2.1.6. Primary disease

The risk of oGVHD is greater in the patients with malignant primary disease than in patients with non‐malignant disease as they have been exposed to stronger chemotherapy/irradiation and thus have more tissue damage that makes the tissue more susceptible to GVHD.

2.1.7. TBI

The risk of oGVHD is greater in the patients who receive higher doses of irradiation because of greater tissue damage.

2.1.8. Schirmer's test

Patients with low S1t before transplantation have a greater risk of developing ocular GVHD.

2.1.9. Organs involved in cGVHD

Ocular cGVHD is associated with cGVHD in other organs, and especially in extensive cGVHD compared with limited disease.

2.2. Aims

This PhD study has three main aims:

1. The first aim was to report the incidence of ocular cGVHD among different conditioning regimens before HSCT.

2. The second aim was to identify risk factors for developing ocular cGVHD in both adults and children.

3. The third aim was to investigate the association between ocular cGVHD and cGVHD in other organs and outcomes after HSCT (relapse and non‐relapse mortality).

2.3. List of papers

The thesis is based on the following papers:

• I Chronic Ocular Graft‐Versus‐Host Disease After Allogeneic Haematopoietic Stem Cell Transplantation in Denmark – Factors Associated with Risks and Rates in Adults According to Conditioning Regimen.

Jeppesen H, Sengeløv H, Eriksson F, Kiilgaard JF, Andersen ST, Lindegaard J, Julian HO, Heegaard S; Bone Marrow Transplant. 2021; 56(1):144–54 (DOI: 10.1038/s41409‐020‐0993‐3

• II Ocular Graft‐Versus‐Host Disease and Dry Eye Disease After Paediatric Haematopoietic Stem Cell Transplantation – Incidence and risk factors.

Jeppesen H, Kielsen K, Siersma V, Lindegaard J, Julian HO, Heegaard S, Sengeløv H, Müller K; Bone Marrow Transplant. 2022; 57(3):487–98 (DOI: 10.1038/s41409‐022‐01564‐2

• III Ocular Chronic Graft‐Versus‐Host Disease and Its Relation to Other Organ Manifestations and Outcomes After Allogeneic Hematopoietic Cell Transplantation.

Jeppesen H, Gjærde LK, Lindegaard J, Julian HO, Heegaard S, Sengeløv H; Transplant Cell Therapy 2022 Aug21 (DOI: 10.1016/j.jtct.2022.08.016).

In this thesis, the publications will be referred to by the Roman numerals.

3. MATERIALS AND METHODS

3.1. Material

3.1.1. Patients and clinical data

The studies in this thesis are based on all patients who received allogeneic HSCT from 1980 to 2016 at Copenhagen University Hospital (Rigshospitalet). Data were collected with a follow‐up until June 2019 from haematological and ophthalmological medical records (I, II and III).

During the study period, the patients had an ophthalmological evaluation planned before transplantation and furthermore each year after HSCT for the first 5 years. In addition to the planned visits, the patients were referred to an ophthalmological evaluation at the hospital if needed. Allogeneic HSCTs were conducted nationwide at Rigshospitalet in children (below 16 years) during the whole follow‐up period, whereas in adults, allogeneic HSCTs were performed nationwide at the hospital until 2009.

A standard ophthalmological evaluation involved visual acuity, tear film break‐up‐time (BUT), Schirmer's test without anaesthesia (S1t), corneal fluorescein staining (CFS), a slit lamp examination of the eyelids, cornea, anterior chamber and funduscopic examination of the posterior segment of the eye. Other clinical data were recorded; use of eye drops, severity of DED symptoms, NIH eye score, age and sex of both recipient and donor, primary disease, relation between donor and recipient, donor–recipient sex combination, stem cell source, conditioning, use of TBI, occurrence of systemic acute GVHD, chronic GVHD, involved organs and date of death.

The PhD study was conducted in accordance with the Declaration of Helsinki. Approvals were granted from the Danish Data Protection Agency (Jr. No.: VD‐2019‐33) and the Danish Health and Medicines Authority (Sundhedsstyrelsen, Jr. No.: 3‐3013‐595/1).

3.1.2. Discussion of the material

The retrospective nature of the studies leads to several limitations, we would not see in a prospective study. The quality of chart registrations may have changed over time. The studies have a time span of almost 40 years, the charts have transitioned from medical records written on paper to completely paperless medical records, and a huge evolution has happened in the era of HSCT leading to better survival of the transplantation patients. Despite challenges in interpreting results of the studies due to evolving knowledge, treatment options and diagnostic procedures over time, we chose to include these patients over the large timespan, to get the full picture of the transplant patients, taking the limitations of this decision into account when interpreting the results.

Exclusion of patients due to missing ophthalmological examinations before transplantation only happened in 32 cases of the 1452 adults and 38 of the 484 children, which indicate that the quality of the material is good, although there are missing values within the examinations as described in the papers.

The strengths of this large material are that it is derived from a single institution with the same follow‐up procedures during the years and a long follow‐up period. These studies are some of the most comprehensive studies of ocular GVHD in both adults and children.

3.2. Methods

3.2.1. Inclusion criteria

All patients who underwent true allogeneic HSCT (no syngeneic HSCTs) at Rigshospitalet who had a baseline ophthalmic evaluation before transplantation were included in the study.

Patients with DED before HSCT were excluded from the risk analyses. In both adults and children, the diagnosis of oGVHD was based on the criteria proposed by the ICCGVHD (Ogawa et al., 2013). These criteria are based on both subjective and objective measurements (S1t, CFS, OSDI, conjunctival injection and whether the patient has systemic cGVHD). Because of the retrospective nature of our studies, it was not feasible to acquire an OSDI score (a questionnaire used to describe the severity of symptoms (Schiffman et al., 2000). The information from the medical charts made it possible to obtain NIH eye score to assess symptoms (1: none, 2: mild, 3: moderate and 4: severe). It was necessary to use a modified scale where NIH eye score was used instead of OSDI to graduate the symptoms in the four categories (Table 1) (Jeppesen et al., 2021).

TABLE 1.

Chronic ocular GVHD diagnostic score^* (I).

Score (points)	Schirmer's test (mm)	CFS (points)	NIH eye score (points)**	Conjunctival injection (points)	Systemic GVHD
0	>15	0	0	None	No
1	11–15	<2	1	Mild/moderate
2	6–10	2–3	2	Severe	Yes
3	≤5	≥4	3	(Max 2 points)	(0 or 2 points)

^*Modified diagnostic criteria for chronic GVHD proposed by the International Chronic Ocular GVHD Consensus Group (Ogawa et al., 2013): The total of the diagnostic scores (‘Schirmer test score’ + ‘CFS score’ + ‘NIH Eye score’ + ‘Conjunctival injection score’ + ‘Systemic GVHD score’): Score <6: ‘None’, score 6–7: ‘probable’ and score ≥8: ‘definite’ ocular GVHD. As modification from the original score, NIH eye score was used instead of OSDI score in our studies. CFS, corneal fluorescein staining; GVHD, graft‐versus‐host disease; NIH, National Institutes of Health; OSDI, Ocular Surface Disease Index.

^**NIH eye score: (Jagasia et al., 2015). (0) No symptoms. (1) Mild dry eye symptoms not affecting activities of daily living (ADL) (requiring eye drops ≤3 times per day) OR asymptomatic signs of keratoconjunctivitis sicca (KCS). (2) Moderate dry eye symptoms partially affecting ADL (requiring eye drops >3 times per day or punctal plugs) WITHOUT vision impairment. (3) Severe dry eye symptoms significantly affecting ADL (special eyewear to relieve pain) OR unable to work because of ocular symptoms OR loss of vision caused by KCS.

According to the ICCGVHD criteria, a score of 6–7 implies that the patient ‘probably’ has oGVHD, and a score of 8 or above implies that the patient ‘definitely’ has oGVHD. Rapoport et al. sought to validate the ICCGVHD criteria where they found that the ICCGVHD diagnosis of ‘probable’ and ‘definite’ oGVHD combined was in better agreement with the oGVHD diagnosis determined by best clinical practice (Rapoport et al., 2017). To obtain an oGVHD diagnosis in our studies, the ICCGHVD score of the most affected eye had to be 6 or higher.

Reproduced with permission from Springer Nature (DOI: Jeppesen et al., 2021).

3.2.2. Discussion of methods

We used the diagnostic criteria proposed by the International Chronic Ocular GVHD Consensus Group criteria (ICCGVHD), because they are the most validated criteria to date, and they can be used reproducibly (Rapoport et al., 2017). The use of NIH eye score instead of OSDI score may have led to bias. We chose this option because it was not possible to obtain an OSDI questionnaire retrospectively. Alternatively, we could have chosen the NIH criteria for diagnosing ocular cGVHD (Jagasia et al., 2015). We decided to use the ICCGVHD criteria, not only because they were the most validated, but also because they included several more important variables (other than Schirmer's test) to make the diagnosis of ocular cGVHD.

In children, S1t (which is one of the main diagnostic values) was only performed sporadically in our cohort depending on patient symptoms, age and cooperation. It was not possible to use multiple imputation like we did in the diagnosis of adults to deal with missing values, because the proportion of missing S1ts was too large. We chose to acknowledge this limitation and still use the ICCGVHD criteria knowing it probably led to an underestimation of ocular cGVHD or a delay in diagnosis (Jeppesen, Gjærde et al., 2022).

One of the great strengths of the studies were the ophthalmological evaluation before transplantation making it possible to assess the condition of the patient's eyes and exclude patients with DED before transplantation.

3.2.3. Statistical analyses

The cumulative incidences of events were analysed in competing risk analyses with death and secondary graft infusions (transplant or donor lymphocyte infusion) as competing events (I and II). Risk factors were investigated using hazard ratios in both unadjusted and adjusted analyses.

In Study I, it was possible to handle missing covariate data by multiple imputation.

In Study III, chi‐squared test (or Fisher's exact test if the expected number was 5 or less) was applied to compare the relative frequencies of ocular cGVHD stratified by non‐ocular cGVHD in other organs. In the analyses of outcomes (III), a cause‐specific Cox proportional hazards model was performed to determine the overall mortality, the non‐relapse mortality (with relapse as competing event) and relapse (with non‐relapse mortality as competing event). In these analyses, ocular cGVHD was treated as a time‐dependent predictor (III).

3.2.4. Discussion of statistical methods

After HSCT, time to event is an important factor which may induce bias in statistical analyses (Klein et al., 2001). Survival analysis deals with censored follow‐up data (the patients in our study have different lengths of follow‐up time). Competing risk analysis is a form of survival analysis you can use in the presence of competing events (Andersen et al., 2012). In our studies, competing events of developing ocular GVHD, were death and secondary graft infusions. The patient group we are observing have a high mortality rate. By using death as a competing event, we reduce the risk of a variable being observed as a risk factor just because a group lives longer than the group of comparison, and by using secondary graft infusion as competing event, we reduce the risk of a variable being observed as a risk factor just because the group had multiple transplantations leading to increased risk of cGVHD.

Cumulative incidences are usually most relevant when the aim is to provide predictions, whereas cause‐specific hazards may be most relevant, when the aim is to study disease aetiology (Andersen et al., 2012). We chose to report both results as recommended by statisticians (I, II) (Latouche et al., 2013). To reduce the risk of confounders interfering with our results, we reported both unadjusted and adjusted analyses with confounders selected according to results from previous studies and limitations of the data material (I, II).

Chi‐squared test was used to compare organ involvement of cGVHD and the relative frequencies of ocular cGVHD (III). This method does not take competing events or confounders into account, which is a major limitation. Statistical methods to overcome these issues have been discussed with two different statisticians without finding a solution, which is why we used this simplified method. In the same article (III), we also investigated ocular cGVHD and its association with overall mortality, non‐relapse mortality and relapse, where it was possible to take the time to event and confounders into account. Jacobs et al. conducted a study where they found a better survival in patients with ocular cGVHD, but they did not address immortal time bias (which might cause better survival because the patient in the ocular cGVHD group could not experience the outcome death from time of HSCT to onset of ocular cGVHD) (Jacobs et al., 2012). In our study (III), we avoided immortal time bias by conducting a cause‐specific Cox proportional hazards models of our outcomes, coding ocular cGVHD as a time‐dependent predictor in both unadjusted and adjusted analyses (Jeppesen, Kielsen et al., 2022).

4. RESULTS

The number of patients receiving HSCT in the study period was 1936 (1452 adults and 484 children). Only 32 (2% of) adults, and 38 (8%) of paediatric patients, had missing ophthalmic examination before transplantation (I, II and III). The number of patients with DED before transplantation was 186 (13%) in adults and 25 (6%) in children. In total, 1189 adult patients and 418 children met the inclusion criteria for the primary risk analysis described in the methods section (I, II).

In adults, we found a higher 5‐year cumulative incidence of oGVHD after NMA conditioning (35% [95% CI: 30–39]) compared with after MA conditioning (18% [95% CI: 15–21]). The majority of patients developing ocular cGVHD were diagnosed within the first 3 years after transplantation (I). Risk factors of ocular cGVHD in the MA group were as follows: malignant primary disease, low S1t before transplantation, the use of a matched unrelated donor or female donor, PB as stem cell source, acute GVHD (in both unadjusted and adjusted analysis) and higher recipient age (only in unadjusted analysis) (I). Risk factors of ocular cGVHD in the NMA group were as follows: low S1t before transplantation and higher recipient age (I).

In children (II), the cumulative incidence of chronic oGVHD was 2.4% (95% CI: 1.2–4.0) 2 years after HSCT, 4.3% (95% CI: 2.7–6.6) 5 years after HSCT and the 6.0% (95% CI: 3.9–8.6) 10 years after HSCT. The median time to diagnosis of chronic oGVHD was 1056 days (range: 134–4122) after HSCT (II). The 10‐year cumulative incidence of DED (new onset) was 8.7%, and for new onset CFS, it was 12.7% (II). In adjusted analyses, malignant disease was a risk factor for developing CFS. Lower recipient age was associated with reduced risk of DED, and patients receiving cyclophosphamide as part of conditioning also had lower risk of developing DED (II).

In study III, 1221 adult patients met the inclusion criteria. During the follow‐up period, 311 (25%) of the included HSCT patients developed ocular cGVHD and 601 (49%) developed non‐ocular cGVHD. In most patients, non‐ocular cGVHD was diagnosed first (87%), whereas ocular cGVHD was diagnosed first in 8% of the patients, and in 5% they were diagnosed simultaneously within 1 week. Some patients (11%) were diagnosed with ocular cGVHD without having non‐ocular cGVHD (III). Ocular cGVHD was diagnosed on average after 518 days in adults (median, range: 34–3683), and non‐ocular cGVHD was diagnosed on average after 196 days (median, range: 41–5014) (III). The percentage of patients developing ocular cGVHD (N = 279) out of all the adult patents developing non‐ocular cGVHD (N = 601) was 46% during the study period (III). Ocular cGVHD was more frequent in patient with extensive cGVHD (51%) than in patients with limited cGVHD (29%), p < 0.001 (III). When we investigated the different organ sites involved in cGVHD, we found that ocular cGVHD was more frequent in patients with skin cGVHD (p < 0.001), oral cGVHD (p = 0.0024), genital cGVHD (p = 0.0023) or nail cGVHD (p = 0.031) (III). In the same study, we found that 70% of the patients with skin sclerosis developed oGVHD, which was higher than the patients with skin cGVHD without sclerosis (49%, p = 0.0003). A large proportion (97%) of the patients with ocular cGVHD had cGVHD in ectodermal derived organs, and ocular cGVHD was also more frequent in the group with ectodermal cGVHD than in the cGVHD group without ectodermal involvement (p < 0.0001) (III).

5. GENERAL DISCUSSION

This PhD project had three aims. The first was to report the incidence of ocular cGVHD among different conditioning regimens in both adults and children. Ocular chronic GVHD was less common in children than in adults as hypothesized. In adults, the cumulative incidence of ocular cGVHD was 18% after MA conditioning and 35% after NMA conditioning, whereas in children (only MA), the cumulative incidence was 6% (I, II). This difference may be explained by the fact that younger recipients/children still have a functioning thymus helping with the negative selection (destruction of alloreactive T cells), which induces tolerance to the recipient's tissues (Cooke et al., 2017). The incidence of ocular cGVHD has been investigated mostly in adults, where the results vary greatly from 5.5 to 60% ascribed to various populations, study designs and the use of different diagnostic criteria (Jacobs et al., 2012; Na et al., 2015; Pellegrini et al., 2021; Shikari et al., 2013; Tabbara et al., 2009; Wang et al., 2015; Westeneng et al., 2010). Our results are divided into the different conditioning regimens which stands out from previous research (I). We thought it was important to distinguish between these two types of transplantation due to the different procedure, indications and patient groups (i.e. age, comorbidities).

In our study (I), ocular chronic GVHD was not more common after MA conditioning (high dose regimen) as hypothesized. There are several factors, which may explain the higher incidence in the NMA group; there was a higher risk of death (competing event) in the group receiving MA conditioning which shortens the time frame where the patients are exposed to ocular cGVHD before they die, leading to a lower probability of ocular cGVHD. Furthermore, the age of the patients receiving NMA conditioning is higher, which is found to be a risk factor of ocular cGVHD (I).

The second aim of this PhD project was to identify risk factors for developing ocular cGVHD in both adults and children. Several factors increased the risk for developing ocular cGVHD in the adult group, which were consistent with our established hypothesis of possible risk factors, and they are discussed in detail in Study I (Jeppesen et al., 2021) (the use of PB as stem cell source, higher recipient age, MUD, female donor, malignant disease, low S1t before HSCT and aGVHD). To our surprise, the use of TBI was not a risk factor for developing oGVHD. Our hypothesis, that the tissue damage caused by TBI in the conditioning regimen would result in more ocular cGVHD, was not confirmed in our study. TBI is a known risk factor for non‐ocular aGVHD (Hahn et al., 2008), and one study showed that TBI was associated with skin sclerosis (Martires et al., 2011).

We hypothesized that the risk factors were similar in both children and adults. In the paediatric group (II), we found no risk factors for developing ocular cGVHD. An explanation for this could be the low incidence of ocular cGVHD and thereby low statistical power. However, malignant disease significantly increased the risk of new onset CFS in children indicating that the potent cytotoxic treatment used to eradicate the primary disease, is a contributing factor in the development of DED after transplantation (II).

Another hypothesis was that ocular chronic GVHD would occur at the same time after HSCT in children and adults. In adults (III), the median time to diagnosis of ocular cGVHD was 518 days (range: 34–3683) after HSCT, whereas in children (II), the diagnosis was made later with a median 1056 days after HSCT (range: 134–4122). In adults (I), most patients were diagnosed during the first 3 years after HSCT (the 3‐year cumulative incidence was 16% in the MA group and 28% in the NMA group), whereas the diagnose seemed to be made later in children (II) (2‐year cumulative incidence was 2.4%, and the 5‐year cumulative incidence was 4.3%). Bias to these results could be the difficulties performing slit lamp examination, Schirmer's test or other difficulties in assessing symptoms in children, which could potentially lead to a delayed diagnosis (or missed diagnosis). Better diagnostic guidelines for ocular GVHD are needed in the younger patient groups to make sure the paediatric population receive correct diagnosis and optimal treatment.

The third aim of this PhD project was to investigate the association between ocular cGVHD and cGVHD in other organs and outcomes after HSCT (relapse and non‐relapse mortality). Our hypothesis was confirmed that ocular cGVHD was associated with cGVHD in other organs, and non‐ocular cGVHD was diagnosed before ocular cGVHD in most cases (III). Almost half of the patients with non‐ocular cGVHD developed ocular cGVHD at some point during follow‐up and was more frequent when the patients had more widespread cGVHD (extensive) than limited disease (III). Furthermore, we found that cGVHD in organs derived from the ectoderm (skin, mouth, genitals and nails) were associated with cGVHD in the eyes which is also derived from ectoderm. This indicates that these ectodermal derived organs may share features increasing their cGVHD susceptibility (III).

The frequency of ocular cGVHD was notably high (70%) in the cGVHD group who developed skin sclerosis (III). Skin sclerosis is located in the dermis, which is of mesodermal origin. This could contradict our germ‐layer hypothesis as mentioned in our paper (III), but all the patients with skin sclerosis also had cGVHD in ectodermal derived organs elsewhere. An explanation to why ocular cGVHD is associated with cGVHD in dermis (mesoderm) could be the fact that skin sclerosis occurs late, and the patients would therefore have prolonged time to develop cGVHD in other tissues (III).

We hypothesized that patients who develop ocular cGVHD have a lower relapse rate and better survival because of the ‘graft‐versus‐tumour effect’. In our study (III), we found no association with relapse rate, but we found that the risk of non‐relapse mortality was higher in patients that developed ocular cGVHD (in the cGVHD group), which was not consistent with our original hypothesis. The association between increased non‐relapse mortality and ocular cGVHD may be caused the fact that ocular cGVHD is associated with extensive cGVHD (more advanced systemic cGVHD where the mortality rate is higher).

A task force report from NIH regarding the biology of chronic GVHD describes some of the gaps in understanding cGVHD biology (Cooke et al., 2017). One of their concluding questions is as follows; ‘What are the target antigens of chronic GVHD in humans?’ Our study (III) potentially contributes to one of many future answers to this question with the finding of the association between ocular cGVHD and cGVHD in other ectodermal derived organs. Further studies are needed to explore whether specific ectodermal target antigens contribute to the complex pathophysiology of ocular cGVHD. New and more specified approaches could be developed if these potential pathways are manipulated to induce immunological tolerance towards these target antigens (antigen‐specific immunotherapy, ASI). In mice, ASI has been investigated as treatment for autoimmune disease with promising results, but promising results have yet to be found in humans (Hirsch & Ponda, 2015; Richardson & Wraith, 2021).

DED was found in both adults (13%) and paediatric patients (6%) before transplantation. In adults (I), there was significant more DED in patients with malignant disease (p = 0.005), and the results in children (II) showed a similar trend towards increased DED before HSCT in the group with malignant disease (p = 0.08). In adults (NMA group), we found that the cumulative incidence of ‘assumed oGVHD’ was significantly increased in patients with ‘DEWS score 1’ before HSCT compared with the group without DED before HSCT (43% vs. 35%, p = 0.04) (Study I). In children (II), we found that 29% of the patients with DED before HSCT developed signs of ocular cGVHD, but the results were not statistically significant (p = 0.16). These results (including the results of S1t before HSCT) underscores the importance of an ophthalmological evaluation before HSCT in order to be able to detect and commence treatment in these patients, when DED is detected.

One of the greatest limitations of our studies is the retrospective nature of the design (I, II, III). Some of the annual examinations are missing, and some patients have missing clinical values when an ophthalmological examination has been performed, which might have caused bias (i.e. delayed diagnosis or underestimation of ocular manifestations). The conversion of NIH eye score to OSDI score might also have resulted in bias by underestimating symptoms. Furthermore, challenges in examining children in the slit lamp, completing Schirmer's test, and reporting symptoms could also cause bias. A factor that could have strengthened the study could be that more than one person read the charts and entered information in the database to ensure correct information.

The strengths of our studies are that data come from a single centre, and the patents are consecutive with a baseline eye examination performed before HSCT. The follow‐up period was long, and only a limited number of patients were lost to follow‐up (for other reasons than death or a new graft infusion) (I, II, and III).

It has been debatable whether acute oGVHD exists, and ocular manifestations are not described in the existing consensus reports of acute GVHD (Harris et al., 2016; Inamoto et al., 2019; Przepiorka et al., 1995). Our study in children (II) showed that some of the patients with systemic aGVHD also had conjunctival inflammation simultaneously, suggesting an aGVHD response in the eyes, but because of the retrospective nature of our study, it was not possible to distinguish whether it was caused by alloreactivity or virus/allergy. Another research group found T‐lymphocytes derived from donor in ocular pseudomembranes in a patient with systemic aGVHD, which indicates that aGVHD can affect the eyes (Saito et al., 2002). In accordance with this, our study (II) showed a higher risk of acute oGVHD when busulfan was used in the conditioning, which is consistent with busulfan being a risk factor for developing non‐ocular aGVHD in other studies (Andersson et al., 2002; Bartelink et al., 2016). More studies are needed to clarify the pathophysiology of acute oGVHD to implement diagnostic criteria and improve treatment.

The current available diagnostic criteria for ocular cGVHD are not optimal either, especially not for children, but also not for adults. One of the important questions is, how do we distinguish ocular cGVHD from DED from other causes? Kantor et al. have tried to answer this question (Kantor et al., 2024). Both eye diseases cause similar symptoms and objective findings of the ocular surface like signs of dry eye, visual disturbances, decreased S1t, MGD and epithelial disruption (Kantor et al., 2024). Some of the few distinctions are pseudomembranous conjunctivitis in ocular aGVHD, and fibrosis in ocular cGVHD that is not commonly seen in DED. They share cellular and soluble mediators, but there are still many gaps in the literature on oGVHD (Kantor et al., 2024). New approaches to diagnostics are needed to ensure that the patients get the right diagnosis early, and the right treatment can be commenced as soon as possible to try to avoid the severe late complications of ocular GVHD.

6. FUTURE PERSPECTIVES

In the future, hopefully we will know more about which factors may predict the development of oGVHD. Identification of biomarkers could improve diagnostics and optimize treatment. This will enable early detection of the disease and distinguish it from DED caused by other reasons. The tears are easily accessible and may be sampled by non‐invasive methods, which makes it ideal to examine biomarkers for detecting and monitoring oGVHD (when the disease is not too progressed and the lacrimal gland not too destructed to produce tears). Cytokine profiling, proteomics, lipid profile, leukocyte profile in tear fluid and the microbiome of the ocular surface are some of the approaches to detect oGVHD biomarkers (Bohlen et al., 2024). Some studies have been conducted, but more are needed to find biomarkers that can be used to diagnose or monitor a patients disease in the clinic one day.

Numerous factors contribute to the development of oGVHD, and when these factors have been clarified, a more personalized approach in the treatment of oGVHD could be introduced. Perhaps the evolution of artificial intelligence (AI) will assist the risk assessment of GVHD, diagnostics and enable personalized treatment (Mushtaq et al., 2023). Today, researchers may use statistical tools like multivariate Cox proportional hazard models and logistic regression to assess the individual risk of GVHD development, but the existing statistical models have limitations in the accuracy of prediction, generalizability and the models are often unavailable to the clinicians (Arai et al., 2019; Mushtaq et al., 2023). AI could potentially revolutionize the procedures of analysing large and complex HSCT registries, resulting in optimized donor selection, personalized risk assessment, early diagnosis and treatment. However, a lot of obstacles must be resolved before this could be a reality. For instance, very large‐scale collaborations are necessary to create the high‐quality datasets required for generalizable AI algorithms, cross‐centre generalizability will be challenging due to bias/confounders that do not apply the total HSCT population, and unambiguous ethical standards must be decided before introducing methods based on AI into the clinic. Furthermore, it is crucial to ensure the security and privacy of the clinical data and secure the autonomy of the patients (Mushtaq et al., 2023). Overall, this field of research faces an exciting future.

7. GENERAL CONCLUSIONS

Despite the many years HSCT has existed as a curative treatment option, GVHD remains a difficult challenge. We found no significant difference in the incidence of ocular cGVHD in either adults or children during the long timespan of our studies emphasizing, that ocular cGVHD remains a substantial problem. The studies in this thesis have contributed with knowledge of risk factors for the development of ocular GVHD in both children and adults. The risk factors included DED and low Schirmer's test before transplantation, malignant diagnosis of recipient, the use of an unrelated or female donor, use of PB as stem cell source, acute GVHD (grade III‐IV) and higher recipient age at transplantation.

Ocular cGVHD is associated with cGVHD in other ectodermal derived organs and is more common when cGVHD is extensive. Ectodermal target antigens, involved in the pathophysiology of oGVHD, may exist. Non‐relapse mortality was increased in patients developing ocular cGVHD (with patients developing cGVHD as reference).

Even though we have identified several risk factors of developing oGVHD, the puzzle of ocular GVHD has not been solved. The pathophysiology is very complex and far from known. The diagnosis of ocular GVHD can be difficult to assess. We need more research in new ways to predict, diagnose and treat oGVHD. Treatment is often commenced during the symptomatic stage, where damage to the lacrimal gland and conjunctival scarring might already be permanent. Therefore, it is important with early detection of the disease.

ASSESSMENT COMMITTEE

Sarah Linea von Holstein, Associate Professor, MD, PhD, Department of clinical Medicine, University of Copenhagen, Denmark (Chair).

Henrik Vorum, Professor, MD, DMSc, Aalborg University Hospital, Denmark.

Geir E. Tjønnfjord, Professor emeritus, MD, DMSc, Oslo University Hospital, Norway.

LIST OF ACADEMIC SUPERVISORS

Principal supervisor: Steffen Heegaard, Professor, MD, DMSc, Departments of Pathology and Ophthalmology, Rigshospitalet, Copenhagen University Hospital, Copenhagen, Denmark.

Co‐supervisors:

Henrik Sengeløv, Department of Haematology, Professor, MD, DMSc, The Bone Marrow Transplantation Unit, Rigshospitalet, Copenhagen University Hospital, Copenhagen, Denmark (Primary co‐supervisor).

Jens Lindegaard, PhD, MD, Copenhagen Eye Infirmary, Copenhagen, Denmark.

Hanne Olsen Julian, PhD, MD, Department of Ophthalmology, Mølholm Private Hospital, Denmark.

GLOSSARY

Allogeneic transplantation	Transplantation with graft from a donor.
Autologous transplantation	Transplantation with harvesting and re‐infusion of the patients own cells.
Conditioning regimen	Treatment given to prepare the patient for stem cell transplantation which may include chemotherapy and total body irradiation (TBI). Purpose of this treatment is to kill cancer cells and prevent rejection of the transplanted cells.
Graft‐versus‐host disease (GHVD)	A toxic immunological response mediated by donor‐derived T cells towards the recipient's cells.
Graft‐versus‐leukaemia effect	Toxic destruction of tumour cells mediated by the donor‐derived T cells.
Haematopoietic stem cell	An immature cell of the haematopoietic system that give rise to all the cells of the blood system.
Myeloablative transplantation (MA)	Myeloablative conditioning regimen (high dose) is given to prepare the patient for transplantation and causes irreversible cytopenia.
Non‐myeloablative transplantation (NMA)	Non‐myeloablative conditioning regimen (low dose) is given to prepare the patient for transplantation and causes reversible cytopenia.
Syngeneic transplantation	Transplantation with graft from a donor with identical genotype as recipient (identical twins).

Jeppesen, H. (2025) Ocular graft‐versus‐host disease: Risk factors of ocular graft‐versus‐host disease after allogeneic haematopoietic stem cell transplantation in Denmark. Acta Ophthalmologica, 103(Suppl. 286), 3–19. Available from: 10.1111/aos.17452