Clinical usefulness of prethymectomy plasmapheresis in patients with myasthenia gravis: a systematic review and meta-analysis

Abstract

OBJECTIVES

Our goal was to evaluate, through a systematic review, the efficacy of plasmapheresis in the preoperative preparation of the patient for a thymectomy for the treatment of myasthenia gravis.

METHODS

MEDLINE, Embase, LILACS, Scopus and CENTRAL databases were searched. The following outcomes were evaluated: myasthenic crisis, mortality, pneumonia, bleeding, use of mechanical ventilation, length of hospital stay and intensive care unit (ICU) stay. RevMan 5.3 software provided by the Cochrane Collaboration was used for the meta-analysis.

RESULTS

The total number of patients evaluated in the 7 included studies was 360. Plasmapheresis during the preoperative period did not decrease the myasthenic crisis [risk ratio (RR) 0.36, 95% confidence interval (CI) 0.08–1.66; I² = 44%; 5 studies, 243 patients]. There was also no change in the mortality rate (RR 0.7, 95% CI 0.11–4.62; I² = 0%; 3 studies, 172 patients) or pneumonia cases (RR 0.28, 95% CI 0.07–1.09; I² = 27%; 5 studies, 272 patients). Bleeding was greater in patients who underwent plasmapheresis (mean difference 34.34 ml; 95% CI 24.93–43.75; I² = 0%). We evaluated the following outcomes: need for mechanical ventilation, hospital stay, ICU stay and mechanical ventilation, but these outcomes were not adequate to perform the meta-analysis due to the high heterogeneity among the studies. Subgroup analysis showed that plasmapheresis performed during the preoperative period in patients with severe disease (Osserman III and IV) decreased the myasthenic crisis postoperatively (RR 0.12, 95% CI 0.02–0.65; I² = 63%).

CONCLUSIONS

Plasmapheresis may reduce the myasthenic crisis during the postoperative period in patients with severe disease but may produce little or no difference in patients with mild clinical expression of the disease.

Myasthenia gravis, Thymectomy, Plasmapheresis, Systematic review

INTRODUCTION

Myasthenia gravis (MG) is an autoimmune disease in which antibodies are produced against acetylcholine receptors, with damage to neuromuscular transmission. Consequently, this process generates the main clinical symptom of the disease, muscle weakness, which increases with the use of muscles and tends to decrease with rest [1].

The choice of treatment is individualized because it depends on the need for urgency in dealing with weakness. The removal of autoantibodies by plasmapheresis is applied to patients with severe MG, usually during a myasthenic crisis, or during the preoperative preparation of the patient for a thymectomy for the treatment of MG [2]. However, controversies exist regarding the use of plasmapheresis in the preoperative preparation for a thymectomy, and those who advocate its application believe that it reduces the length of mechanical ventilation (MV) and time in the intensive care unit (ICU) [3] in addition to reducing the risk of a postoperative myasthenic crisis [4]. Those who defend the non-use of the preoperative preparation argue that there may be an increase in the possibilities for complications related both to the use of the central catheter [5] and to the plasmapheresis itself [6, 7]. In addition, some studies have not shown relevance in reducing the risk of myasthenic crisis or mortality during the postoperative period in patients undergoing plasmapheresis [3]. The cost of plasmapheresis, which can range from US$680 to US$1400 [6], prevents it from being widely used worldwide.

The aim of this study was to verify, through a systematic review, whether plasmapheresis conducted during the period before a thymectomy in patients with MG is effective in reducing complications during the postoperative period.

METHODS

Inclusion criteria

Types of studies: We identified experimental or observational studies that compared the evolution during the postoperative period of patients who had a thymectomy with or without plasmapheresis in the preoperative period.

Participants: Patients included those with MG and candidates for thymectomy.

Interventions: Patients who had surgical treatment with plasmapheresis during the preoperative period comprised the study group.

Controls: Those who had surgical treatment without plasmapheresis during the preoperative period comprised the control group.

Outcomes evaluated

The primary outcome was a myasthenic crisis during the postoperative period. Secondary outcomes included death, pneumonia, bleeding, need for MV, lengths of stays in the hospital and ICU and length of MV.

Search methods

We searched the following databases: MEDLINE via PubMed, LILACS (Latin American and Caribbean Literature in Health Sciences), Scopus, Embase, CENTRAL (Cochrane Central Register of Controlled Trials) and additional sources of published and unpublished trials. We also searched the Experimental Clinical Trials database (http://clinicaltrials.gov) for data from any ongoing studies. The following search strategy was used in the descriptor Medical Subject Headings Section of PubMed/MEDLINE, Web of Science and Cochrane Library: (Myasthenia Gravis or Myasthenia Gravis, Ocular or Ocular Myasthenia Gravis or Myasthenia Gravis, Generalized or Generalized Myasthenia Gravis) and (Plasmapheresis or Plasmaphereses) and (Thymectomy or Thymectomies). In Embase: (Myasthenia Gravis) or (Myasthenia Gravis, Ocular) or (Ocular Myasthenia Gravis) or (Myasthenia Gravis, Generalized) or (Generalized Myasthenia). In LILACS: (Myasthenia Gravis or Miastenia Gravis or Miastenia Grave) and (Plasmapheresis or Plasmaféresis or Plasmaferese) and (Thymectomy or Timectomia).

Data collection and analysis

Selection of studies

Duplicate items were initially deleted. Included in this review were experimental or observational studies that compared the evolution during the postoperative period of patients undergoing thymectomy with and without plasmapheresis in the preoperative period. To do so, the studies were grouped and evaluated by their titles and abstracts by 2 researchers (T.A.R. and D.C.C.), independently and blindly, to remove the irrelevant articles; the full text of the potentially relevant studies was reviewed in a second step. To resolve any disagreements, a discussion was held; if the discussion did not result in consensus, the opinion of the third author was decisive.

Data extraction and management

A protocol was used to extract the data from the studies eligible for this review. The resulting summary contained the basic characteristics of the participants, interventions, the number of participants included, outcome measures and length of follow-up. Two reviewers (T.A.R. and D.C.C.) independently extracted all data from the interventions studied. Any disagreements were resolved by discussion.

Bias risk assessments

Two authors (T.A.R. and A.J.M.C.) independently assessed the risk of bias in each study and any disagreements were resolved by consensus or with the consultation of a third party. For the bias risk analysis of the experimental studies, we used the ‘Risk of bias’ tool of the Cochrane Collaboration [8]. For the observational studies, we used the Robins-I tool [9].

Measures of treatment effect

Dichotomic outcomes were presented as risk ratios (RRs) with their respective 95% confidence intervals (CIs). For the continuous outcomes, the mean differences with the corresponding 95% CIs were analysed.

Data synthesis

For outcomes with heterogeneity equal to 0, a fixed effect model was used, according to the Mantel–Haenszel meta-analysis method [10]. For the outcomes with heterogeneity >0, the random effects model was used. The analysis was done with the Review Manager 5.3 program. The analysed outcomes were considered with a 95% CI. The results were summarized using the forest plot in which each horizontal line represents an included study. The estimated effect is represented by a square and the size of the square corresponds to the weight of the study in question. The combined effect estimate is represented by a diamond located at the bottom of the chart. In order to quantify the inconsistencies of the studies used in the meta-analysis, the heterogeneity test I² = [(Q − df)/Q] × 100% was used, where Q is the χ²and df is the degree of freedom. Substantial heterogeneity was considered when I² >75%. In this case, variability in estimates and effect may be a result of heterogeneity rather than sampling error [11], and it is inappropriate to perform the meta-analysis.

Subgroup analysis

The plan was to analyse separately young women with MG for <2 years and elderly men with the disease for more than 2 years. The plan also included analysing subgroups with less advanced disease (Osserman II) and more advanced disease (Osserman III and IV).

RESULTS

The searches of the electronic databases to detect papers that met the objectives of the study were carried out in November 2018 and yielded 384 articles from PubMed (136), Embase (141), Web of Science (86), LILACS (18) and Cochrane (3). After the manual deletion of duplicates, 315 articles remained. We verified, after an initial analysis of the respective titles and abstracts, that 302 articles did not warrant a complete reading of the text. Thirteen studies were chosen for the complete reading of their texts. Six were excluded for various reasons; the 7 remaining studies were included in the meta-analysis (Fig. 1).

Figure 1:

Eligibility flow chart for articles for meta-analysis.

We were unable to locate 2 of the 6 studies that were excluded [12, 13], despite intense efforts. We tried unsuccessfully to acquire these articles in several ways: purchase, through the Program of Bibliographic Switching and through contact with the authors. Of the 4 other studies that were excluded, Alipour-Faz et al. [14] compared the IG results with plasmapheresis; El-Bawab et al. [15] evaluated the indication for plasmapheresis in the preoperative period and not its effects in the postoperative period; Sigal et al. [16] used plasmapheresis in the preoperative and postoperative periods but only emphasized outcomes related to muscle weakness in the late postoperative period; and Dias-Tosta and Morato-Fernandes [17] presented a series of cases comparing different possibilities of treatment of MG but did not present the outcomes proposed in this systematic review.

The 7 studies that met the previously defined inclusion criteria involved 360 patients operated on between 1975 and 2011. There were 2 prospective randomized and 5 observational studies. Five studies reported the number of plasmapheresis sessions. Two studies did not report the length of the postoperative period follow-up, which ranged from 12 to 125 months. All studies stratified their patients into the Osserman classification (Table 1).

Table 1:

Main features of included studies

Author (year of publication)		Intervention/ control	d’Empaire et al. [18] (1985)	Iváñez et al. [19] (1994)	Seggia et al. [20] (1995)	Nagayasu et al. [21] (2005)	Kamel and Essa [3] (2009)	Nishtar et al. [22] (2012)	Saeteng et al. [6] (2013)
Country			USA	Spain	Brazil	Japan	Egypt	Pakistan	Thailand
Type of study			RTP	RCT	RTP	RTP	RCT	RTP	RTP
Centre (uni-/multicentric)			Uni	Uni	Uni	Uni	Uni	Multi	Uni
Period of study		PPG	1975–1983	NA	1984–1993	1980–1997	2004–2008	2002–2009	2005–2011
		NPPG			1975–1983
Postoperative follow-up time (months)		PPG	NA	12	12	100.2 ± 41.2	19.8 ± 7.5	12	NA
		NPPG				125.1 ± 77.5	18.9 ± 10.2
Participants		PPG	11	12	40	19	19	20	33
		NPPG	26	12	40	32	16	27	53
Osserman, n (%)	I	PPG	0	0	0	0	0	0	2 (6)
	IIA		0	0	0	8 (42.1)	6 (31.6)	0	6 (18.2)
	IIB		11 (100)	12 (100)	19 (47.5)	11 (57.9)		8 (40)	21 (63.3)
	III		0	0	16 (40)	0	9 (47.4)	8 (40)	4 (12.1)
	IV		0	0	5 (12.5)	0	4 (21.1)	4 (20)	0
	I	NPPG	0	0	0	0	0	0	5 (9.4)
	IIA		0	0	0	20 (62.5)	8 (50)	0	14 (25.9)
	IIB		26 (100)	12 (100)	18 (45)	12 (37.5)		17 (63)	31 (58.4)
	III		0	0	20 (50)	0	6 (37.5)	5 (18.5)	3 (5.7)
	IV		0	0	2 (5)	0	2 (12.5)	5 (18.5)	0

NA: not available; NPPG: no plasmapheresis in the preoperative period group; PPG: plasmapheresis during the preoperative period group; RCT: randomized clinical trial; RTP: retrospective.

Type of intervention: All 7 studies compared the postoperative evolution of thymectomized patients with and without plasmapheresis in the preoperative period.

Types of outcomes: Five of the 7 studies evaluated the prevalence of a myasthenic crisis postoperatively; 3 evaluated the length of stay in the ICU; 2 assessed the length of the hospital stay; 3 reported the number of deaths; 2 evaluated bleeding; 3 evaluated the need for postoperative period MV; 5 evaluated the incidence of pneumonia; and 2 evaluated the length of MV in the postoperative period.

Risk of bias: For the case-control studies, the risk analysis of biases was performed using the Robins-I tool [9] (Table 2). The overall risk of bias in each study is the risk attributed to the domain with the highest risk of bias. To assess the risk of bias in randomized controlled trials, we used Cochrane Collaboration criteria for the development of systematic reviews of intervention (Table 3).

Table 2:

Risk of bias in case-control studies

Bias domains	Author (year of publication)
	d’Empaire et al. [18] (1985)	Seggia et al. [20] (1995)	Nagayasu et al. [21] (2005)	Nishtar et al. [22] (2012)	Saeteng et al. [6] (2013)
Confounding	Low	Low	Serious	Low	Low
Selection of participants	Serious	Serious	Critical	Serious	Serious
Classification of interventions	Low	Moderate	Low	Low	Low
Deviations from intended interventions	Moderate	Low	Low	Low	Low
Missing data	Low	Low	Low	Low	Low
Measurement of outcomes	Low	Moderate	Low	Low	Low
Selection of the reported result	Low	Low	Low	Low	Low
Overall bias	Serious	Serious	Critical	Serious	Serious

Table 3:

Risk of bias in randomized controlled trials

Author (year)	Domain of biases
	Random sequence generation	Allocation concealment	Blinding of participants and professionals	Outcome evaluator blinding	Incomplete outcomes	Report of selective outcome
Iváñez et al. [19] (1994)	Uncertain	Uncertain	Uncertain	Uncertain	Low	Uncertain
Kamel and Essa [3] (2009)	Uncertain	Uncertain	Uncertain	Uncertain	Low	Uncertain

Effects of intervention

Primary outcome

Five studies involving 243 people evaluated myasthenic crisis in the postoperative period. Plasmapheresis during the preoperative period did not decrease the myasthenic crisis in the postoperative period (RR 0.36, 95% CI 0.08–1.66; I² = 44%; P = 0.13; Fig. 2A).

Figure 2:

(A) Forest plot of the risk ratio (RR) for myasthenic crisis in the postoperative period of patients who underwent PPG compared to those who did not. Meta-analysis applying random effect in 5 studies (RR 0.36, 95% CI 0.08–1.66; I² = 44%; P = 0.13). (B) Forest plot of the RR for death of patients who underwent plasmapheresis compared to those who did not: meta-analysis applying fixed effect in 3 studies (RR 0.70, 95% CI 0.11–4.62; I² = 0%; P = 0.44). (C) Forest plot of the RR for pneumonia in patients who underwent plasmapheresis compared to those who did not: meta-analysis applying random effect in 5 studies (RR 0.28, 95% CI 0.07–1.09; I² = 27%; P = 0.25). (D) Forest plot of mean bleeding difference: patients who underwent plasmapheresis bled about 34 ml more than those who did not (DM + 34.34 ml; 95% CI 24.93–43.75; I² = 0%; P = 0.57). CI: confidence interval; df: degree of freedom; M-H: Mantel–Haenszel; NPPG: control: no plasmapheresis during the preoperative period; PPG: experimental: plasmapheresis during the preoperative period; SD: standard deviation.

Secondary outcomes

Three studies involving 172 people evaluated the mortality rate. Plasmapheresis during the preoperative period did not alter the mortality rate (RR 0.7, 95% CI 0.11–4.62; I² = 0%; P = 0.44; Fig. 2B).

Five studies involving 272 people evaluated pneumonia. Plasmapheresis during the preoperative period did not reduce the risk of pneumonia in the postoperative period (RR 0.28, 95% CI 0.07–1.09; I² = 27%; P = 0.25), but plasmapheresis tended to reduce this risk (Fig. 2C).

Two studies involving 121 people evaluated bleeding. Plasmapheresis during the preoperative period increased bleeding in the patients who had it compared to those in the control group (mean difference 34.34 ml; 95% CI 24.93–43.75; I² = 0%; P = 0.57; Fig. 2D).

Three studies with 213 people evaluated the need for MV in postoperative period. Due to the high degree of heterogeneity, the meta-analysis for this outcome was considered inappropriate. In 2 studies, the need for MV was lower in the group that underwent plasmapheresis, but in 1 study there was no difference between the groups (Fig. 3A).

Figure 3:

(A) Forest plot of the risk ratio for the need for mechanical ventilation in the postoperative period of patients who underwent plasmapheresis in the preoperative period compared to those who did not: meta-analysis not performed due to high heterogeneity (I² = 80%). (B) Forest plot of the mean difference in hospitalization time between patients who underwent plasmapheresis and those who did not: meta-analysis not performed due to high heterogeneity (I² = 87%). (C) Forest plot of the mean difference in length of stay in the intensive care unit between patients who underwent plasmapheresis and those who did not: meta-analysis not performed due to high heterogeneity (I² = 98%). (D) Forest plot of the mean mechanical ventilation time difference between patients who underwent plasmapheresis and those who did not: meta-analysis not performed due to high heterogeneity (I² = 83%). CI: confidence interval; df: degree of freedom; M-H: Mantel–Haenszel; NPPG: control: no plasmapheresis during the preoperative period; PPG: experimental: plasmapheresis during the preoperative period; SD: standard deviation.

Two studies involving 121 people evaluated length of hospital stay. Due to the high degree of heterogeneity, a meta-analysis for this outcome was considered inappropriate. In 1 study, there was a shorter hospital stay in patients who had plasmapheresis, but in the other there was no difference (Fig. 3B).

Three studies involving 158 people evaluated length of stay in the ICU. Due to the high degree of heterogeneity, the meta-analysis for this outcome was considered inappropriate. In 2 of these studies, the patients who underwent plasmapheresis had a shorter stay in the ICU, but in the other there was no difference (Fig. 3C).

Two studies involving 72 people evaluated MV time. Due to the high degree of heterogeneity, the meta-analysis for this outcome was considered inappropriate. In both studies, patients who had plasmapheresis had shorter MV times (Fig. 3D).

Subgroup analysis

It was not possible to separate young women with MG for <2 years and elderly men with the disease for more than 2 years. Because the studies did not separate patients’ progress according to the Osserman classification, the subgroups were divided as follows: subgroup 1 (less advanced disease) included studies with 80–100% of patients in stage II; subgroup 2 (more advanced disease) included studies with 40% or more of the patients in stages III and IV. For the subgroups, only a meta-analysis of the primary outcomes was performed because of the limited number of studies that separated the more severe patients from the less severe ones.

Three studies involving 161 people evaluated myasthenic crisis in the postoperative period in subgroup 1. Plasmapheresis during the preoperative period did not decrease the myasthenic crisis in the postoperative period in this subgroup (RR 0.93, 95% CI 0.07–12.14; I² = 63%; P = 0.07; Fig. 4A).

Figure 4:

(A) Forest plot of the risk ratio (RR) for myasthenic crisis in the postoperative period of the patients who underwent plasmapheresis in the preoperative period compared to those who did not in the subgroup with more than 80% of the patients in the Osserman II group: meta-analysis applying random effect in 3 studies (RR 0.93, 95% CI 0.07–12.14; I² = 63%; P = 0.07). (B) Forest plot of the RR for myasthenic crisis in the postoperative period of patients who underwent plasmapheresis in the preoperative period compared to those who did not in the subgroup with more than 40% of patients in the Osserman III group: meta-analysis applying fixed effect in 2 studies (RR 0.12, 95% CI 0.02–0.65; I² = 0%; P = 0.51). CI: confidence interval; df: degree of freedom; M-H: Mantel–Haenszel; NPPG: control: no plasmapheresis during the preoperative period; PPG: experimental: plasmapheresis during the preoperative period.

Two studies involving 82 people evaluated myasthenic crisis in the postoperative period in subgroup 2. Plasmapheresis during the preoperative period decreased the myasthenic crisis in the postoperative period in this subgroup (RR 0.12, 95% CI 0.02–0.65; I² = 0%; P = 0.51; Fig. 4B).

DISCUSSION

There is no consensus in the literature concerning the use of plasmapheresis during the preoperative period before a thymectomy for the treatment of MG. Many authors state that reducing the level of autoantibodies in the circulation using plasmapheresis leads to better results in the postoperative period in relation to the occurrence of myasthenic crises and other complications [21]. In addition, performing plasmapheresis during the preoperative period could help in the differential diagnosis of myasthenic crisis and postoperative cholinergic crisis. Because both have similar symptoms and antagonistic treatments, the probability of having a myasthenic crisis would be reduced given that plasmapheresis was performed preoperatively [23]. Other authors emphasize that plasmapheresis is costly, does not result in effective improvement of these parameters and leads to complications that may represent a threat to life, due to the procedure itself and to the implantation of another central venous catheter for the procedure [7].

Five of the 7 included studies evaluated the myasthenic crisis in the postoperative period. Only 1 study demonstrated the efficacy of plasmapheresis in reducing the crisis [21], but in that study, plasmapheresis was used in the best patients, that is, in those who exhibited symptoms for a shorter time. Thus, we performed a sensitivity test, withdrawing this study and obtaining a combination with the rate of risk approaching even closer to the nullity line. In the other studies, there was no difference between the 2 groups. According to the meta-analysis, plasmapheresis did not reduce the risk of a myasthenic crisis postoperatively. In the analysis of subgroups of patients with more advanced disease, we noted a protective effect of plasmapheresis, i.e. it reduced the risk of a myasthenic crisis in these patients. But the same result was not observed in the subgroup of patients with less advanced disease in whom plasmapheresis did not alter the risk of myasthenic crisis.

Plasmapheresis did not reduce pneumonia in the postoperative period, but there was a tendency towards protection against its occurrence. Some authors have suggested that a greater number of infections could occur after plasmapheresis because of the decrease in the concentration of antibodies, including antibodies against microorganisms [24]. Only 1 study [20] showed that pneumonia occurred more frequently in patients who did not undergo plasmapheresis; the other studies showed no difference between the 2 groups. There were no prospective studies on the frequency or severity of any MG-related infection before or after thymectomy [25].

We noted a significant increase in the average amount of blood loss during surgery in the group undergoing plasmapheresis. Only 2 studies presented this outcome [3, 6]; only 1 of them noted this result [6], which was highly weighted in the meta-analysis. The greater amount of bleeding in patients in the group who had plasmapheresis during the preoperative period could be explained by the use of anticoagulants, which are used in plasmapheresis, such as sodium citrate [7], but patients who did not undergo plasmapheresis lost 34 ml of blood less than those who did, which is an insignificant amount for haemodynamic balance and, in theory, is not clinically significant.

For some outcomes, such as MV time and ICU and hospital stays, it was not possible to perform the meta-analysis due to the high heterogeneity among the studies. Nishtar et al. [22] and Seggia et al. [20] showed that plasmapheresis reduced the need for MV, but Saeteng et al. [6] found no difference between the groups.

Most of the studies in this review were case-control studies, which are subject to a high risk of bias. However, some characteristics of these studies may reduce this risk, such as the fact that the control group was selected from the same population as the intervention group. Other studies may have selection biases; in 1 study [6], the group with plasmapheresis during the preoperative period comprised patients who had more myasthenic crises prior to surgery than patients in the group who did not have plasmapheresis during the preoperative period. In another study [22], those with a higher risk of MV in the postoperative period were selected for the group with plasmapheresis during the preoperative period. In the study by Nagayasu et al. [21], the patients in the group who did not have plasmapheresis consisted of patients who refused to have it or who had contraindications, such as duration of infection or use of anticoagulants. In addition, we observed that the group who had plasmapheresis during the preoperative period comprised patients with more recent symptoms than those in the group that did not have plasmapheresis during the preoperative period, which represents a critical selection bias in the study. The study by Seggia et al. [20] presented heterogeneous groups because they were followed up in different decades. In addition, we found an important heterogeneity in the number of plasmapheresis sessions among studies in this review. However, the fact that most of them were retrospective decreased the detection bias because the results were taken from previously written medical records. In the 2 experimental studies [3, 19], the authors did not care about the description of the method of randomization and did not report if there was blinding of the participants or the evaluators.

The difficulty of randomization and double-blind studies of surgical patients has been widely debated in the literature. Although we found biases in the papers included in this review, the final interpretation of the results should be taken into account by the physician, because it is the best evidence to date on the outcomes addressed.

CONCLUSIONS

Plasmapheresis during the preoperative period prior to a thymectomy may reduce myasthenic crisis postoperatively in patients with more advanced disease (Osserman III and IV) but may make little or no difference in patients with less advanced disease (Osserman II).

Implications for the research

Further good quality clinical trials should be performed to conclude whether or not we should perform plasmapheresis in the period prior to a thymectomy in the patient with MG.

Implications for practice

In the patient with MG, a thymectomy could be performed without plasmapheresis during the preoperative period. However, due to the evidence provided only by studies with a low level of evidence, it is prudent to perform plasmapheresis in patients with advanced disease (Osserman III and IV).

ACKNOWLEDGEMENTS

The authors express their acknowledgment to the authors of the primary studies included in the present review and the patients who participated in the clinical trials.

Conflict of interest: none declared.