Thymectomy in ocular myasthenia gravis before generalization results in a higher remission rate

Abstract

Open in new tabDownload slide

OBJECTIVES

This study aimed to compare the outcomes of patients with ocular myasthenia gravis (OMG) who underwent thymectomy before generalization with the outcomes of those who underwent thymectomy after generalization.

METHODS

We retrospectively reviewed patients who underwent robotic thymectomy for myasthenia gravis between January 2003 and February 2018. Patients who presented with purely ocular symptoms at myasthenia gravis onset were eligible for inclusion. Exclusion criteria were patients who were lost to follow-up and patients who underwent re-thymectomy. Patients with OMG who developed generalization before thymectomy were categorized into gOMG group and those who did not were categorized into OMG group. The primary outcome was complete stable remission according to the Myasthenia Gravis Foundation of America Post-Intervention Status (MGFA-PIS).

RESULTS

One hundred and sixty-five (66 males and 99 females) out of 596 patients with myasthenia gravis were eligible for inclusion. Of these, there were 73 and 92 patients undergoing thymectomy before and after the generalization of OMG, respectively. After propensity score matching, a data set of 130 patients (65 per group) was formed and evaluating results showed no statistical differences between the 2 groups. The estimated cumulative probabilities of complete stable remission at 5 years were 49.5% [95% confidence interval (CI) 0.345–0.611] in the OMG group and 33.4% (95% CI 0.176–0.462) in the gOMG group (P = 0.0053). Similar results were also found in patients with non-thymomatous subgroup [55 patients per group, OMG vs gOMG, 53.5% (95% CI 0.370–0.656) vs 28.9% (95% CI 0.131–0.419), P = 0.0041].

CONCLUSIONS

Thymectomy in OMG before generalization might result in a higher rate of complete stable remission than thymectomy after generalization.

INTRODUCTION

Ocular myasthenia gravis (OMG) is frequently defined as patients who present with purely ocular symptoms at myasthenia gravis (MG) onset [1]. About 60% OMG will progress to secondary generalization (gOMG), characterized by generalized skeletal muscle weakness, typically within 2 years [2]. Retrospective studies demonstrated that the use of immunosuppressive medications is likely to improve symptoms and reduce the risk of secondary generalization in patients with OMG [1, 3, 4]. Moreover, epidemiological studies also showed a lower generalization rate of OMG after the use of immunosuppressive medications [4], compared to that before the era of immunosuppression [5]. However, even low-to-moderate dose oral corticosteroid treatment can result in adverse effects, such as heartburn, insomnia, weight gain and myalgia, in about 35% patients with OMG [6].

Thymectomy is an indispensable treatment for patients with MG [7]. In 2016, a randomized trial of extended transsternal thymectomy in non-thymomatous MG patients determined the therapeutic value of thymectomy [8]. Later in 2019, the 5-year results of non-thymomatous MG patients persistently favour thymectomy combined with prednisone therapy over prednisone alone [9]. Currently, controversy persists regarding the indications of thymectomy for patients with OMG. Some studies suggested no benefit for OMG patients who underwent thymectomy as compared to medical treatment [4, 10]. However, a well-designed retrospective study demonstrated that early treatment, especially early thymectomy, is a significant predictor of stable remission in patients with OMG [11]. Importantly, a meta-analysis has analysed retrospective studies of thymectomy in patients with non-thymomatous OMG, concluding thymectomy being a beneficial treatment with a pooled rate of complete stable remission (CSR) of 0.5074 [12]. However, there is no high-level evidence whether delaying thymectomy until the time of generalization can result in any harm for patients with OMG. Therefore, we aimed to compare the clinical outcomes of patients with OMG and patients with gOMG after robotic thymectomy.

MATERIALS AND METHODS

Study design

This is a monocentric retrospective observational study comparing the clinical outcomes of OMG patients undergoing robotic thymectomy before and after secondary generalization. The study was approved by the Ethics Committee of Charité Universitaetsmedizin Berlin (Berlin, Germany) and performed according to the Declaration of Helsinki and its amendments. Written informed consent for surgery was obtained from each patient, but written informed consent for the use of clinical data was not required according to § 25 Landeskrankenhausgesetz Berlin.

Patients

From January 2003 to February 2018, a total of 596 patients with MG underwent robotic thymectomy in our department. Data were retrospectively collected from the hospital information system. Patients with MG who presented with purely ocular symptoms at disease onset were eligible for inclusion. Exclusion criteria were patients who were lost to follow-up and patients who underwent re-thymectomy. Patients with OMG who developed generalized symptoms, even within 1 month from disease onset, before thymectomy were categorized into the gOMG group and those who did not were categorized into the OMG group. We did not exclude patients who were anti-acetylcholinesterase receptor (AChR) antibody-negative in order to get a greater sample size and minimize the selection bias. Moreover almost 50% of patients with pure ocular myasthenia, which is frequently diagnosed after a 2–3 year course of the disorder, are negative for antibodies against AChR or MuSK in standard assays. However, most of these patients probably have antibodies against a cluster of rapsyn and AChR, which has been demonstrated by using a cell-based assay [13].

The diagnosis of OMG was made based on the presence of relevant symptoms (ptosis and/or diplopia) and either seropositive status of autoantibody, electrophysiological tests or response to therapy. The secondary generalization (gOMG) was defined as the presence of generalized symptoms other than ocular symptoms, including bulbar symptoms and weakness of the jaw, neck, arms and legs.

Surgical procedure

As no consensus has been reached regarding the surgical treatment for patients with OMG, we provided a comprehensive discussion about the pros and cons of surgical and non-surgical treatment for each patient who was contemplating undergoing thymectomy. All patients made their own decision on whether to undergo surgical treatment.

The authors preferred the left-sided 3-trocar approach for robotic thymectomy in patients with non-thymomatous MG [14]. For patients with a suspected thymoma, a unilateral 3-trocar approach for robotic thymectomy was chosen according to the location of the lesion. Typically, an en bloc resection for patients with MG includes the thymic gland, thymoma (if any) and all mediastinal fat (especially anterior mediastinal fat and fat in the pericardiophrenic angles, aortopulmonary window, cervical region, lateral to phrenic nerves and aortocaval groove). Besides, the ‘no touch’ technique was used during the operation for patients with a suspected thymoma. The neurologists of the Integrated Myasthenia Gravis Center helped to administer the postoperative medications for MG.

Follow-up

Patients were followed up by both thoracic surgeons and neurologists every 3 or 6 months according to the requirement of adjusting the medications. Patients who stopped visiting us in the clinic were called for a final follow-up. Based on the response to a questionnaire and current medications in need, the therapeutic effect was assessed according to the Myasthenia Gravis Foundation of America Post-Intervention Status (MGFA-PIS). ‘CSR’ was defined as no symptoms or signs of MG as well as no therapy for MG for at least 1 year. Change in status was also assessed for each patient according to MGFA-PIS.

Statistical analysis

R statistical software 3.5.2 was used to perform the statistical analyses. Continuous variables, if normally distributed, were summarized as means ± standard deviations (SDs) and analysed by independent-samples T-test or paired Student’s t-test; if not, were summarized as median [interquartile range (IQR)] and analysed by Mann–Whitney test or Wilcoxon signed-rank test. Normal distribution was checked by using histogram and calculating skewness. Continuous variables were considered normally distributed if data distribution was sufficiently symmetric and absolute value of skewness was smaller than 0.5. Categorical variables were expressed by numbers of observation and proportions and analysed by χ² test or χ² test combined with Monte Carlo method in case more than 20% of cells have expected frequencies <5.

To reduce the selection biases related to this non-randomized cohort, a propensity score model was used to explore the effect of timing of thymectomy (OMG vs gOMG) on the clinical outcomes of patients with OMG [15]. The propensity score model was conducted using ‘matchit’ package (Daniel Ho et al., 9 January 2018), using a 1:1 matching protocol without replacement, with a caliper width of 0.2 SD of the logit of the propensity score. Standardized difference in mean (SDM) was estimated for all baseline covariates to assess prematch imbalance and postmatch balance. SDM <0.1 for a given covariate was considered a relatively small imbalance. Covariates with an SDM larger than 0.1 after matching were managed by a double adjustment. Patients with missing data were excluded because missing values only accounted for 0.51% of all values. Covariates, which are likely to influence the development of secondary generalization and the clinical outcomes after thymectomy, were included in the propensity score model after a discussion among thoracic surgeons, a neurologist and a statistician: gender, age at onset, anti-AChR antibody status, time from onset to thymectomy and thymic histopathology. We applied this method on all eligible patients with MG and then on the subset of patients with non-thymomatous MG. Once a matched sample was formed, paired comparisons were performed with the use of paired Student’s t-test or Wilcoxon signed-rank test for continuous variables and χ² test or χ² combined with Monte Carlo method for categorical variables.

Survival curves were plotted using the Kaplan–Meier method and comparison was investigated by the log-rank test using ‘Survival’ package (Terry M. Therneau and Thomas Lumley, 1 April 2019). Multivariable Cox proportional hazard model of time to CSR was performed both in the matched whole group and in the matched non-thymomatous group. Variables with a P-value smaller than 0.1 at log-rank test were included for a final multivariable analysis. P-value <0.05 was considered statistically significant.

RESULTS

Patients

Two hundred and forty out of 596 patients with MG presented with purely ocular symptoms at disease onset. After excluding 4 patients who underwent re-thymectomy and 71 patients who were lost to follow-up, 165 patients (66 males and 99 females) were eligible for inclusion. The mean follow-up time was 73.1 ± 46.03 months, 121 patients had a follow-up time more than 3 years and 84 patients more than 5 years. Of these, 73 patients underwent thymectomy before the generalization of OMG (OMG group), 92 patients underwent thymectomy after the generalization of OMG (gOMG group). Not surprisingly, time from symptoms onset to thymectomy was significantly shorter in the OMG group than that in the gOMG group [10 (5–17) vs 14 (8–26) months, P = 0.003]. Table 1 describes the characteristics of both groups.

Clinical outcomes of the matched groups

A propensity score matching was performed to reduce the selection bias between the 2 groups. After matching, a data set including 130 patients (65 per group) was formed. SDMs for all covariates except for gender (SDM = 0.127) were <0.1. There were no significant differences between the 2 groups after matching. Detailed characteristics of the matched groups and the results are shown in Table 1.

CSR was achieved in 39.2% of the matched cohort; there was a significant difference in the percentage of patients reaching CSR between the 2 groups (OMG vs gOMG, 50.8% vs 27.7%; P = 0.007) without a difference in follow-up time (P = 0.579). The median time to achieve CSR was 27 months (IQR 17–41 months), patients in the OMG group [24 (16–34) months] tended to have a shorter time to achieve CSR after thymectomy than patients in the gOMG group [37.5 (IQR 22–57) months] did (P = 0.095). Table 2 describes the clinical outcomes of the 2 matched groups in detail.

We used Kaplan–Meier method and log-rank test to explore the role of timing of thymectomy in the achievement of CSR, resulting in a significant difference between the 2 groups (Fig. 1A, P = 0.0053). The estimated cumulative probabilities of CSR were 0.495 [95% confidence interval (CI) 0.345–0.610] in the OMG group and 0.334 (95% CI 0.176–0.462) in the gOMG group at 5 years after thymectomy. Besides, thymectomy before generalization of OMG (adjusted hazard ratio (HR) 2.27, 95% CI 1.27–4.1; P = 0.006) was still associated with the achievement of CSR after adjusting for gender which had a relatively larger SDM (0.127) after matching (Fig. 1B). Furthermore, Kaplan–Meier curves of CSR in the matched whole cohort stratified by each covariate are detailed in Supplementary Material, Figures S1–S5. Age at onset of 40 years or younger (P = 0.002) and thymic hyperplasia (P = 0.0067) were significantly associated with a higher cumulative probability of CSR after thymectomy. However, thymectomy before generalization remained significant in the matched cohort after adjusting for age at onset and thymic histopathology (Fig. 2, adjusted HR 2.35, 95% CI 1.32–4.19; P = 0.004). Then, we plotted Kaplan–Meier curves of CSR stratified by each covariate in these 2 groups separately (Supplementary Material, Figures S6–S10). Significant differences in cumulative probability of CSR were only found in the OMG group stratified by age at onset (Supplementary Material, Figure S7.1) and thymic histopathology (Supplementary Material, Figure S10.1).

Clinical outcomes of the matched groups with non-thymomatous myasthenia gravis

A propensity score matching with the same covariates was then performed in patients with non-thymomatous OMG (n = 136). We obtained 110 patients (55 per group) after matching and only anti-AChR antibody had an SMD larger than 0.1 in the matched group. Table 3 describes the characteristics of the matched groups and the evaluating results of the matching. CSR was achieved in 42 (38.2%) patients with non-thymomatous MG, 54.5% in the OMG group and 21.8% in the gOMG group (P = 0.000). The detailed clinical outcomes of the matched groups are shown in Table 4.

Kaplan–Meier curves of CSR in both groups were plotted and a statistical difference was identified by log-rank test (Fig. 3A, P = 0.00041). The estimated cumulative CSR rates were 53.5% (95% CI 0.370–0.656) in the OMG group and 28.9% (95% CI 0.131–0.419) in the gOMG group at 5 years after thymectomy. Besides, thymectomy before generalization of OMG (adjusted HR 3.1, 95% CI 1.58–6.1; P < 0.001) was still associated with the achievement of CSR after adjusting for anti-AChR antibody which had a relatively larger SDM (0.172) after matching (Fig. 3B). Likewise, Kaplan–Meier curves of CSR in the matched non-thymomatous cohort stratified by each covariate are also plotted (Supplementary Material, Figure S11–S15). Age at onset of 40 years or younger (P = 0.017), time from symptoms onset to thymectomy <1 year (P = 0.056) and thymic histopathology (P = 0.036) were likely associated with a higher cumulative probability of CSR after thymectomy. However, thymectomy before generalization remained significant in the matched non-thymomatous cohort after adjusting for age at onset, time from symptoms onset to thymectomy and thymic histopathology (Fig. 4, adjusted HR 3.42, 95% CI 1.74–6.71; P < 0.001). Then, we plotted Kaplan–Meier curves of CSR stratified by each covariate in these 2 groups separately (Supplementary Material, Figures S16–S20). Statistical differences in cumulative probability of CSR were only found in the non-thymomatous OMG group stratified by age at onset (Supplementary Material, Figure S17.1) and thymic histopathology (Supplementary Material, Figure S20.1).

DISCUSSION

Thymectomy in ocular myasthenia gravis: a long-existing dilemma

Thymectomy is an effective treatment for patients with early onset generalized MG who are anti-AChR antibody positive [16–18]. For patients with OMG, however, only a few neurologists and thoracic surgeons consider thymectomy to be a reasonable treatment. Besides, many patients with OMG would compromise quality of life rather than undergoing traumatic thymectomy. Apart from that, data focusing on the role of thymectomy in patients with OMG are also limited due to the relative rarity of the disease. In general, the baseline characteristics of our cohort were in line with that in previous reports [11, 19–22]. It is worth noting that patients in our study seemed to have an older age at disease onset than those from previous reports [11, 19, 22]. Accordingly, we did find a lower percentage of thymic hyperplasia in our cohort, compared to those previously reported [11, 19, 21, 22]. As it is difficult to design a prospective randomized trial to deal with this dilemma, retrospective evidence should be appreciated to shed light on the role of thymectomy in OMG.

Thymectomy in ocular myasthenia gravis: an evidence-based approach?

Previous studies have shown that the use of immunosuppressive medications in patients with OMG may improve ocular symptoms and reduce the risk of secondary generalization [3–5]; however, the adverse effects are well known [6, 23, 24]. Prospective evidence has shown that thymectomy is an effective approach to improve symptoms and reduce the use of prednisone in patients with generalized MG [8], which might also benefit patients with OMG in the same manner. Besides, some retrospective studies focusing on patients with OMG have suggested the efficacy of thymectomy in terms of inducing remission [11, 19–22, 25]. The range of the reported remission rate is quite wide, from 16.7% to 64%, possibly due to different surgical approaches, different definitions of remission and different lengths of follow-up periods [11, 19–22, 25]. Recently, a meta-analysis has summarized the evidence regarding the efficacy of thymectomy in patients with non-thymomatous OMG, concluding that thymectomy is a beneficial treatment with a pooled remission rate of 0.5074 [12]. In 2013, Mineo and Ambrogi [11] conducted a direct comparison of the outcomes between surgical patients and non-surgical controls, concluding thymectomy being a marginal predictor of remission. In our study, the estimated CSR rate at 5 years was 49.5% in patients with OMG and 53.5% in patients with non-thymomatous OMG, comparable to that (53%) reported by Mineo and Ambrogi [11]. On the other hand, OMG is, more frequently, underestimated due to the so-called ‘mildness of symptoms’, which can be profoundly visually disabling and considerably impair the quality of life. Therefore, considering the efficacy of thymectomy, the use of thymectomy, especially minimally invasive thymectomy, in patients with non-thymomatous OMG should not be controversial. Although there is a lack of high-level evidence comparing surgical treatment and non-surgical treatment in this population, previous retrospective evidence the well-known immunological role of thymectomy in MG and the Myasthenia Gravis Thymectomy Trial (MGTX) have made the surgery decision in OMG straightforward.

Thymectomy in ocular myasthenia gravis: until the time of generalization?

The main controversy is whether thymectomy before generalization of OMG can reduce the risk of generalization and result in better outcome. Previous surgical series reported a rate of generalization between 0% and 4.5% in patients with OMG [11, 19–22, 25], which seems to be lower than that (9.4–73.3%) reported in patients with non-surgical treatment [3, 26, 27]. Given the variable rate of generalization in prior studies, however, it is still difficult to draw a solid conclusion that surgical treatment exerts a greater effect on preventing generalization of OMG, compared to non-surgical treatment. On the other hand, would thymectomy result in overtreatment for patients who would reach a spontaneous remission and those who would never develop a secondary generalization? Even if about 20% patients with MG can achieve spontaneous remission, usually within 1 year after symptom onset, most of them will relapse with a median duration of remission lasting <1 year [2, 5, 28]. Regarding patients with OMG who would never develop a secondary generalization, most of them would need immunosuppressive medications because cholinesterase inhibitors alone does not usually relieve the symptoms [29]. However, there is uncertainty whether waiting to treat until the time of generalization brings any disadvantages. To approach this remaining uncertainty, we compared the outcomes between the OMG group and the gOMG group. Our propensity score analysis suggested that delaying thymectomy until the time of generalization results in a lower rate of CSR compared to thymectomy before generalization.

On the other hand, early thymectomy and mild disease classification (MGFA I–II) are frequently reported to be predictors of remission after thymectomy [30]. Regarding thymectomy in OMG, several retrospective studies indicated that only early thymectomy is likely to result in better clinical outcome [11, 19, 22]. In our cohort, thymectomy before generalization, age of 40 years or younger at disease onset and thymic hyperplasia were found to be potential predictors of CSR after thymectomy. This is possibly due to the different patient cohorts and small sample size. Besides, early thymectomy and thymectomy before generalization actually support each other to some extent. More importantly, previous studies are more likely to include patients with a younger age at disease onset and those with thymic hyperplasia, which makes it difficult to find a difference in cumulative probability of CSR in the subgroups of age at onset and thymic histopathology [11, 19, 22]. It is worth mentioning that although patients with thymic hyperplasia were more likely to achieve CSR after thymectomy, patients with a normal or involuted thymus in the matched non-thymomatous cohort were also doing quite well with an accumulative probability of CSR of 36.1% (95% CI 21.0–48.3) at 5 years. Altogether, we consider thymectomy as a reasonable and effective treatment for patients with OMG.

Limitations

The retrospective nature of our study is a major limitation. Although we used a propensity score model to reduce selection bias between the 2 groups, unadjusted confounder may still influence treatment selection. To enhance our matching process and to include as many patients as possible in the matched groups, we included only variables hypothesized to be associated with both groups (OMG vs gOMG) and outcome (CSR) judged from a clinical point of view. In order to increase the granularity of the propensity scores, we considered to use interactions and polynomial terms in the matching process. We only identified an interaction between ‘Age at disease onset’ and ‘Thymic histopathology’ in the non-thymomatous group. After including this interaction into the matching process, we not only lost patients (55 vs 51 per group) but also not observed a higher granularity of the propensity scores (data not shown). Therefore, we decided to not include interactions and polynomial terms in the final matching process. Besides, only 13 out of 73 patients in the OMG group generalized after surgery, whereas all patients in the gOMG group generalized after a certain period with purely ocular symptoms, which contributes towards heterogeneity between the 2 groups. Small sample size is another limitation of our study. However, this is a monocentric observational study from a high-volume centre of thymectomy for MG; an experienced surgeon operated on almost all patients through an extended robotic approach. These data might be useful for planning a randomized controlled trial determining the efficacy of minimally invasive thymectomy in OMG.

CONCLUSIONS

Our results suggest that thymectomy before generalization of OMG is likely to result in a higher rate of CSR than thymectomy after generalization of OMG. Hence, thymectomy may be a first-choice treatment for patients with OMG.

Presented at the 27th European Conference on General Thoracic Surgery, Dublin, Ireland, 9–12 June 2019.

ACKNOWLEDGEMENTS

The authors are thankful to Ulrike Grittner, Department of Biostatistics and Clinical Epidemiology, Charité Universitätsmedizin Berlin, for her assistance in statistical analyses. F.L. thanks China Scholarship Council for the financial support during his study in Germany.

Conflict of interest: Dr Meisel reports personal fees from Alexion, Bristol Myers Squipp, Grifols and Hormosan, and research grants from Alexion and Octapharma outside the submitted work. Other authors have nothing to disclose.

Author contributions

Feng Li: Conceptualization; Investigation; Methodology; Project administration; Resources; Software; Visualization; Writing - Original Draft; Writing - Review & Editing. Zhongmin Li: Investigation; Methodology; Resources; Software; Writing - Original Draft; Writing - Review & Editing. Yanli Chen: Formal analysis; Investigation; Methodology; Resources; Software; Writing - Original Draft. Gero Bauer: Data curation; Investigation; Methodology; Resources; Software; Writing - Original Draft. Deniz Uluk: Investigation; Methodology; Resources; Software; Writing - Original Draft. Aron Elsner: Investigation; Methodology; Resources; Software; Supervision; Writing - Original Draft. Marc Swierzy: Investigation; Methodology; Resources; Software; Supervision; Writing - Original Draft. Mahmoud Ismail: Investigation; Methodology; Resources; Supervision; Writing - Original Draft. Andreas Meisel: Formal analysis; Investigation; Methodology; Project administration; Resources; Supervision; Writing - Original Draft; Writing - Review & Editing. Jens-C. Rückert: Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Project administration; Resources; Software; Supervision; Validation; Visualization; Writing - Original Draft; Writing - Review & Editing.

REFERENCES

ABBREVIATIONS

ABBREVIATIONS
 
• AChR

  Acetylcholinesterase receptor

 
• CI

  Confidence interval

 
• CSR

  Complete stable remission

 
• HR

  Hazard ratio

 
• IQR

  Interquartile range

 
• MG

  Myasthenia gravis

 
• MGFA-PIS

  Myasthenia Gravis Foundation of America Post-Intervention Status

 
• OMG

  Ocular myasthenia gravis

 
• SD

  Standard deviation

 
• SDM

  Standardized difference in mean