Predictors of rapidly progressive interstitial lung disease and mortality in patients with autoantibodies against melanoma differentiation-associated protein 5 dermatomyositis

Abstract

Objective

Anti-melanoma differentiation-associated protein 5 (MDA5)-positive DM is associated with rapidly progressive interstitial lung disease (RP-ILD) and high mortality. This multicentre retrospective study aimed to identify predictors of mortality and RP-ILD.

Methods

Anti-MDA5-positive DM patients were identified from the Hong Kong Myositis Registry and the Clinical Data Analysis and Reporting System. Clinical characteristics were reviewed. Risk factors for mortality and RP-ILD were identified.

Results

Among the 116 recruited patients, 100 (86.2%) had ILD, 47 (40.5%) had RP-ILD and 44 (37.9%) patients died. Cox regression analysis revealed RP-ILD [hazard ratio (HR) 9.735 (95% CI 3.905, 24.272)], age >52 years [HR 4.750 (95% CI 1.692, 13.333)], ferritin level >2800 pmol/l [HR 3.042 (95% CI 1.323, 6.997)] and lactate dehydrogenase (LDH) >400 IU/l [HR 2.290 (95% CI 1.009, 5.198)] were independent predictors of mortality. With regard to RP-ILD, analyses showed that potential predictors at baseline included age >50 years [HR 2.640 (95% CI 1.277, 5.455)], LDH >300 IU/l [HR 3.189 (95% CI 1.469, 6.918)], fever [HR 1.903 (95% CI 0.956, 3.790)] and neutrophil:lymphocyte ratio >7.0 [HR 1.967 (95% CI 0.942, 4.107)]. We proposed a prediction model based on fever, LDH, age and white cell count (FLAW) to stratify the risk of development of RP-ILD. The probability of RP-ILD in a patient with a score of 4 was 100%. A small internal validation cohort showed the odds of RP-ILD with FLAW scores of 0, 1, 2 and 3 were 0%, 0%, 42.9% and 75%, respectively.

Conclusions

Anti-MDA5-associated RP-ILD is significantly associated with poor survival rates. The FLAW model maybe useful to predict the development of RP-ILD.

Introduction

Idiopathic inflammatory myopathies (IIMs) are a heterogeneous group of disorders characterized by muscle or skin inflammation and associated with various systemic complications. Myositis-specific antibodies (MSAs) are pivotal biomarkers for predicting clinical manifestations and disease prognosis in IIMs. Anti-melanoma differentiation-associated protein 5 (MDA5) antibody was first identified in 2005 by Sato et al. and is one of the most commonly detected MSAs, particularly in East Asians [1, 2]. It is associated with clinically amyopathic DM (CADM) and skin ulcerations [3]. Anti-MDA5 antibody–related DM is gaining increasing attention due to its complication of rapidly progressive interstitial lung disease (RP-ILD) and high mortality [3–5]. A recently published cohort study of 90 anti-MDA5-positive DM patients showed that 38.9% (35/90) developed RP-ILD and 24.4% (22/90) died within 6 months of diagnosis. Yet the association of RP-ILD and mortality was not assessed [6].

Gono et al. [7] recently proposed a risk prediction model for mortality in patients with IIM-related ILD, based on a combination of biomarkers at baseline, including the presence of anti-MDA5 antibody, Kerbs von den Lungen 6 (KL-6) and CRP level, using data from the Japanese Patients with Myositis-ILD (JAMI) database. However, only a limited number of candidate biomarkers were analysed according to the availability of information in the database. Also, serum KL-6 is not a readily available test in daily clinical practice. Another group designed the FLAIR model using data from a single centre, which includes ferritin, lactate dehydrogenase (LDH), anti-MDA5 antibody, high-resolution CT (HRCT) score and the presence RP-ILD, to predict mortality in CADM patients with ILD [8]. However, this model requires the anti-MDA5 antibody titre, which may not be standardized because of the use of different assays. Also, the HRCT score involved a complex formula and clinicians may need training to achieve accurate assessments of CT features.

Early and aggressive treatments are generally recommended to improve outcome in anti-MDA5 DM patients. However, potent immunosuppressants carry the risk of serious infections and other drug-specific side effects [9]. Risk stratification of DM patients with anti-MDA5 in terms of RP-ILD development at diagnosis is required in order to guide management. At present, there is no clinical prediction model to identify patients with a high risk of RP-ILD.

We hypothesized that RP-ILD is associated with increased mortality in patients with MDA-5-associated DM. We conducted a multicentre retrospective study to identify poor prognostic factors associated with mortality in anti-MDA5-positive DM and to ascertain the predictors for the development of RP-ILD.

Methods

Study design

This was a multicentre, retrospective cohort study. The primary objective of the study was to identify risk factors of mortality and RP-ILD in anti-MDA5-positive DM patients. The secondary objective of the study was to generate a prediction model for RP-ILD.

The study was conducted in accordance with the Declaration of Helsinki. Institutional ethics approvals were obtained for this study. Consents were waived in view of the retrospective nature of the study.

Patients and data source

Ten regional public hospitals participated in this study. Patients with IIMs diagnosed between 1 January 2015 and 31 December 2020 were identified from MyoHK and the Clinical Data Analysis and Reporting System (CDARS). MyoHK is a territory-wide, population-based registry that was set up in 2019 that aims to systemically collect clinical information on IIM patients in Hong Kong. The CDARS is an electronic system created by the Hong Kong Hospital Authority in 1991, mainly for audit and research purposes. The system has been extensively used in large-scale epidemiological studies [10, 11]. In order to avoid missing cases, identification of patients from MyoHK was supplemented by a CDARS search using the International Classification of Disease, 9th Revision, Clinical Modification (ICD9-CM) codes for myositis (729.1), DM (710.3) and polymyositis (710.4). All participating hospitals and their specialist clinics used the same computer software platform for clinical information and records. All clinical data were stored in this electronic patient record (ePR) system. The clinical records of all IIM patients identified via MyoHK and CDARS were reviewed by the investigator individually via the ePR to screen for eligibility for recruitment into the study. Inclusion criteria included adult-onset disease (≥18 years); diagnosis of DM based on Bohan and Peter’s criteria or the EULAR/ACR 2017 classification criteria [12, 13]; diagnosis of CADM, defined as typical cutaneous features of DM confirmed by a rheumatologist or dermatologist but minimal or no clinical features of myositis (hypomyopathic and amyopathic) according to the Sontheimer’s criteria [14]; and a positive anti-MDA5 antibody result. Exclusion criteria included myositis secondary to trauma, seizure, alcohol or drug abuse, myopathic drugs, such as statin and fibrates, or following a recent viral illness and patients with other concomitant connective tissue diseases. Further recruitment of anti-MDA5-positive DM patients diagnosed between 1 January 2021 and 30 October 2021 was performed for internal validation of the FLAW (fever, LDH, age and white cell count) model.

Clinical data

Data for the recruited patients were reviewed and collected from the date of DM diagnosis until either death or the end of study. Basic demographics including age, sex, smoking history, ethnicity and detailed drug history were assessed. Patients with cancer diagnosed 1 year before or 3 years after the diagnosis of DM were documented. The clinical presentation at diagnosis, including fever, muscle weakness, dysphagia and cutaneous features, was recorded. Infections at and after diagnosis were reviewed. Fever was defined as a body temperature ≥38°C. Patients with ILD and RP-ILD were identified. ILD was defined by typical radiological features of ILD on CT or high-resolution CT (HRCT) [15]. RP-ILD was defined as worsening of radiological interstitial change, progressive dyspnoea and hypoxaemia within 1 month of onset of respiratory symptoms regardless of treatment [16, 17]. Muscle biopsy, electromyography and laboratory findings at diagnosis, including creatine kinase (CK), LDH, ferritin, liver function test, white blood cell count with differential count, ANA, MSAs and inflammatory markers including ESR and CRP were obtained. The neutrophil:lymphocyte ratio (NLR), a potential biomarker for RP-ILD that has been shown to be useful in predicting acute and severe lung disease, was calculated by dividing the absolute neutrophil count by the absolute lymphocyte count [18, 19]. Treatments received were reviewed. High-dose glucocorticoids were defined as prednisolone >0.5 mg/kg/day. Time from symptoms onset to diagnosis of DM, development of RP-ILD and death were ascertained. Causes of death were reviewed.

Identification of anti-MDA5 antibody

Commercial line blot immunoassay kits [Euroline Autoimmune Inflammatory Myopathies 15 Ag (IgG); Euroimmun, Lübeck, Germany) were used to detect the MSAs. Borderline anti-MDA5 antibody results were not regarded as positive.

Statistical analysis

Descriptive statistics were used to present the demographic, clinical and treatment variables. Univariate analysis was performed to identify potential risk factors associated with mortality and RP-ILD. Independent samples Student’s t-test was used to compare continuous variables and Fisher’s exact test or the chi-squared test were used for categorical variables. The Mann–Whitney U test was used for variables with a skewed distribution. Biologically plausible clinical factors with a P-value <0.1 in the univariate analyses were included in the multivariable analyses. Receiver operating characteristics (ROC) curves with multivariable analysis were used for identification of potential biomarkers. The optimal cut-off values for predicting all-cause mortality and RP-ILD were determined by calculating the area under the curve (AUC) and the Youden index. Dichotomous variables were applied to the Cox regression analysis to determine the independent predictors with individual hazard ratios (HRs) for mortality and RP-ILD, respectively, with additional adjustment for age and gender. A prediction model for RP-ILD was generated by using clinical factors with a P-value <0.1 from the Cox regression analysis. Kaplan–Meier curves were used to visualize the cumulative survival rates. A P-value <0.05 was considered statistically significant. Statistical analyses were performed by using the Statistics Package for Social Sciences (version 24.0; IBM, Armonk, NY, USA).

Results

Patients’ demographics, clinical characteristics and outcomes

A total of 116 anti-MDA5-positive DM patients diagnosed between January 2015 and December 2020 were recruited into the study. The mean follow-up duration was 19.3 months (s.d. 21.6). Their baseline demographics and outcomes are shown in Table 1. The clinical characteristics, laboratory findings and treatments are shown in Supplementary Table S1A–D, available at Rheumatology online. There was a slight female predominance (56%). The mean age at diagnosis was 52 years (s.d. 13; range 21–88). The great majority of the patients were Chinese [112/116 (97%)]. A total of 38 (32.8%) and 78 (67.2%) patients had DM and CADM, respectively. Of the 116 patients, 100 (86.2%) had ILD and 47 (40.5%) developed RP-ILD. The majority [41/47 (87.2%)] of the patients with RP-ILD developed the condition within 3 months of diagnosis of DM, as shown in Fig. 1. The overall mortality of patients with anti-MDA5-related DM was 37.9% (44/116). More than half [28/47 (59.5%)] of the patients with RP-ILD died within 3 months of diagnosis.

Comparison of clinical manifestations and laboratory features between survivors and non-survivors

Univariate analysis revealed that older age of onset (60 years (s.d. 9) vs 48 (13), P = 0.005), smoking [6/44 (13.6%) vs 3/72 (4.2%), P = 0.074], DM subtype [20/44 (45.5%) vs 18/72 (25%), P = 0.023], heliotrope rash [27/44 (61.4%) vs 27/72 (37.5%), P = 0.012], RP-ILD [37/44 (84.1%) vs 10/72 (13.9%), P < 0.001], fever [30/44 (68.2%) vs 27/72 (37.5%), P = 0.002] and infection [16/44 (36.4%) vs 11/72 (15.3%), P = 0.009] at initial presentation were significantly associated with mortality (Supplementary Table S2A, available at Rheumatology online). Non-survivors had a tendency towards a higher prevalence of pneumothorax [9/44 (20.5%) vs 6/72 (8.3%), P = 0.059] and cancer [4/44 (9.1%) vs 1/72 (1.4%), P = 0.068]. Of 44 non-survivors, 34 (77.3%) were diagnosed within 3 months of symptoms onset. More non-survivors were diagnosed within 3 months than those who survived [42/44 (95.5%) vs 49/72 (68.1%), P < 0.001], had higher CK [550 U/l (s.d. 778) vs 211 (218), P < 0.001], LDH (663 IU/l (s.d. 578) vs 351 (132), P < 0.001], CRP [22.7 mg/dl (s.d. 29.9) vs 11.6 (17.6), P < 0.001] and ferritin [14 309 ng/ml (s.d. 35 760) vs 4100 (5589), P = 0.006] levels at baseline. There was a trend towards a higher NLR [8.1 (s.d. 6.3) vs 6.3 (4.9), P = 0.054] in non-survivors (Supplementary Table S2B, available at Rheumatology online).

Using ROC curve analysis, the optimal cut-offs for the potential poor prognostic factors at diagnosis associated with an increased mortality are age >52 years (AUC 0.770, sensitivity 0.840, specificity 0.667, P = 0.000), LDH >394 IU/l (AUC 0.776, sensitivity 0.760, specificity 0.689, P = 0.000), ferritin >2835 pmol/l (AUC 0.653, sensitivity 0.800, specificity 0.533, P = 0.034) and CK >134 IU/l (AUC 0.694, sensitivity 0.840, specificity 0.556, P = 0.008) (Supplementary Table S2C, available at Rheumatology online). The individual serum biomarker cut-off values were rounded off for further analysis.

Independent predictors of mortality among anti-MDA5-positive DM patients

In multivariable Cox regression analysis, RP-ILD was the strongest predictor of mortality [HR 9.735 (95% CI 3.905, 24.272), P < 0.001] after adjusting for other covariates, as shown in Table 2. Other independent predictors of mortality included age >52 years [HR 4.750 (95% CI 1.692, 13.333), P = 0.003], serum LDH >400 IU/l [HR 2.29 (95% CI 1.009, 5.198), P = 0.047] and serum ferritin level >2800 pmol/l [HR 3.042 (95% CI 1.323, 6.997), P = 0.009].

The cumulative survival for anti-MDA5-positive DM patients is shown in the Kaplan–Meier curve in Supplementary Fig. S1, available at Rheumatology online. Significantly lower cumulative survival was observed in patients with RP-ILD (P = 0.000) (Fig. 2).

Comparison of clinical manifestations and laboratory features in patients with and without RP-ILD

In univariate analysis, patients with RP-ILD were generally older than those without RP-ILD [57 years (s.d. 10) vs 50 (13), P = 0.040] (Supplementary Table S3A, available at Rheumatology online). More patients with RP-ILD were diagnosed within 3 months of symptoms onset than those without [RP-ILD 43/47 (91.5%) vs non-RP-ILD 48/69 (69.6%), P = 0.005]. Fever [RP-ILD 31/47 (66.0%) vs non-RP-ILD 26/69 (38.2%), P = 0.003] and infection [RP-ILD 16/47 (34.0%) vs non-RP-ILD 11/69 (15.9%), P = 0.024] at presentation were significantly associated with an increased risk of RP-ILD. There was also a trend towards more V sign [RP-ILD 10/47 (21.3%) vs non-RP-ILD 7/69 (10.1%)] noted in patients with RP-ILD, whereas higher LDH [RP-ILD 591 IU/l (s.d. 548) vs non-RP-ILD 382 (194), P = 0.028], CRP [RP-ILD 24.2 mg/dl (s.d. 30.3) vs non-RP-ILD 9.8 (14.6), P < 0.001] and NLR [RP-ILD 8.4 (s.d. 6.4) vs non-RP-ILD 5.9 (4.5), P = 0.006] were significantly associated with RP-ILD (Supplementary Table S3B, available at Rheumatology online).

The ROC analyses were performed to reveal the optimal cut-off level for age, serum LDH, CRP and NLR at diagnosis for RP-ILD (Supplementary Table S3C, available at Rheumatology online). The AUC of age, serum LDH and CRP at diagnosis in predicting RP-ILD were 0.691 (sensitivity 0.786, specificity 0.551, P = 0.002), 0.681 (sensitivity 0.881, specificity 0.469, P = 0.003) and 0.669 (sensitivity 0.429, specificity 0.898, P = 0.006), respectively, with cut-off values of 50 years old, 311 IU/l and 18 mg/l, respectively. The optimal cut-off value for NLR in predicting RP-ILD was 7 (sensitivity 0.429, specificity 0.776, P = 0.060), with an AUC of 0.607. The individual serum biomarkers cut-off values were rounded off for further analysis.

Independent predictive factors for the development of RP-ILD among anti-MDA5-positive DM patients

In multivariable Cox regression analysis, age >50 years [HR 2.640 (95% CI 1.277, 5.455), P = 0.009] and serum LDH >300 IU/l at diagnosis [HR 3.189 (95% CI 1.469, 6.918), P = 0.003] were independent risk factors for RP-ILD, as shown in Table 3. Fever [HR 1.903 (95% CI 0.956, 3.790), P = 0.067] and serum NLR >7.0 at diagnosis [HR 1.967 (95% CI 0.942, 4.107), P = 0.072] also appeared to be associated with an increased risk of RP-ILD.

Prediction model for RP-ILD—the FLAW model

Potential independent risk factors at baseline identified from Cox regression analysis were included in the prediction models for RP-ILD. The prediction model for RP-ILD was based on fever, LDH >300 IU/l, age >50 years and NLR >7 at diagnosis. The FLAW score was defined as the number of risk factors for RP-ILD. The odds of RP-ILD for patients with FLAW scores of 0, 1, 2, 3 and 4 were 7.1%, 17.9%, 38.7%, 57.6% and 100%, respectively, as shown in Table 4. An internal validation with 16 anti-MDA5-positive DM patients recruited from December 2020 onward showed the odds of RP-ILD for patients with FLAW scores of 0, 1, 2 and 3 were 0%, 0%, 42.9% and 75%, respectively (no patient scored 4) (Table 4).

Discussion

This study is one of the largest cohorts that comprehensively reviewed the demographic, clinical and laboratory characteristics as well as outcomes of DM patients with anti-MDA5. The high overall mortality of our patients with anti-MDA5-related DM (37.9%) was similar to their Japanese counterparts (36–41%) [3, 20] but higher than European patients (27.3%) [21]. Consistent with previous studies [3, 6, 22, 23], we found that most patients indeed died early in the disease course, with 3 month and 6 month mortality rates of 28.4% and 31%, respectively.

In our study, we noted that RP-ILD was the most important predictor of mortality, with a 9.7-fold increased risk of death. Nearly 90% of the RP-ILD patients in our cohort developed this complication within 3 months of diagnosis of DM. Age >52 years, serum ferritin >2800 pmol/l at presentation and LDH >400 IU/l at presentation were other risk factors for mortality. Given its high incidence (40.5%), acute worsening nature, strong association with mortality and potential reversibility, we believe RP-ILD risk stratification in anti-MDA5-positive DM patients at diagnosis is most crucial clinically so that prompt appropriate treatments can be given to improve a patient’s outcome. In this study we found that age >50 years and LDH >300 IU/l at baseline were independent predictors of RP-ILD. Fever and an NLR >7 at presentation were also potentially associated with the development of RP-ILD.

LDH is highly expressed in skeletal muscles and is released during tissue damage [24]. The LDH concentration in the blood reflects disease activity and severity in idiopathic pulmonary fibrosis [25]. High LDH levels reflect the degree of lung injury in severe pulmonary fibrosis. In patients with RP-ILD, we found that the LDH level, but not the CK level, was significantly higher in patients with RP-ILD. Serum LDH >300 IU/l and 400 IU/l at diagnosis were associated with a 3.2-fold and 2.3-fold increased risk of RP-ILD and mortality, respectively.

Lymphopenia was observed in anti-MDA5-positive patients [26]. It is postulated that a low lymphocyte count in the blood may be due to transfer of lymphocytes from the blood to lung tissue due to a local immune response [27]. NLR is a simple method to assess systemic inflammation and has been used to predict mortality in cancers, major cardiac events and sepsis [28–31]. In normal healthy individuals, the mean NLR was 1.65 (range 0.78–3.53) [32]. An NLR >2.4 may predict ILD in patients with connective tissue disease [33]. In DM patients, an NLR >3.98 may be useful to differentiate patients with and without ILD [34]. In our study, the mean NLR was significantly higher in patients with RP-ILD compared with those without RP-ILD. An NLR >7 appeared to be an independent risk factor for RP-ILD.

Herein we propose a new risk prediction model for anti-MDA5-positive DM patients—the FLAW model. Compared with other prediction models, this is a simple pragmatic model designed for anti-MDA5-positive DM patients. LDH and NLR are inexpensive biomarkers that are readily available. The model offers graded risk estimation. Importantly, 10 of 10 patients with a score of 4 had RP-ILD, while only 1 of 13 with a score of 0 eventually developed RP-ILD. This model allows rapid clinical risk stratification at the diagnosis of anti-MDA5-related DM for the development of RP-ILD. In view of the high mortality risk within the first 3 months of diagnosis, potent immunosuppressants should be considered in patients at high risk of RP-ILD, even before the onset of ILD. The preliminary validation showed promising results. Future studies are required to validate this novel prediction model prospectively and to assess its generalizability in other populations.

There are some limitations in this study. First, this was a retrospective and observational study in which incomplete or unstandardized data collection could be an important source of systematic error. Identification and documentation of subtle but clinically relevant features such as vasculitis rash and cutaneous ulcers may be missed. Second, there may be survival bias in the study, as patients with severe disease who died early, especially those with CADM and RP-ILD, may not have been tested for anti-MDA5 antibody and may thus be underrepresented. Third, the effect of the treatments may be confounded by indications. The occurrence of RP-ILD may also be affected by the immunosuppressive treatments the patients received at diagnosis. Fourth, the clinical deterioration of patients could be attributed to chest infection rather than RP-ILD, although subgroup analyses excluding those with infection at presentation identified largely similar predictors of mortality as well as RP-ILD (Supplementary Tables S4 and S5, available at Rheumatology online). Lastly, ILD patterns and CT extent of involvement were not assessed in this study. Future studies could investigate the CT findings and their association with RP-ILD and mortality.

Conclusion

In this large retrospective study, we reported the rapid deterioration and high mortality risk in anti-MDA5-positive DM patients. RP-ILD was the single most important predictor of mortality. Older age (>52 years), higher ferritin level (>2800 pmol/l) and high LDH level (>400 IU/l) were also independently associated with mortality. We propose that the FLAW model, consisting of fever, LDH (>300 IU/l), age (>50 years) and white cell count (NLR >7) at presentation, can be used in predicting the development of RP-ILD. With better awareness and risk stratification, hopefully more timely and appropriate treatment can be offered in order to improve the outcomes of anti-MDA5-positive DM patients.

Acknowledgements

We would like to express our gratitude to all recruited patients and medical staff. All authors critically revised the manuscript for important intellectual content. J.S., H.S., C.C.M. and L.S.T. were responsible for study design. J.S., H.S., P.C.H.W., L.H.P.T., T.T.O.L., C.C.M., C.H.T., Y.K.C., V.T.L.W., T.Y.W., W.L.L., R.H., C.H. and C.S.L. were responsible for data collection. J.S. and H.S. were responsible for data analysis. J.S., H.S. and L.S.T. drafted the manuscript. Ethical approval was provided by the Joint Chinese University of Hong Kong–New Territories East Cluster Clinical Research Ethics Committee (2020-0743), New Territories West Cluster Research Ethics Committee (20132), Hong Kong East Cluster Clinical Research Ethics Committee (2021-0010), Hong Kong West Cluster Research Ethics Committee (2021-0037), Research Ethics Committee Kowloon Central/Kowloon East (2021-0022) and Kowloon West Cluster Research Ethics Committee (2020-0194).

Funding: Funding was provided by the Hong Kong Society of Rheumatology Project Fund 2019.

Disclosure statement: The authors have declared no conflicts of interest.

Data availability statement

Data available on request.

Supplementary data

Supplementary data are available at Rheumatology online.

References

Author notes