Immunotherapy for IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathies

See: Summary of findings for the main comparison Rituximab versus placebo for treating IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathy

Summary of findings for the main comparison. Rituximab versus placebo for treating IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathy

Should rituximab versus placebo be used for IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathies?
Patient or population: people with IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathies Setting: hospital and outpatient treatment centres Intervention: rituximab Comparison: placebo
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)	Comments
	Risk with placebo	Risk with rituximab
Number of participants improved in disability Assessed with: INCAT score or INCAT Leg Disability Score Follow‐up: range 8 to 12 months¹	Study population		RR 3.51 (1.30 to 9.45)	73 (2 RCTs)	⊕⊕⊝⊝ low²	Statistically significant effect in meta‐analysis
	74 per 1000	260 per 1000 (96 to 700)
Mean improvement in disability Assessed with: INCAT score or INCAT leg disability score Scale from: 0 to 12 in INCAT and 0 to 7 in INCAT leg disability. Follow‐up: range 8 to 12 months¹	The mean improvement in INCAT score (see text) at 8 to 12 months was ‐0.18	The mean improvement in INCAT score (see text) at 8 to 12 months in the intervention group was 0.45 lower (0.85 lower to 0.05 lower)	‐	73 (2 RCTs)	⊕⊕⊝⊝ low²	Statistically significant improvement versus placebo but probably less than MCID
Improvement in 10‐metre walk time at 8 to 12 months¹	The mean improvement in 10‐metre walk time at 8 to 12 months was 0.14 seconds	The mean improvement in 10‐metre walk time at 8 to 12 months in the intervention group was 0.35 seconds more (1.89 more to 1.19 less)	‐	68 (2 RCTs)	⊕⊕⊕⊝ moderate⁴	MCID for 10‐metre walk approximately 0.4 seconds. Borders on clinically significant improvement, but wide CIs
Participant subjective impression of change stable or improved at 8 to 12 months Assessed with: VAS from: 0 to 10 Follow‐up: mean 8 months¹	Study population		RR 1.86 (1.27 to 2.71)	70 (2 RCTs)	⊕⊕⊝⊝ low³	Patient global impression of change improved similarly in both studies in a time‐dependent manner (see data tables)
	447 per 1000	832 per 1000 (568 to 1000)
Change in serum IgM paraprotein concentration 8 months after treatment¹	The mean change in IgM level 8 months after treatment was 32.3 mg/L	The mean change in IgM level 8 months after treatment in the intervention group was 287.7 mg/L lower (328.98 lower to 244.42 lower)	‐	26 (1 RCT)	⊕⊕⊕⊝ moderate⁴	An unsurprising reduction in IgM in the rituximab‐treated group
Any adverse event	Study population		RR 1.18 (0.84 to 1.66)	80 (2 RCTs)	⊕⊝⊝⊝ very low⁵	No statistically significant difference in adverse effects. Serious adverse effects too few to make comment. Consistency of adverse event collection always suspect
	561 per 1000	662 per 1000 (471 to 931)
*The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; IgM: immunoglobulin M; INCAT score: Inflammatory Neuropathy Cause and Treatment score;MCID: minimum clinically important difference; RR: risk ratio; VAS: visual analogue scale
GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect
¹The outcomes in the rituximab study were recorded at between 8 and 12 months, which is rational for the treatment of a disease with rituximab. Thus we have reported this interval for this intervention (see text). ²Downgraded twice for imprecision (small underpowered studies) and unresponsive outcome measures. ³Downgraded twice for imprecision (small underpowered studies) and indirectness (patient impression of change). ⁴Downgraded once for imprecision (small underpowered studies). ⁵Downgraded three times for imprecision (small studies and small even numbers), trial design and inconsistent adverse event recording.

Background

Up to 10% of people with a peripheral neuropathy which is not secondary to another primary illness have a paraprotein in the serum (Kelly 1981). A paraprotein is an abnormal protein produced by bone marrow cells and may belong to one of the three classes of immunoglobulin (Ig), IgG, IgA or IgM. The majority of paraproteins in people with peripheral neuropathy are IgM, produced by an initially benign condition of blood cells called monoclonal gammopathy of undetermined significance (MGUS). Of this group of paraproteins, 50% react with the CD57/HNK‐1 carbohydrate epitope found on myelin‐associated glycoprotein (MAG). This epitope is also found on other peripheral nerve myelin molecules (myelin protein zero, peripheral myelin protein 22, sulphated glucuronyl paragloboside (SGPG), sulphated glucuronyl lactosaminyl paragloboside (SGLPG) and others) but the antibodies are commonly referred to as anti‐MAG.

Description of the condition

The clinical neuropathy associated with anti‐MAG antibodies is usually relatively phenotypically homogeneous, and consists of a slowly progressive distal sensorimotor neuropathy with a variable degree of ataxia and prominent tremor (Chassande 1998; Smith 1983; Smith 1987; Yeung 1991). The neurophysiological features are demyelination, characteristically with more slowing of conduction in the distal than the proximal nerve segments (Kaku 1994). The term anti‐MAG IgM paraproteinaemic demyelinating peripheral neuropathy (PDPN) is therefore sometimes used to refer to this condition.

Other antibody reactivities exist in people with neuropathy and an IgM paraprotein and they have tentatively been classified into small homogeneous groupings. Examples include motor neuropathies with anti‐GM1 antibodies (Steck 1996), chronic sensory ataxic neuropathy associated with antibodies to GQ1b and other gangliosides with at least two sialic acid groups called CANOMAD (Willison 2001), and sensory demyelinating neuropathies with anti‐sulfatide antibodies (Ferrari 1998). However, the most clearly described, relatively homogeneous group remains the neuropathy associated with IgM anti‐MAG antibodies, the subject of this review.

Considerable evidence exists that anti‐MAG antibodies are pathogenic. Deposits of the IgM paraprotein have been demonstrated in sural nerves of affected people (Takatsu 1985). The paraprotein is bound to myelin in a similar distribution to MAG. In some cases the binding has been associated with the deposition of activated complement components (Hays 1988; Monaco 1990). Demyelination has also been induced in some animal models by the passive transfer of anti‐MAG antibodies (Tatum 1993; Willison 1988), and more recently by immunisation with SGPG (Ilyas 2008). People have been reported to respond to immunotherapies, the improvement sometimes correlating with a reduction in levels of IgM (Wilson 1999).

Description of the intervention

There are a number of possible therapies which have been described as effective in the treatment of anti‐MAG paraproteinaemic neuropathy, including plasma exchange, corticosteroids, intravenous immunoglobulin (IVIg), cyclophosphamide, fludarabine and cytarabine, chlorambucil, alpha‐interferon, rituximab or combinations of these agents. These are described below in the Results and Discussion sections.

Currently, treatment is largely decided on the informed choice of the individual. Many people are too old or too mildly affected to make the risks of treatment worthwhile. There are only eight randomised controlled trials (RCTs) testing the efficacy of the various immunosuppressive regimens in IgM paraproteinaemic neuropathies (Comi 2002; Dalakas 1996; Dalakas 2009; Dyck 1991; Léger 2013; Mariette 1997; Mariette 2000; Niermeijer 2007; Oksenhendler 1995). Most case series have been small, included diverse groups of participants and have not presented results of efficacy clearly.

How the intervention might work

Treatment strategies have either aimed to reduce the IgM paraprotein concentration, by removing the antibody or targeting the presumed monoclonal B‐cell clone and reducing its production, or to interfere with the presumed effector mechanisms such as complement activation or macrophage recruitment.

Objectives

To assess the effects of immunotherapy for IgM anti‐MAG paraprotein‐associated demyelinating peripheral neuropathy.

Methods

Criteria for considering studies for this review

Types of studies

We included RCTs and quasi‐RCTs using any immunotherapy in anti‐MAG paraproteinaemic demyelinating peripheral neuropathy (PDPN).

Types of participants

We included participants of any age with a diagnosis of MGUS, demyelinating neuropathy and anti‐MAG antibodies. Other causes of peripheral neuropathy should have been ruled out.

Paraproteins were to be of the IgM class and shown to be reactive with MAG or SGPG by a validated method which could be:

western blotting against human sciatic nerve homogenate;
enzyme‐linked immunosorbent assay (ELISA) for MAG or SGPG or SGLPG;
complement fixation test against human sciatic nerve homogenate (confirmed by a second method);
thin layer chromatographic immuno‐overlay against SGPG or SGLPG.

As anti‐MAG methods varied, we accepted positive anti‐MAG results as reported by the trial authors and we did not define a threshold titre for positivity.

The neuropathy was to be typical distal symmetrical sensory or sensorimotor and should fit published criteria for slowing of motor nerve conduction in chronic inflammatory demyelinating polyradiculoneuropathy (Ad Hoc 1991 or Nicolas 2002), with or without prominent slowing in distal nerve segments. The occurrence of conduction block in this neuropathy is debated and was not an exclusion criterion.

We included studies that did not exactly fulfil these criteria, if necessary after consultation with the original study authors, provided the review authors agreed that IgM anti‐MAG‐associated demyelinating neuropathy was the preferred diagnosis. We noted any departure from the diagnostic criteria.

Types of interventions

We included any type of immunotherapy used for the treatment of IgM paraprotein‐associated demyelinating peripheral neuropathy with anti‐MAG antibodies. We considered the following therapies for inclusion: plasmapheresis or plasma exchange or selective apheresis, intravenous immunoglobulin (IVIg), corticosteroids (prednisolone, prednisone, methylprednisolone, dexamethasone), chlorambucil, cyclophosphamide, azathioprine, fludarabine,cladribine, interferon alfa‐2a, adriamycin, melphalan, and monoclonal antibody‐based therapies (for example, rituximab). We considered comparisons versus placebo, another treatment, or an alternative dosage or treatment protocol. Therapies could be administered using any protocol (for example, single agent, combined therapy or sequential administration). If the control arm received a co‐intervention then the experimental arm had to also receive that same treatment.

Types of outcome measures

Primary outcomes

Predefined primary outcome measures were numbers of participants improved in disability at six months after randomisation assessed with either or both of:
(a) the Neuropathy Impairment Scale (NIS) (Dyck 1980; Dyck 2005) by at least 10% (maximum score 244);
(b) the modified Rankin Scale (mRS) (Bamford 1989) (scale 0 to 6).

We selected the NIS and mRS as primary outcome measures, as we considered them to be broad, commonly‐used scores that were potentially easy to retrospectively derive from collected data. We predefined six months as the favoured time point for re‐evaluation on the basis that IgM anti‐MAG paraprotein‐associated neuropathy is a chronic and slowly progressive disorder. A reduction in either score indicates improvement and as there is no published minimum clinically important difference for the NIS, we defined a 10% change as improvement.

Secondary outcomes

Secondary outcome measures were:

mean improvement in disability assessed with either the NIS or the mRS, or both, 12 months after randomisation;
change in impairment as measured by improvement in 10‐metre walk time six and 12 months after randomisation (improvement is reduction in walk time);
change in a validated linear disability measure such as the Rasch‐built Overall Disability Scale (R‐ODS) at six and 12 months after randomisation;
change in subjective clinical scores at six and 12 months after randomisation;
change in electrophysiological measures:
1. reappearance of sural sensory nerve action potentials (SNAPs) or compound muscle action potentials (CMAPs) in previously inexcitable nerves, or
2. electrophysiological change in at least two nerves (where improvement was defined as more than a 20% increase in motor or sensory nerve conduction velocities or more than a 20% decrease in motor distal latencies) compared to baseline at six and 12 months after randomisation;
change in serum IgM paraprotein concentration or anti‐MAG titre (significant improvement defined as at least a 20% reduction in IgM or a 50% reduction of anti‐MAG titre compared to baseline at six months after randomisation);
adverse effects from treatment during the trial period, graded as:
- minor ‐ not requiring action;
- moderate ‐ requiring alteration in dosage, drug regimen or other intervention; or
- severe ‐ requiring withdrawal from study or resulting in hospitalisation or death.

Search methods for identification of studies

We searched the Cochrane Neuromuscular Specialised Register (1 February 2016), the Cochrane Central Register of Controlled Trials (CENTRAL) (2016, Issue 1 in the Cochrane Library), MEDLINE (January 1966 to January 2016) and Embase (January 1980 to January 2016) for randomised controlled trials. We reviewed bibliographies to identify other controlled trials. We contacted trial authors and experts in the field to identify additional published or unpublished data.

We also searched ClinicalTrials.gov (ClinicalTrials.gov) and the World Health Organization International Clinical trials Registry Platform (ICTRP) (www.who.int/ictrp/en/), on 23 October 2014.

The detailed search strategies are in the appendices: MEDLINE (Appendix 1), Embase (Appendix 2), CENTRAL (Appendix 3) and Cochrane Neuromuscular Specialised Register (Appendix 4). We searched clinical trials registries from within their search engines with the terms "myelin associated glycoprotein", "anti‐MAG neuropathy" and "paraproteinaemic neuropathy" (Appendix 5).

Data collection and analysis

Selection of studies

Two review authors independently checked titles and abstracts identified from database searches and bibliographies. Both review authors obtained and assessed the full text of all potentially relevant studies. Both review authors decided which trials fitted the inclusion criteria, resolving disagreements about inclusion by discussion.

Data extraction and management

Both review authors independently performed data extraction onto a custom‐designed data extraction sheet, cross‐checked the data and resolved differences by discussion. We requested and obtained missing data from the trial authors whenever possible.

Assessment of risk of bias in included studies

The review authors assessed the risk of bias in included studies using the methods described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). They considered: sequence generation; allocation concealment; blinding of participants, personnel and outcome assessors; incomplete outcome data; selective outcome reporting; and other potential sources of bias.

We graded these items as at low risk of bias, high risk of bias or unclear. Each review author graded the risk of bias independently; we then compared the results and reached agreement about differences by consensus.

Data synthesis

We pooled trial outcomes for interventions when possible. We calculated a weighted treatment effect across trials (using a fixed‐effect model) with the Cochrane statistical package Review Manager 5 (RevMan 2014). We expressed results as risk ratios (RRs) with 95% confidence intervals (CIs) for dichotomous outcomes, and mean differences (MDs) with 95% CIs for continuous outcomes. We tried to analyse all the primary and secondary outcomes under consideration. Because very little meta‐analysis was possible, we did not perform sensitivity analysis. We did not plan any subgroup analyses.

`Summary of findings' tables

We included 'Summary of findings' tables where it was possible to populate these with meaningful data. Where possible, we reported all outcomes at six months after randomisation:

number of participants improved in disability assessed with either or both of the NIS and the mRS;
mean improvement in disability assessed with either the NIS or the mRS;
change in impairment as measured by improvement in 10‐metre walk time;
change in subjective clinical scores (participant subjective impression of change stable or improved);
change in serum IgM paraprotein concentration; and
any adverse event.

Results

Description of studies

Results of the search

Up to 2006, a search of the Cochrane Neuromuscular Specialised Register revealed 12 possible trials. Searching MEDLINE and Embase with the same strategy and handsearching bibliographies failed to reveal further trials. A further search to July 2009 for the update identified trials and papers as follows: CENTRAL: 4, MEDLINE: 679 (13 papers reviewed), and Embase: 212 (seven reviewed). Review of the full texts identified two new eligible studies (Dalakas 2009; Niermeijer 2007). A further search for the 2012 update identified the following numbers of papers: Cochrane Neuromuscular Specialised Register: 3, CENTRAL: 52, MEDLINE: 116, and Embase: 317, with no new trials identified. In 2015, searches identified the following numbers of papers: Cochrane Neuromuscular Specialised Register: 27 papers, CENTRAL: 76 papers, MEDLINE: 65 new papers (1035 total), Embase: 112 new papers (1645 total). A search of the Database of Abstracts of Reviews of Effectiveness (DARE) identified four papers and searches of ClinicalTrials.gov and ICTRP (www.who.int/ictrp/en/) identified three each (identified from the resources above).

Figure 1 presents a flow chart of the study selection process for this update.

Figure 1

PRISMA chart for searched to included studies of IgM paraproteinaemic neuropathy

Included studies

The eight included studies enrolled 236 participants (see Characteristics of included studies; Figure 1), of whom 104 were treated at some stage with the study intervention. In only three trials were all the predeclared inclusion criteria met (Comi 2002; Dalakas 2009; Niermeijer 2007). Three studies stated that the participants had benign monoclonal gammopathy or monoclonal gammopathy of undetermined significance (Comi 2002; Dalakas 2009; Niermeijer 2007). Other studies may have included participants with Waldenstrom's macroglobulinaemia (Mariette 1997; Mariette 2000; Oksenhendler 1995), or participants with other causes of neuropathy (Dalakas 1996). Three studies stated that all participants had a demyelinating neuropathy (Comi 2002; Dalakas 1996; Dalakas 2009), but only one of these stipulated the electrophysiological criteria for demyelination used (Comi 2002). All of the included participants had serum IgM anti‐MAG activity tested but this was established by various methods of western blotting, immunofluorescence or anti‐SGPG ELISA testing. In total, 194 of 236 participants had anti‐MAG activity, and in four of eight studies all the participants in the study had anti‐MAG activity (Dalakas 2009; Léger 2013; Mariette 1997; Mariette 2000). Niermeijer 2007 stated that numbers of participants were too small to establish an effect of anti‐MAG antibodies and no original data were available to analyse the subgroup of anti‐MAG participants in the other three studies.

The interventions and trial designs of the eight included studies were widely different. Two studies compared IVIg with placebo in a placebo‐controlled cross‐over design (Comi 2002; Dalakas 1996). Both of these studies included a one‐month minimum washout period between the cross‐over arms, considered rather short given the half‐life of IgG, even though Comi 2002 gave a single dose of IVIg compared to the three monthly doses in Dalakas 1996. One study compared interferon alfa‐2a with IVIg in an open‐label design (Mariette 1997) and another with placebo in a double‐blind design (Mariette 2000). One study compared chlorambucil alone with a combination of chlorambucil and plasma exchange (Oksenhendler 1995). The most recent trials studied the effect of cyclophosphamide and prednisolone (Niermeijer 2007), or rituximab (Dalakas 2009; Léger 2013). We included all eight studies, despite the lack of fulfilment of all inclusion criteria by five of them.

Excluded studies

We excluded 19 studies that had been considered at the title selection stage (see Excluded studies). Of these, we excluded one trial as serum IgM anti‐MAG activity was not tested in the 21 participants with IgM‐associated neuropathy and specific criteria for demyelinating neuropathies were not an entry criterion (Dyck 1991). We excluded two rituximab trials: as one (with seven participants) was open (Renaud 2003), the other (including nine participants) had a non‐randomised control group and unblinded assessments (Pestronk 2003). Other reasons for excluding studies are shown in Characteristics of excluded studies.

Risk of bias in included studies

See Figure 2 for a summary of the review authors' 'Risk of bias' assessments for the included studies.

Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

All the included trials had a randomised design and participants were randomly assigned to receive the intervention or control for the whole trial, or the first arm in cross‐over studies (see Characteristics of included studies). All but one of the studies stipulated explicit diagnostic criteria (Dalakas 1996). Baseline characteristics were not significantly different in six of the studies, but in Dalakas 2009 the groups were very unbalanced with very few men in the rituximab‐treated group at randomisation. Differences in baseline characteristics were also present in Dalakas 1996.

The trials of IVIg both had low risk of bias (Comi 2002; Dalakas 1996). In Dalakas 1996, the risk of bias was a little greater, as allocation concealment was unclear and the participants randomised to placebo were much older than in the IVIg arm, although this did not reach significance with the nine anti‐MAG participants included. However, the sensory scores were significantly higher in the placebo group than the IVIg‐treated group. Furthermore, this study did not state explicit diagnostic criteria. Both trials suffer from a short washout period and a potential carry‐over effect, although this would be expected to reduce the significance of any therapeutic effect of IVIg.

Mariette et al studied interferon alfa‐2a in two studies (Mariette 1997; Mariette 2000). The second study was blinded and controlled and at a much lower risk of bias than the preliminary randomised open trial. Four of 10 participants in the IVIg group in Mariette 1997 and three of 12 in the interferon group in Mariette 2000 dropped out early in the trial, which may confound results in these small groups.

Oksenhendler 1995 studied plasma exchange plus chlorambucil versus chlorambucil alone. As a result of the plasma exchange procedure, the blinding and allocation concealment were inadequate. Furthermore, although dropouts were balanced in the two groups (four of 22 in one group and four of 23 in the other), and the investigators performed an intention‐to‐treat analysis, the dropout numbers were deemed significant given the small number of included participants. Overall, we judged the risk of bias in this small study to be high.

The first published RCT of rituximab for paraproteinaemic demyelinating peripheral neuropathy (PDPN) was at high overall risk of bias (Dalakas 2009). The treatment groups at entry were unbalanced with respect to sex distribution (PDPN affects more men than women, but only two of 13 participants randomised to treatment were men). The randomisation method was not clear, and although the National Institutes of Health (NIH) pharmacy provided drug and placebo, it is not clear that the trial maintained allocation concealment. Most importantly, one participant was randomised and then removed from the trial having received rituximab and suffered an adverse event; this participant's data were not analysed. A further participant was removed at the analysis stage when it became clear the person did not fulfil the inclusion criteria; statistical analyses were done including and excluding this participant. We judged a second trial of rituximab to be at low risk of bias. The trial was entered in ClinicalTrials.gov, but the published protocol does not contain all outcomes presented in the trial report, which appear to have been added post protocol publication. To clarify the meta‐analysis, we requested trial data from the authors of these two studies. Data were provided for Léger 2013. The authors of Dalakas 2009 were unable to provide any outcome data for their trial, which were lost with a retiring colleague.

We considered the study of cyclophosphamide and steroid to be at low risk of bias in all domains (Niermeijer 2007). We analysed the first six months of the trial, as after this point participants who experienced a clinical decline were rerandomised to treatment with a different corticosteroid or cyclophosphamide.

Effects of interventions

General comment

Two trials compared IVIg with placebo (Comi 2002; Dalakas 1996). Their designs differed in IVIg dose and outcome measures and neither provided six‐ or 12‐month assessments.

Two trials tested rituximab in a very similar group of participants, using fairly similar outcomes. In Dalakas 2009, 26 participants were randomised to rituximab or a placebo infusion. There were concerns about the validity of the statistical manipulations used in this trial (see above and below) and corroborative data were not available. Léger 2013 also studied rituximab and randomised 54 participants to rituximab or placebo. The study produced similar results to Dalakas 2009 (see analysis below). The trial data were analysed at slightly different time points, but the review authors considered eight‐ and 12‐ month outcomes similar enough in this slowly changing disease.

The other four trials tested different interventions or combinations of treatment. Oksenhendler 1995 compared plasma exchange with no plasma exchange in participants receiving chlorambucil. Mariette et al compared IVIg with interferon alfa‐2a (Mariette 1997), and, in a follow‐up study, interferon alfa‐2a with placebo (Mariette 2000). Niermeijer 2007 randomised participants to cyclophosphamide and prednisolone pulses or placebo in a six‐month‐long phase one of a 24‐month trial. To satisfy ethical concerns, any participant who deteriorated in phase one was reassigned to cyclophosphamide and prednisolone (if they had received placebo in phase one) or dexamethasone (if they had received cyclophosphamide and prednisolone) (termed phase 2 of the study), making the 24‐month outcomes difficult to assess. In phase two the placebo group was not truly randomised, as phase two contained only participants who had not improved on placebo, nine of 19 participants having switched to active treatment in phase two. After discussion of the trial data released by the trial authors, we did not include six‐month to 24‐month outcomes in the analysis.

The primary and secondary outcome measures used in all these trials varied widely and those predefined for this review were rarely provided.

Intravenous immunoglobulin versus placebo

Investigated in Comi 2002 and Dalakas 1996. See 'Summary of findings' (Additional Table 1).

Table 1. 'Summary of findings' table: Intravenous immunoglobulin (IVIg) versus placebo

IVIg versus placebo for IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathies
Patient or population: people with IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathies Settings: hospital and outpatient treatment centres Intervention: IVIg versus placebo
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)	Comments
Outcomes	Risk with placebo	Risk with IVIg	Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)	Comments
Number of participants improved in disability score at 6 months ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Short‐term outcomes only, not reported at 6 months
Mean improvement in disability score at 6 months ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Short‐term outcomes only, not reported at 6 months
Improvement in 10‐metre walk time at 6 months ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Short‐term outcomes only, not reported at 6 months
Participant subjective impression of change at 6 months ‐ number of participants reporting improvement ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Short‐term outcomes only, not reported at 6 months
Change in serum IgM paraprotein concentration 8 months after treatment ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Short‐term outcomes only, not reported at 6 months
Any adverse event	Data for the anti‐MAG subgroup were not available. In one study, one participant had an aseptic meningitis and a rash after IVIg treatment, with 2 moderate adverse events in the placebo group. In the other trial, there were no serious adverse events with IVIg, but mild and transient effects were "more common in the IVIg than the placebo group".		Not estimable	‐	See comment	No further analysis was possible.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

¹Incomplete reporting of adverse events in one study.

Primary outcome measures

One trial with 22 participants (Comi 2002), of IVIg versus placebo in IgM paraprotein‐associated neuropathy used the modified Rankin Scale (mRS). However, this was a short trial and the investigators only measured the primary outcome at two and four weeks, not at six months. Eleven of 22 participants had anti‐MAG antibodies but the data concerning the anti‐MAG participant subgroup were not available. At two weeks, the mRS score showed a significant improvement with IVIg (‐0.38, standard deviation (SD) 0.58) over placebo (+0.19, SD 0.51) at two weeks (P = 0.008), a difference that may not be clinically significant. No significant difference between the groups was present at four weeks.

The primary outcome of Comi 2002 was the Inflammatory Neuropathy Cause and Treatment (INCAT) disability score at four weeks. This was not one of our prespecified outcome measures (see Discussion). After two weeks, the INCAT disability score did not change significantly; there was no difference between the groups and no evidence of a carry‐over effect. After four weeks, the INCAT disability score decreased by a mean of 0.55 (SD 0.67) grades in the IVIg period (P = 0.001), with no significant change (mean 0.05 (SD 0.90)) in the placebo period. Only the mean difference (MD) between the treatment effects in the two four‐week periods was significant (0.5 of a grade, 95% confidence interval (CI) 0.00 to 1.00).

Secondary outcome measures

The secondary outcomes at time points selected for this review were not available because the trials were both so short.

Related secondary outcomes were measured at different time points (two and four weeks) and the results were presented as MD between treatments.

a. One trial assessed the 10‐metre walk (Comi 2002), but only at four weeks in the first arm of the study, as there was significant evidence of a carry‐over effect into the second arm. There was a significant reduction in 10‐metre walk time with IVIg compared with placebo in the whole IgM treatment group (MD 2.77 seconds, 95% CI 0.01 to 5.54). No data were available on the anti‐MAG participants alone, but the trial authors state that MAG antibodies "did not influence the response to IVIg treatment".

b. Serum data: not enough information was available on anti‐MAG titres or serum IgM concentrations in Dalakas 1996 to make a meaningful comment.

No 12‐month outcomes were available.

Adverse events

Two studies compared IVIg and placebo (Comi 2002; Dalakas 1996), but data for the anti‐MAG subgroup were not available. In Comi 2002, one participant had an aseptic meningitis and a rash after IVIg treatment. It is not clear whether she completed the three‐month trial period. Two moderate adverse events occurred in the placebo treatment arm in Comi 2002, one retinal vein thrombosis and one episode of transient diplopia. In Dalakas 1996, no serious adverse events occurred as a result of IVIg infusion, but mild and transient effects were "more common in the IVIg than the placebo group". No further analysis was possible.

For adverse events of IVIg in the interferon alfa‐2a versus IVIg trial (Mariette 1997), see below.

Interferon alfa‐2a versus placebo or IVIg

Investigated in Mariette 1997 and Mariette 2000.

Primary outcome measures

The mRS was not assessed. The Neuropathy Impairment Scale (NIS) was the primary outcome in both studies involving interferon alfa‐2a (referred to as the Clinical Neuropathy Disability Score (CNDS), a scale derived with minor modifications from the NIS, maximum score 93 (higher scores indicate greater disability)). In the initial randomised but open parallel‐group study of interferon alfa‐2a versus IVIg (Mariette 1997), participants in the IVIg group worsened by a mean of 2.3 (SD 7.6) points on the NIS at six months, and those in the interferon alfa‐2a group improved by 7.5 (11.1) points, a MD of 9.80 (95% CI 1.46 to 18.14, n = 20) in favour of interferon alfa‐2a (Analysis 1.1). However, Mariette 2000 (n = 24), a randomised and blinded trial of interferon alfa‐2a versus placebo at low risk of bias, failed to confirm the beneficial effect of interferon alfa‐2a in the primary outcome at six months (Analysis 2.1).

Secondary outcome measures

Neither trial assessed the 10‐metre walk at any time point.

The two trials derived a subjective clinical score (Mariette 1997; Mariette 2000). In the unblinded 1997 study, the subjective scores improved significantly in favour of interferon alfa‐2a over IVIg (MD 3.10, 95% CI 1.02 to 5.18, n = 20; Analysis 1.2). In the later blinded and reported 'negative' placebo‐controlled study of interferon alfa‐2a, the subjective scores were not presented in the published data.

The presence of sural sensory nerve action potentials (SNAPs) was examined at six months in both studies. The number of participants with detectable sural nerve SNAPs did not improve with either interferon alfa‐2a or IVIg in the first study and exhibited "no significant improvement" in the second (data not presented). More participants had recordable median SNAPs after six months treatment with interferon alfa‐2a (2/7 improved to 5/7) compared to IVIg (4/8 declined to 1/8) in the first study (Mariette 1997).

In Mariette 1997, IgM paraprotein bands and anti‐MAG activity were still detectable at six months in all participants but only decreased significantly in two participants receiving interferon alfa‐2a (50% reduction each). No 12‐month outcome data for serum data were published.

Adverse events

In both interferon alfa‐2a trials all included participants had anti‐MAG serum activity (Mariette 1997; Mariette 2000). Interferon alfa‐2a caused flu‐like symptoms in all 10 participants in the open study (Mariette 1997), which were persistent and required tapering of the dose in three (moderate side effects). A further three participants required dose‐tapering for systemic adverse effects (also moderate). One of 10 participants in the IVIg treatment group withdrew because of self‐limiting erythroderma (Mariette 1997). No other mild or moderate side effects were reported with IVIg.

In the placebo‐controlled, blinded study, two of 12 participants withdrew from the study because of side effects (severe), one with diarrhoea and one with flu‐like symptoms (Mariette 2000). One participant withdrew with worsening of neuropathy. There were no other severe side effects reported for interferon alfa‐2a.

Chlorambucil and plasma exchange versus chlorambucil alone

Investigated in Oksenhendler 1995.

Primary outcome measures

In a comparison of chlorambucil and plasma exchange versus chlorambucil alone in IgM paraprotein‐associated neuropathy, the NIS at four months showed no statistically significant difference in outcome (Oksenhendler 1995). The trialists assessed neither the NIS nor the mRS at the six‐month time point specified for this review. Data were not available for the anti‐MAG subgroup alone, as participants with anti‐MAG activity were not differentiated from those without.

Secondary outcome measures

The change in NIS at 12 months was no different between the plasma exchange plus chlorambucil and the chlorambucil‐alone groups.

The 10‐metre walk time was not recorded at six months.

A subjective clinical score was not recorded at six months. There was "no difference" in the subjective clinical score at 12 months between the groups.

SNAPs were not recorded at six months.

IgM paraprotein concentrations and anti‐MAG (anti‐myelin IgM) titres were determined prior to entry but no results were given for later time points other than a comment in the paper that the response was not associated with a significant decrease in the serum IgM concentration.

Adverse events

In Oksenhendler 1995, there were no serious adverse events from plasma exchange, but side effects of chlorambucil were common. Ten of the 44 participants required temporary suspension or tapering of the chlorambucil dosage because of haematotoxicity, but none had to cease treatment.

Rituximab versus placebo

In 2012 only one study had been published using rituximab as an intervention for paraproteinaemic neuropathy that fulfilled our inclusion criteria (Dalakas 2009). A second study called RiMAG has now been published and we include it here (Léger 2013). See summary of findings Table for the main comparison.

Dalakas 2009 randomised 26 participants to receive either rituximab or placebo at a standard dose of four infusions of 375 mg/m² every week for four weeks. Outcomes were measured at eight months (which for the purposes of this review we have used as six‐month outcomes). One randomised participant had a severe anaphylactic adverse event with the first infusion of the drug. The participant was replaced and not included in any further analysis. One further participant randomised to rituximab and treated in accord with the protocol was found to have had an INCAT leg score at entry of zero (inadvertently assigned as 1 at entry). He could not therefore have improved. We have performed the analysis here with the participant included as in the protocol for this review.

Some discrepancies exist in the results in the published paper (see below) (Dalakas 2009). These were resolved through enquiry with the trial author. Original trial data were not made available in 2012 and further requests in 2014 and 2015 were not fruitful. Trial data were "entered directly by the statistician who retired more than 10 years ago" and he was untraceable.

Léger 2013 randomised 54 participants to receive either rituximab in a standard 375 mg/m² dose or placebo. Outcomes were measured at 0 and 12 months, but upon request, the trial authors also made nine‐month data available for the INCAT scale, change in INCAT scale and IgM levels, to correspond to outcomes of Dalakas 2009. We requested data for the leg subscale of the INCAT score but these were not available. The outcomes in the published and amended ClinicalTrials.gov protocol are limited and do not fully correspond to the outcomes in the methods section of the published paper (which are those with results). There is no suggestion of suppression of negative outcome data or selective reporting, but these outcomes could have been decided post hoc.

Primary outcome measures

These two studies included neither the NIS nor the mRS as outcomes. The INCAT leg score reported in Dalakas 2009 corresponds broadly to the mRS and thus we used this in our analysis both as a continuous and a dichotomous outcome. For the dichotomous improvement we have compared the leg score at eight months from Dalakas 2009 with the dichotomous complete INCAT score at 12 months from Léger 2013 (Analysis 3.1; Figure 3). The RR for improvement was 3.51, 95% CI 1.30 to 9.45 (I² = 0%, 73 participants). We will perform a six‐month meta‐analysis if other studies become available. Using the published eight‐month data from Dalakas 2009 and 12‐month data from Léger 2013, the mean improvement in the INCAT leg disability score (MD ‐0.45, 95% CI ‐0.85 to ‐0.05; I² = 0%) was statistically significant, whether the INCAT 0‐score participant from Dalakas 2009 was included in the analysis or not (summary of findings Table for the main comparison; Analysis 3.2). Participant‐level data obtained for nine‐month improvement from Léger 2013 (not published), which are at a more comparable time point to Dalakas 2009, give a MD of ‐0.33, 95% CI ‐0.73 to 0.07 (I² = 0%, 70 participants) in the same direction as the longer time point but including the possibility of no effect (Analysis 3.3).

Figure 3

Forest plot of comparison: 3 Rituximab versus placebo, outcome: 3.1 Number of participants improved on INCAT score (see text) at 8‐12 months.

(Note: in the published paper of Dalakas 2009 there is a discrepancy between the data for the INCAT leg score in the text and in the published table. We checked the data with the trial authors who gave verbal clarification and we included the correct figures. No recorded trial data were made available.)

Secondary outcome measures

Secondary outcomes that we could include in our analyses were the 10‐metre walk time, the change in IgM levels, and anti‐MAG titre, but these were all measured at eight months in Dalakas 2009 and at 12 months in Léger 2013. Nine‐month data were not available for Léger 2013 except for the IgM levels.

The 10‐metre walk times did not improve significantly, either in terms of time to walk (MD ‐0.35 seconds, 95% CI ‐1.89 to 1.19; I² = 0%, 68 participants) or numbers of participants improved (RR 1.80, 95% CI 0.83 to 3.92; 26 participants; Dalakas 2009 only ‐ data from Léger requested; Analysis 3.5; Analysis 3.6). As might be expected, there was a measurable and significant decrease in the level of serum IgM eight months after rituximab (reduction of 286 mg/dL, 95% CI 329 to 244; 26 participants; Analysis 3.7). Less expected (from reports of other studies of treatment of anti‐MAG neuropathy) but not surprising was a decrease in the titre of anti‐MAG activity of ‐17.79 units/mL (95% CI ‐33.33 to ‐2.25; Analysis 3.8; I² = 0%, 71 participants).

Outcome measures not prespecified in this protocol

Both studies included participant "clinical assessments and questionnaires" (not further described). The RR for "stable or improved" at eight to 12 months was 1.86 (95% CI 1.27 to 2.71; I² = 0%, 70 participants) in favour of rituximab (Analysis 3.9). For participants "improved" only at eight to 12 months, the RR was 9.67 with a wide CI (95% CI 1.84 to 50.85; I² = 0%, 70 participants) albeit statistically significant (Analysis 3.10). Léger 2013 presented results for the Short Form 36 Health Survey (SF36) with a MD improvement of 15.50 (95% CI 5.24 to 25.76; 37 participants) in the physical subscores (Analysis 3.11) but no statistically significant change in the mental subscores (MD 6.60, 95% CI ‐0.35 to 13.55; 41 participants; Analysis 3.12). These assessments were not prespecified as primary or secondary outcomes.

Adverse events

There were not significantly more adverse events in the rituximab group (RR 1.18, 95% CI 0.84 to 1.66; I² = 0%, 80 participants; Analysis 3.13). Most adverse events were minor, including mild temperature increases and chills, headaches and mild hypotension, nausea, vomiting, dizziness and lightheadedness, and rash. One severe adverse event of bronchospasm led to the participant being removed from Dalakas 2009, despite receiving the drug; his data were not included in the published results. In Léger 2013, one participant with rituximab was withdrawn from the study after developing a bradycardia. The RR for severe adverse events was 3.11 (95% CI 0.34 to 28.54; I² = 0%, 80 participants; Analysis 3.14), but this figure should be interpreted with caution, given the small number of participants and single‐integer events.

Cyclophosphamide and prednisolone combination versus placebo

The only trial to use this combination as an intervention was Niermeijer 2007. See 'Summary of findings' Additional Table 2. Values for outcomes in this study were presented as means and 95% CIs. We calculated SDs from the 95% CIs.

Table 2. 'Summary of findings' table: Cyclophosphamide and prednisone versus placebo

Cyclophosphamide and corticosteroids compared to placebo for IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathies
Patient or population: people with IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathies Settings: hospital and outpatient treatment centres Intervention: cyclophosphamide and corticosteroids Comparison: placebo
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)	Comments
Outcomes	Risk with placebo	Risk with cyclophosphamide and corticosteroids	Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)	Comments
Number of participants improved in disability score at 6 months Modifed Rankin Scale Scale from: 0 to 6 (normal to death) Follow‐up: mean 6 months	Study population		RR 3.56 (0.41 to 30.99)	35 (1 study)	⊕⊕⊕⊝ moderate¹
	53 per 1000	189 per 1000 (22 to 1000)	RR 3.56 (0.41 to 30.99)	35 (1 study)	⊕⊕⊕⊝ moderate¹
Mean improvement in disability score at 6 months Modifed Rankin Scale Scale from: 0 to 6 (normal to death) Follow‐up: 6 months²	The mean improvement in disability score at 6 months in the control groups was 0.11 units	The mean improvement in disability score at 6 months in the intervention groups was 0.31 lower (0.61 lower to 0.01 lower)		34 (1 study)	⊕⊕⊕⊝ moderate¹
Improvement in 10‐metre walk time at 6 months ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Not reported
Participant subjective impression of change at 6 months ‐ number of participants reporting improvement ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Not reported
Change in serum IgM paraprotein concentration after treatment ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Not reported
Any adverse event ‐ not measured	See comment	See comment	Not estimable	‐	See comment	Adverse events not specifically collected but text notes in the paper (see text of review)
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; IgM: immunoglobulin M; RR: risk ratio
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

¹Dowgraded once for imprecision (single small study).
²Phase 1 of cross‐over trial only.

The trial authors made original trial data available and we considered whether to carry out further statistical analysis to use the data after six months from phase two (see Effects of interventions above). However, after discussion we did not consider the data to be valid, as the re‐allocation of participants was not truly random and the results would therefore be significantly biased.

Primary outcome measures

The mRS at six months was included as one of the outcomes of phase one of this trial, the only disability data that could be included in this review. At six months a (barely) significant improvement in the mean mRS was present, with a MD of ‐0.31 points (95% CI ‐0.61 to ‐0.01; 34 participants; Analysis 4.1), with no significant difference in the number of participants improved at six months when we extracted the dichotomous data from the dataset provided by the authors (3/16 improved with cyclophosphamide and prednisolone combination versus 1/19 with placebo, RR 3.56, 95% CI 0.41 to 30.99; 35 participants; Analysis 4.2.

Secondary outcome measures

The authors of this trial used the Rivermead Mobility Index (RMI) as their primary outcome measure. We included the RMI as a secondary outcome under subjective clinical scores. No significant improvement in the RMI was present after six months of cyclophosphamide and prednisolone (MD 0.42, 95% CI ‐0.41 to 1.25; 34 participants; Analysis 4.3).

Adverse events

Adverse events were not specifically sought in this trial. One participant receiving cyclophosphamide withdrew because of angina and was lost to follow‐up, and one participant withdrew because of rapid progression of neuropathy. A further participant, who was also cyclophosphamide‐treated, withdrew during the seven‐month to 12‐month phase. The trialists reported that nausea was significantly more common in the cyclophosphamide‐treated group (P = 0.001), but provided no further details.

Three of 35 participants (8.6%) who had received cyclophosphamide developed an immunocytoma not requiring further treatment at five years of follow‐up.

Discussion

Randomised controlled trials (RCTs)

We identified eight RCTs for this update. Four fulfilled all our strict predefined inclusion criteria (Comi 2002; Dalakas 2009; Léger 2013; Niermeijer 2007). The trials of Comi, Leger and Niermeijer had a low risk of bias, whereas Dalakas 2009 had a high risk of bias in multiple domains (Figure 2). We included the four other RCTs that did not fulfil all the inclusion criteria, because their inclusion criteria closely approximated to those intended.

According to evidence from two trials, IVIg may produce some short‐term benefit in the treatment of anti‐MAG IgM paraproteinaemic neuropathy. One double‐blind, placebo‐controlled study with minimal bias showed benefit of IVIg at four weeks in one of the primary outcome measures (mRS) and the 10‐metre walk time secondary outcome measure but no other significant outcomes (Comi 2002). The outcomes were available only at four weeks and were not available for the predefined endpoints of six and 12 months; it is not clear whether any short‐term benefit would be sustained or whether IVIg is clinically useful. Furthermore, this trial was of IgM paraproteinaemic neuropathy and not all the included participants had associated anti‐MAG activity. Evidence for any benefit from IVIg should be regarded with caution.

Two trials with interferon alfa‐2a gave contradictory results. Interferon alfa‐2a appeared to be of benefit in treating anti‐MAG neuropathy when tested in an open trial with IVIg (Mariette 1997). Because this trial was not blinded it was at high risk of bias. The later, less biased, double‐blind, placebo‐controlled study of interferon alfa‐2a did not demonstrate any significant benefit from interferon alfa‐2a (Mariette 2000).

A single trial assessed the use of plasma exchange as additional treatment to chlorambucil (Oksenhendler 1995). No additional benefit was gained from the addition of plasma exchange in the predefined outcome measures at four or 12 months. This relatively large study lacked blinding and allocation concealment, and the results should therefore be interpreted with caution. Since the trial showed no additional benefit from plasma exchange, any possible bias did not lead to a significant result.

A number of case reports and small case series have described use of rituximab for the treatment of paraproteinaemic neuropathy (Barohn 2005; Benedetti 2005; Benedetti 2007; Benedetti 2008; Canavan 2002; Delmont 2010; Goldfarb 2005; Gono 2006; Gorson 2007; Kelly 2006; Kilidireas 2006; Latov 1999; Levine 1999; Pestronk 2003; Renaud 2003; Renaud 2006; Weide 2000; Niermeijer 2009). The results were varied, with most reporting positive outcomes in the majority of participants but some very usefully reporting no response in all their reported cases (Barohn 2005; Rojas‐García 2003), and some reporting deteriorations (Broglio 2005; Gironi 2006; Noronha 2006). We included the first rituximab RCT in the 2012 update (Dalakas 2009), and we now include a further RCT (Léger 2013). Dalakas 2009 had a high risk of bias in two assessed domains and unclear risks in three others of the six assessed. As such, we assessed the evidence from this trial as very low quality across the outcomes stipulated. The results should therefore be regarded with caution. With the removal of a participant from the analysis who should perhaps not have been randomised, the trial reported improvements in the INCAT leg score and the 10‐metre walk time, and in the whole group, the IgM levels and anti‐MAG titres. The study reported seven of 13 rituximab‐treated participants compared to none of the 13 participants in the placebo group experiencing improvement in "patient clinical assessments and questionnaires". In our analysis, although there was a trend in favour of the treatment group in all analyses, none of the clinical outcomes were statistically significantly different from placebo. However, rituximab did reduce IgM levels and anti‐MAG titres, as would be expected. The interpretation of the results in the original paper by the trial authors concludes that "The results warrant confirmation with a larger trial". We have now added the data from the larger and less biased Léger 2013 to the meta‐analysis. Léger 2013 was reported as demonstrating no change in its primary and secondary outcomes. However, the outcome changes, where they are comparable, are all in a similar direction and of similar magnitude to those of Dalakas 2009 (I² = 0% throughout). In meta‐analysis, comparing rituximab to placebo, significant improvements are seen in numbers of participants improved on the INCAT score (although the magnitude of this does not reach clinical significance), the biological parameters (IgM, anti‐MAG titres), the physical subscores of the SF36 (but not the mental), and perhaps importantly the patient clinical impression of change scores, both for improvement alone and stabilisation and improvement. The numbers of participants are small but the results demonstrate no heterogeneity. The results of Léger 2013 are more robust, at less risk of bias and more reliable, and support an effect of rituximab in stabilising or improving anti‐MAG neuropathy.

Niermeijer 2007 trialed cyclophosphamide and corticosteroids (CP) in combination, versus placebo. Ethical concerns led to modification of the protocol such that we could not use the data after the sixth month in the meta‐analysis. We found no significant improvements in the functional scales used (mRS and RMI), as displayed here, from moderate‐quality evidence. The trial reports significant improvements in validated impairment measures of strength and sensory dysfunction (not predefined outcomes in this meta‐analysis), which did not translate into functional benefits.

This review has identified many problems with the trials of treatment for this indication, including variations in trial design, participant inclusion and exclusion criteria, analysis points, and outcome measures. Only six of the eight trials were blinded. Some trials employed a cross‐over design, which must be allowed for in analysis. In one study an alteration in the trial design for ethical reasons meant that data beyond the six‐month time point were not usable in this analysis (Niermeijer 2007). Most studies did not use validated clinical and disability scales that are likely to detect changes (for example, the INCAT Sensory Sum Score (Merkies 2000)) and these should be considered in future studies. Further, more sensitive scales measuring both disability and specific domains of impairment that are affected in this predominantly sensory neuropathy are being developed. Although Comi 2002 was possibly too short to detect meaningful changes, it did include validated, clinically useful, and reproducible endpoints such as the sensory sum score, 10‐metre walking time, nine‐hole peg test, the Rotterdam Scale (a handicap score), and the SF36 quality‐of‐life scale. Dalakas 2009 also used more up‐to‐date outcomes, although the trial was probably too small to detect significant change, and bias makes the results unreliable. Further work is ongoing to develop valid disability measures (see for example the PeriNomS project). Objective endpoints such as presence of SNAPs were infrequently used, but in clinical practice these seldom change in the short to medium term. No study has included a measure of fatigue (Merkies 1999), which is prevalent in chronic inflammatory demyelinating polyradiculoneuropathy but has not been investigated in anti‐MAG neuropathy. Furthermore, effective treatments, if identified, are likely to be expensive and possibly invasive, and investigators should therefore consider measures of quality of life and cost effectiveness in all trials.

Only four trials confined recruitment to participants with anti‐MAG antibodies (Dalakas 2009; Léger 2013; Mariette 1997; Mariette 2000). The anti‐MAG neuropathies constitute the majority of the IgM paraproteinaemic demyelinating neuropathies in terms of immunoreactivity. They are relatively homogeneous in their clinical features and are distinct from some other rare but identified subgroups (for example, those with IgM anti‐GM1/GD1b reactivity). Many IgM paraproteinaemic neuropathies have no identified antigen target and yet are not clinically very different from those with anti‐MAG antibodies. This is often used as an argument that anti‐MAG antibodies are not relevant in the pathogenesis. Although this may be the case, there is a considerable body of evidence in favour of anti‐MAG antibodies causing demyelination, as described in the Background. Lack of anti‐MAG antibodies may simply reflect low detection through inadequate immunological identification, or alternative antigenic targets yet to be described. Subgroup analysis by anti‐MAG activity should be reported in future trials.

Cochrane is increasingly concerned about the quality of data supplied for systematic reviews from original trials where a number of instances of fraudulent publishing have been uncovered. Any future trials in the area should eradicate the need for suspicion about the data published, by publishing, in full, the trial protocol on a clinical trials website, making minimal change (but being transparent with the change if required), publishing all the predefined outcomes for all trials, and making the data available for scrutiny.

Non‐randomised studies

As part of our systematic search of the literature for RCTs, we identified many case reports and small non‐randomised case series. These included all the therapies covered in the RCTs above, as well as other therapies not subjected to RCTs. These are documented in Table 3 but this should not be regarded as a systematic presentation of the non‐randomised literature.

Table 3. Details of non‐randomised studies

Study ID	Study design	Particiant number (total)	Participant number (IgM)	Participant number (MAG)	Intervention	Number improved (IgM)	Number improved (MAG)
Latov 1980	Case report	1	1	1 (anti‐myelin Abs)	Prednisolone + chlorambucil + plasma exchange	1/1	‐
Dalakas 1981	Uncontrolled case series (response to Rx recorded in Latov 1988)	11	4	?	a. Prednisolone b. Prednisolone + chlorambucil	a. 0/1 b. 1/1	‐
Melmed 1983	Uncontrolled case series (response to Rx recorded in Latov 1988)	3	3	3	a. Corticosteroids b. Steroids + plasma exchange + chlorambucil	‐	a. 3/3 b. 0/3
Stefansson 1983	Case report	1	1	1	Plasmapheresis + immunosuppressant	‐	0/1
Meier 1984	Case report	1	1	1	Plasmapheresis + prednisolone + chlorambucil	‐	1/1
Sherman 1984	Retrospective case series	10	6	?	Plasmapheresis + chlorambucil	2/5	‐
Bland 1985	Case report	1	1	1	Plasmapheresis + chlorambucil	‐	1/1
Frayne 1985	2 cases	2	2	‐	Plasma exchange	2/2	‐
Ernerudh 1986	Retrospective non‐randomised	3	3	3 (anti‐myelin Abs)	Plasma exchange	2/3 (temporary)	‐
Hafler 1986	Uncontrolled case series (response to Rx recorded in Latov 1988)	9	9	7	a.Corticosteroids b. Plasma exchange c. Cyclophosphamide	‐	a. 0/5 b. 1/5 c. 1/1
Smith 1987	Prospective non‐randomised case series	13	8 (2 lymphoma)	4 (1 lymphoma)	a. Chlorambucil + plasma exchange b. Plasma exchange	a. 3/4 b. 1/2	a. 1/2 b. 1/1
Haas 1988	Case report	1	1	1	Plasma exchange	1/1	1/1
Kelly 1988	Uncontrolled case series	10	10	5	Prednisolone + cyclophosphamide (azathioprine or chlorambucil) + plasma exchange	4/4 (all 5 not treated)	2/3 (All 5 not treated)
Latov 1988	Uncontrolled case series	11	11	11	a. Plasmapheresis b. Prednisolone c. Chlorambucil d. Prednisolone + plasma exchange e. Plasma exchange + chlorambucil	‐	a. 3/3 b. 0/1 c. 3/4 d. 0/1 e. 0/1
Donofrio 1989	Case report	1	1	0	a. Corticosteroids b. Plasma exchange + corticosteroids	a. 0/1 b. 1/1	‐
Cook 1990	Case reports	2	2	1	a. IVIg b. prednisolone	a. 2/2	a. 1/1 b. 0/1
Hodgkinson 1990	Uncontrolled case series	8	2	?	Ciclosporin	2/2	‐
Ernerudh 1992	Prospective non‐randomised	5	5	5 (anti‐myelin Abs)	a. Plasma exchange b. Prednisolone c. Melphalan d. Chlorambucil e. Chlorambucil and plasma exchange f. Chlorambucil + prednisolone	a. 1/1 b. 1/2 c. 0/1 d. 0/1 e. 1/1 f. 1/2	‐
Hoang‐Xuan 1993	Open prospective	4	4	3 (2 in Léger 1994)	IVIg	‐	0/1
Léger 1994	Open prospective	7	7	5	IVIg	‐	2/5
Sherman 1994	Abstract	10	10	8	Fludarabine	‐	7/8
Siciliano 1994	Case report	3	1	?	Selective apheresis	1/1	‐
Blume 1995	Uncontrolled case series	4	3	4	Plasma exchange + cyclophosphamide	3/3	4/4
Ellie 1996	Retrospective case series	33	33	33	a. Plasma exchange b. IVIg	a. 2/6 b. 13/17	‐
Notermans 1996	Open prospective	16	11	5	Cyclophosphamide + prednisolone	‐	Unclear
Gorson 1997	Retrospective case series	15	12	8	a. Plasma exchange b. Prednisolone c. IVIg d. Cyclophosphamide + plasma exchange	a. 10/12? b. 1/2 c. 2/6 d. 2/2	Unclear
Rudnicki 1998	Case report	1	1	1	a. Fludarabine + plasma exchange b. Bone marrow transplant		a. 0/1 b. 0/1
Latov 1999	Case reports	2	2	2	Rituximab	‐	1/2
Levine 1999	Probably included in Pestronk 2003 ‐ unclear	‐	‐	‐	Rituximab	‐	‐
Wilson 1999	Open prospective	4	4	4	Fludarabine	‐	4/4
Nobile‐Orazio 2000	Retrospective case series	25	25	25	a. Plasma exchange b. Plasma exchange + chlorambucil c. Prednisolone d. IVIg e. Cyclophosphamide		a. 2/5 b. 2/2 c. 0/6 d. 0/2 e. 1/5
Gorson 2001	Retrospective case series (NB some cases in Gorson 1997)	24	22	24	a. Plasma exchange b. IVIg c. Prednisolone d. Cyclophosphamide e. Plasma exchange + cyclophosphamide f. Interferon alfa‐2a g. Chlorambucil h. Azathioprine	‐	a. 8/20 b. 3/19 c. 0/8 d. 3/6 e. 2/8 f. 1/8 g. 0/2 h. 0/2
Tagawa 2001	Case report	1	1	1	Plasmapheresis + cyclophosphamide	‐	1/1
Canavan 2002	Retrospective case reports (abstract)	3	3	2	Rituximab	‐	1/2
Ghosh 2002	Case report	1	1	1	Cladribine	‐	1/1
Pestronk 2003	Prospective open study	21	21	7	Rituximab	‐	Unclear ‐ improvement
Renaud 2003	Prospective open phase II study	9	9	9	Rituximab	‐	6/9
Barohn 2005	Uncontrolled case series	5	5	3	Rituximab	0/5	0/3
Benedetti 2005	Uncontrolled ?prospective (abstract)	13	13	13	Rituximab	‐	8/13
Broglio 2005	Case report	1	1	1	Rituximab	‐	Worse
Hamidou 2005	Uncontrolled case series	9	9	‐	Cyclophosphamide	7/9 improved, 2 stable	‐
Kilidireas 2006	Uncontrolled case series	4	4	2	Rituximab	3/4 improved
Niermeijer 2006	Open prospective	16	16	6	Fludarabine	3/10 improved, all others stabilised	2/6 improved all others stabilised
Renaud 2006	Prospective follow‐up study to 2003	8	8	8	High‐dose rituximab	‐	4/7 improved (1 death)
Gironi 2006	Case report	1	1	1	Rituximab	‐	Worse
Noronha 2006	Case report	1	1		Rituximab	‐	Worse
Niermeijer 2009	Case series	17	17	6	Rituximab	Improvement in: ODSS 2/17, MRC sum score 11/17, Sensory sum score 10/17	3/6
Delmont 2010	Case report	1	1	0	Rituximab (after IVIg, corticosteroids, chlorambucil)	1	N/A
Gruson 2011	Case series	5	5 (2 MGUS, 3 WM)	5	Rituximab and fludarabine	4/5 improved clinically, electrophysiologically, IgM level and anti‐MAG titre	All MAG
Smith 2011	Phase II non‐randomised case series	21	21	N/K	Rituximab	NIS significant (>10 pts) improvement at 6 months 13/21, and 24 months 8/21	N/A

Abs: antibodies; anti‐MAG; anti‐myelin‐associated glycoprotein; IgM: immunoglobulin M; IVIg; intravenous immunoglobulin; pts: participants; Rx: treatment; WM: Waldenstrom's macroglobulinemia

Plasma exchange and apheresis

Dyck 1991 compared plasma exchange to sham exchange in a parallel‐group RCT with 39 participants, 21 of whom had an IgM paraprotein. We excluded this study, as there were no criteria specified for demyelination and the anti‐MAG status was not clear. Dyck 1991 found a significant improvement in the weakness component of the Neuropathy Disability Score (now known as the Neuropathy Impairment Score) in favour of plasma exchange in the IgG and IgA, but not the IgM subgroups. Functional changes were not reported. In eight case series of plasma exchange as monotherapy, improvement was reported in 24 of 48 cases of IgM paraprotein‐associated neuropathy (Ellie 1996; Ernerudh 1986; Ernerudh 1992; Frayne 1985; Gorson 1997; Gorson 2001; Haas 1988; Hafler 1986; Latov 1988; Nobile‐Orazio 2000; Smith 1987) (see Table 3). When described, the improvement was sustained at between eight and 36 months. Siciliano 1994 reported benefit in one person with an IgM paraprotein treated with selective apheresis, and Niemierko 1999 reported another. Improvement with plasma exchange in combination with either pulsed intravenous cyclophosphamide, chlorambucil, protein A immunoadsorption, melphalan or adriamycin, with or without steroid, has been reported in 25% to 100% of participants, all in small studies (one to eight participants per treatment group) with follow‐up periods ranging from two to 34 months (Bland 1985; Blume 1995; Dubas 1987; Gorson 2001; Kelly 1988; Latov 1980; Latov 1988; Meier 1984; Nobile‐Orazio 2000; Oksenhendler 1995; Rudnicki 1998; Sherman 1984; Smith 1987; Stefansson 1983; Tagawa 2001).

Corticosteroids

Less information is available regarding response to steroid therapy. Four of five people treated with pulsed high‐dose intravenous dexamethasone improved, but the incidence of psychiatric side effects (three of six, one with IgG MGUS) was unacceptably high (Notermans 1997). Oral prednisolone alone objectively improved only three of 30 people treated with monotherapy (Cook 1990; Dalakas 1981; Donofrio 1989; Ernerudh 1992; Gorson 1997; Gorson 2001; Hafler 1986; Latov 1988; Melmed 1983; Nobile‐Orazio 2000). Corticoteroids have been used in combination with azathioprine, cyclophosphamide, chlorambucil, and plasma exchange in several small studies, with improvement or stabilisation in 0% to 100% of participants at 14 to 54 months follow‐up (Dalakas 1981; Donofrio 1989; Ellie 1996; Ernerudh 1992; Kelly 1988; Latov 1988; Melmed 1983; Niermeijer 2007; Nobile‐Orazio 1988; Nobile‐Orazio 2000; Notermans 1996; Stefansson 1983).

Intravenous immunoglobulin

We have described the two RCTs of intravenous immunoglobulin (IVIg) versus placebo above (Comi 2002: Dalakas 1996), which provide low‐quality evidence for very short‐term improvement. Six other uncontrolled studies reported transient improvement, in 22 of 50 participants, with IVIg (Cook 1990; Ellie 1996; Gorson 1997; Gorson 2001; Hoang‐Xuan 1993; Léger 1994), whereas another did not report improvement (Nobile‐Orazio 2000).

Interferon alfa‐2a

A well‐performed randomised, double‐blind, placebo‐controlled study of 24 people (Mariette 2000) confirmed the efficacy of interferon alfa‐2a as suggested by an earlier open study (see above). One of eight people treated with interferon alfa‐2a in a study by Gorson et al improved (Gorson 2001). This participant remained stable (mild bilateral foot drop and pinprick and vibration loss in the feet, mRS 1) off all therapy for three years. Furthermore, the IgM paraprotein and anti‐MAG titre became unrecordable. A group of seven people treated with interferon alfa‐2a, and assessed with a novel measure of postural stability, improved (Pouget 2000).

Cytotoxic therapies

Chemotherapeutic (or cytotoxic) therapies have been used singly or in combination with other drugs, but none (except Niermeijer 2007 using cyclophosphamide and corticosteroids, as above) in a controlled trial. These agents include cyclophosphamide (Blume 1995; Gorson 1997; Gorson 2001; Hafler 1986; Hamidou 2005; Kelly 1988; Niermeijer 2007; Nobile‐Orazio 2000; Tagawa 2001), fludarabine (29 participants, 11 anti‐MAG and nine of these with clinical improvement) (Niermeijer 2006; Rudnicki 1998; Sherman 1994; Wilson 1999), fludarabine and rituximab (Gruson 2011, see below), cladribine (one participant) (Ghosh 2002), azathioprine (Gorson 2001), mycophenolate (Gorson 2004), chlorambucil alone (Andres 2001; Gorson 2001; Latov 1988; Nobile‐Orazio 2000), and melphalan and chlorambucil (Ernerudh 1992). Responses to treatment were variable. More extensive chemotherapy has been used in non‐MGUS associated anti‐MAG neuropathies, which are outside the scope of this review (Andres 2001).

Rituximab

Rituximab has been used in multiple cases in non‐randomised cohorts and cases series (Barohn 2005; Benedetti 2005; Benedetti 2007; Benedetti 2008; Canavan 2002; Delmont 2010; Goldfarb 2005; Gono 2006; Gorson 2007; Kelly 2006; Kilidireas 2006; Latov 1999; Levine 1999; Niermeijer 2009; Pestronk 2003; Renaud 2003; Renaud 2006; Smith 2011; Weide 2000). About 50% to 60% of participants seem to respond in these uncontrolled studies, but two studies failed to show any benefit in five participants (Barohn 2005; Rojas‐García 2003), and three studies reported worsening (Broglio 2005; Gironi 2006; Noronha 2006). Three fully‐published studies included 42 participants with anti‐MAG neuropathy treated with rituximab and documented improvements in strength, neurophysiological indices, and functional score up to two years (Dalakas 2009; Pestronk 2003; Renaud 2003). However, neither Pestronk 2003 nor Renaud 2003 were adequately controlled. Renaud 2003 had no controls and Pestronk 2003 had a semi‐historical non‐randomised control group without blinding of the assessment of outcome measures. We describe the first two RCTs of rituximab above (Dalakas 2009; Léger 2013)

A recent study describes the use of rituximab in combination with fludarabine in five participants, four of whom improved clinically and electrophysiologically, with serum IgM and anti‐MAG titre responses too. Improvement was sustained and, in this small series, treatment was without significant toxicity (Gruson 2011).

Others

Some more novel therapies have been tried, including ciclosporin in two participants (Hodgkinson 1990), and a single participant underwent autologous stem cell transplantation (ASCT) (Rudnicki 1998). In this one person, the ASCT was followed after two years by improvements in symptoms, signs, and neurophysiological indices, although he had been treated with other agents including fludarabine prior to the transplant. Neither ciclosporin nor ASCT have been subjected to a RCT.

Economic considerations

The treatments discussed are all expensive. In 2005, the approximate cost of IVIg was about GBP 3600 for the standard 2.0 g/kg dose in a 70 kg adult, and in 2012 about GBP 4500. Fludarabine and rituximab cost approximately GBP 5000 per course, and an autologous stem cell transplant costs approximately GBP 25,000 to 35,000. We cannot overstate the importance of subjecting such agents to early and adequate clinical trials.

Figure 1

PRISMA chart for searched to included studies of IgM paraproteinaemic neuropathy

Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Figure 3

Forest plot of comparison: 3 Rituximab versus placebo, outcome: 3.1 Number of participants improved on INCAT score (see text) at 8‐12 months.

Analysis 1.1

Comparison 1 IVIg versus interferon alfa‐2a, Outcome 1 Change in Clinical Neuropathy Disability Score (CNDS) at six months.

Analysis 1.2

Comparison 1 IVIg versus interferon alfa‐2a, Outcome 2 Subjective score at six months.

Analysis 1.3

Comparison 1 IVIg versus interferon alfa‐2a, Outcome 3 Number of participants improved by at least 20% on NIS at 6 months.

Analysis 2.1

Comparison 2 Interferon alfa‐2a versus placebo, Outcome 1 Change in the CNDS at 6 months (maximum 93).

Analysis 2.2

Comparison 2 Interferon alfa‐2a versus placebo, Outcome 2 Number improved in CNDS by 20% at 6 months.

Analysis 3.1

Comparison 3 Rituximab versus placebo, Outcome 1 Number of participants improved on INCAT score (see text) at 8‐12 months.

Analysis 3.2

Comparison 3 Rituximab versus placebo, Outcome 2 Mean improvement in INCAT score (see text) at 8 ‐ 12 months.

Analysis 3.3

Comparison 3 Rituximab versus placebo, Outcome 3 Rituximab vs placebo. Mean improvement in INCAT score at 8 ‐ 9 months (post hoc data).

Analysis 3.4

Comparison 3 Rituximab versus placebo, Outcome 4 Mean improvement in NIS at 12 months.

Analysis 3.5

Comparison 3 Rituximab versus placebo, Outcome 5 Improvement in 10‐metre walk time at 8 ‐ 12 months.

Analysis 3.6

Comparison 3 Rituximab versus placebo, Outcome 6 Number improved in 10‐metre walk at 6 months.

Analysis 3.7

Comparison 3 Rituximab versus placebo, Outcome 7 Change in IgM level 8 months after treatment.

Analysis 3.8

Comparison 3 Rituximab versus placebo, Outcome 8 Change in IgM anti‐MAG titre at 8 ‐ 12 months.

Analysis 3.9

Comparison 3 Rituximab versus placebo, Outcome 9 Participant subjective impression of change stable or improved at 8 ‐ 12 months.

Analysis 3.10

Comparison 3 Rituximab versus placebo, Outcome 10 Participant subjective impression of change improved at 8 ‐ 12 months.

Analysis 3.11

Comparison 3 Rituximab versus placebo, Outcome 11 Mean change in SF36 physical subscores at 12 months.

Analysis 3.12

Comparison 3 Rituximab versus placebo, Outcome 12 Mean change in SF36 mental health subscores at 12 months.

Analysis 3.13

Comparison 3 Rituximab versus placebo, Outcome 13 Any adverse event.

Analysis 3.14

Comparison 3 Rituximab versus placebo, Outcome 14 Severe adverse event.

Analysis 4.1

Comparison 4 Cyclophosphamide and steroids versus placebo, Outcome 1 Improvement in modified Rankin Scale at 6 months.

Analysis 4.2

Comparison 4 Cyclophosphamide and steroids versus placebo, Outcome 2 Number of participants improved on modified Rankin Scale at 6 months.

Analysis 4.3

Comparison 4 Cyclophosphamide and steroids versus placebo, Outcome 3 Improvement in Rivermead Mobility Index (subjective participant score) at 6 months.

Summary of findings for the main comparison. Rituximab versus placebo for treating IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathy

Should rituximab versus placebo be used for IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathies?
Patient or population: people with IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathies Setting: hospital and outpatient treatment centres Intervention: rituximab Comparison: placebo
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)	Comments
	Risk with placebo	Risk with rituximab
Number of participants improved in disability Assessed with: INCAT score or INCAT Leg Disability Score Follow‐up: range 8 to 12 months¹	Study population		RR 3.51 (1.30 to 9.45)	73 (2 RCTs)	⊕⊕⊝⊝ low²	Statistically significant effect in meta‐analysis
	74 per 1000	260 per 1000 (96 to 700)
Mean improvement in disability Assessed with: INCAT score or INCAT leg disability score Scale from: 0 to 12 in INCAT and 0 to 7 in INCAT leg disability. Follow‐up: range 8 to 12 months¹	The mean improvement in INCAT score (see text) at 8 to 12 months was ‐0.18	The mean improvement in INCAT score (see text) at 8 to 12 months in the intervention group was 0.45 lower (0.85 lower to 0.05 lower)	‐	73 (2 RCTs)	⊕⊕⊝⊝ low²	Statistically significant improvement versus placebo but probably less than MCID
Improvement in 10‐metre walk time at 8 to 12 months¹	The mean improvement in 10‐metre walk time at 8 to 12 months was 0.14 seconds	The mean improvement in 10‐metre walk time at 8 to 12 months in the intervention group was 0.35 seconds more (1.89 more to 1.19 less)	‐	68 (2 RCTs)	⊕⊕⊕⊝ moderate⁴	MCID for 10‐metre walk approximately 0.4 seconds. Borders on clinically significant improvement, but wide CIs
Participant subjective impression of change stable or improved at 8 to 12 months Assessed with: VAS from: 0 to 10 Follow‐up: mean 8 months¹	Study population		RR 1.86 (1.27 to 2.71)	70 (2 RCTs)	⊕⊕⊝⊝ low³	Patient global impression of change improved similarly in both studies in a time‐dependent manner (see data tables)
	447 per 1000	832 per 1000 (568 to 1000)
Change in serum IgM paraprotein concentration 8 months after treatment¹	The mean change in IgM level 8 months after treatment was 32.3 mg/L	The mean change in IgM level 8 months after treatment in the intervention group was 287.7 mg/L lower (328.98 lower to 244.42 lower)	‐	26 (1 RCT)	⊕⊕⊕⊝ moderate⁴	An unsurprising reduction in IgM in the rituximab‐treated group
Any adverse event	Study population		RR 1.18 (0.84 to 1.66)	80 (2 RCTs)	⊕⊝⊝⊝ very low⁵	No statistically significant difference in adverse effects. Serious adverse effects too few to make comment. Consistency of adverse event collection always suspect
	561 per 1000	662 per 1000 (471 to 931)
*The risk in the intervention group (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; IgM: immunoglobulin M; INCAT score: Inflammatory Neuropathy Cause and Treatment score;MCID: minimum clinically important difference; RR: risk ratio; VAS: visual analogue scale
GRADE Working Group grades of evidence High quality: We are very confident that the true effect lies close to that of the estimate of the effect Moderate quality: We are moderately confident in the effect estimate: The true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different Low quality: Our confidence in the effect estimate is limited: The true effect may be substantially different from the estimate of the effect Very low quality: We have very little confidence in the effect estimate: The true effect is likely to be substantially different from the estimate of effect
¹The outcomes in the rituximab study were recorded at between 8 and 12 months, which is rational for the treatment of a disease with rituximab. Thus we have reported this interval for this intervention (see text). ²Downgraded twice for imprecision (small underpowered studies) and unresponsive outcome measures. ³Downgraded twice for imprecision (small underpowered studies) and indirectness (patient impression of change). ⁴Downgraded once for imprecision (small underpowered studies). ⁵Downgraded three times for imprecision (small studies and small even numbers), trial design and inconsistent adverse event recording.

Summary of findings for the main comparison. Rituximab versus placebo for treating IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathy

Table 1. 'Summary of findings' table: Intravenous immunoglobulin (IVIg) versus placebo

IVIg versus placebo for IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathies
Patient or population: people with IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathies Settings: hospital and outpatient treatment centres Intervention: IVIg versus placebo
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)	Comments
Outcomes	Risk with placebo	Risk with IVIg	Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)	Comments
Number of participants improved in disability score at 6 months ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Short‐term outcomes only, not reported at 6 months
Mean improvement in disability score at 6 months ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Short‐term outcomes only, not reported at 6 months
Improvement in 10‐metre walk time at 6 months ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Short‐term outcomes only, not reported at 6 months
Participant subjective impression of change at 6 months ‐ number of participants reporting improvement ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Short‐term outcomes only, not reported at 6 months
Change in serum IgM paraprotein concentration 8 months after treatment ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Short‐term outcomes only, not reported at 6 months
Any adverse event	Data for the anti‐MAG subgroup were not available. In one study, one participant had an aseptic meningitis and a rash after IVIg treatment, with 2 moderate adverse events in the placebo group. In the other trial, there were no serious adverse events with IVIg, but mild and transient effects were "more common in the IVIg than the placebo group".		Not estimable	‐	See comment	No further analysis was possible.
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
¹Incomplete reporting of adverse events in one study.

Table 1. 'Summary of findings' table: Intravenous immunoglobulin (IVIg) versus placebo

Table 2. 'Summary of findings' table: Cyclophosphamide and prednisone versus placebo

Cyclophosphamide and corticosteroids compared to placebo for IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathies
Patient or population: people with IgM anti‐myelin‐associated glycoprotein paraprotein‐associated peripheral neuropathies Settings: hospital and outpatient treatment centres Intervention: cyclophosphamide and corticosteroids Comparison: placebo
Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)	Comments
	Risk with placebo	Risk with cyclophosphamide and corticosteroids
Number of participants improved in disability score at 6 months Modifed Rankin Scale Scale from: 0 to 6 (normal to death) Follow‐up: mean 6 months	Study population		RR 3.56 (0.41 to 30.99)	35 (1 study)	⊕⊕⊕⊝ moderate¹
	53 per 1000	189 per 1000 (22 to 1000)
Mean improvement in disability score at 6 months Modifed Rankin Scale Scale from: 0 to 6 (normal to death) Follow‐up: 6 months²	The mean improvement in disability score at 6 months in the control groups was 0.11 units	The mean improvement in disability score at 6 months in the intervention groups was 0.31 lower (0.61 lower to 0.01 lower)		34 (1 study)	⊕⊕⊕⊝ moderate¹
Improvement in 10‐metre walk time at 6 months ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Not reported
Participant subjective impression of change at 6 months ‐ number of participants reporting improvement ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Not reported
Change in serum IgM paraprotein concentration after treatment ‐ not reported	See comment	See comment	Not estimable	‐	See comment	Not reported
Any adverse event ‐ not measured	See comment	See comment	Not estimable	‐	See comment	Adverse events not specifically collected but text notes in the paper (see text of review)
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; IgM: immunoglobulin M; RR: risk ratio
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
¹Dowgraded once for imprecision (single small study). ²Phase 1 of cross‐over trial only.

Table 2. 'Summary of findings' table: Cyclophosphamide and prednisone versus placebo

Table 3. Details of non‐randomised studies

Study ID	Study design	Particiant number (total)	Participant number (IgM)	Participant number (MAG)	Intervention	Number improved (IgM)	Number improved (MAG)
Latov 1980	Case report	1	1	1 (anti‐myelin Abs)	Prednisolone + chlorambucil + plasma exchange	1/1	‐
Dalakas 1981	Uncontrolled case series (response to Rx recorded in Latov 1988)	11	4	?	a. Prednisolone b. Prednisolone + chlorambucil	a. 0/1 b. 1/1	‐
Melmed 1983	Uncontrolled case series (response to Rx recorded in Latov 1988)	3	3	3	a. Corticosteroids b. Steroids + plasma exchange + chlorambucil	‐	a. 3/3 b. 0/3
Stefansson 1983	Case report	1	1	1	Plasmapheresis + immunosuppressant	‐	0/1
Meier 1984	Case report	1	1	1	Plasmapheresis + prednisolone + chlorambucil	‐	1/1
Sherman 1984	Retrospective case series	10	6	?	Plasmapheresis + chlorambucil	2/5	‐
Bland 1985	Case report	1	1	1	Plasmapheresis + chlorambucil	‐	1/1
Frayne 1985	2 cases	2	2	‐	Plasma exchange	2/2	‐
Ernerudh 1986	Retrospective non‐randomised	3	3	3 (anti‐myelin Abs)	Plasma exchange	2/3 (temporary)	‐
Hafler 1986	Uncontrolled case series (response to Rx recorded in Latov 1988)	9	9	7	a.Corticosteroids b. Plasma exchange c. Cyclophosphamide	‐	a. 0/5 b. 1/5 c. 1/1
Smith 1987	Prospective non‐randomised case series	13	8 (2 lymphoma)	4 (1 lymphoma)	a. Chlorambucil + plasma exchange b. Plasma exchange	a. 3/4 b. 1/2	a. 1/2 b. 1/1
Haas 1988	Case report	1	1	1	Plasma exchange	1/1	1/1
Kelly 1988	Uncontrolled case series	10	10	5	Prednisolone + cyclophosphamide (azathioprine or chlorambucil) + plasma exchange	4/4 (all 5 not treated)	2/3 (All 5 not treated)
Latov 1988	Uncontrolled case series	11	11	11	a. Plasmapheresis b. Prednisolone c. Chlorambucil d. Prednisolone + plasma exchange e. Plasma exchange + chlorambucil	‐	a. 3/3 b. 0/1 c. 3/4 d. 0/1 e. 0/1
Donofrio 1989	Case report	1	1	0	a. Corticosteroids b. Plasma exchange + corticosteroids	a. 0/1 b. 1/1	‐
Cook 1990	Case reports	2	2	1	a. IVIg b. prednisolone	a. 2/2	a. 1/1 b. 0/1
Hodgkinson 1990	Uncontrolled case series	8	2	?	Ciclosporin	2/2	‐
Ernerudh 1992	Prospective non‐randomised	5	5	5 (anti‐myelin Abs)	a. Plasma exchange b. Prednisolone c. Melphalan d. Chlorambucil e. Chlorambucil and plasma exchange f. Chlorambucil + prednisolone	a. 1/1 b. 1/2 c. 0/1 d. 0/1 e. 1/1 f. 1/2	‐
Hoang‐Xuan 1993	Open prospective	4	4	3 (2 in Léger 1994)	IVIg	‐	0/1
Léger 1994	Open prospective	7	7	5	IVIg	‐	2/5
Sherman 1994	Abstract	10	10	8	Fludarabine	‐	7/8
Siciliano 1994	Case report	3	1	?	Selective apheresis	1/1	‐
Blume 1995	Uncontrolled case series	4	3	4	Plasma exchange + cyclophosphamide	3/3	4/4
Ellie 1996	Retrospective case series	33	33	33	a. Plasma exchange b. IVIg	a. 2/6 b. 13/17	‐
Notermans 1996	Open prospective	16	11	5	Cyclophosphamide + prednisolone	‐	Unclear
Gorson 1997	Retrospective case series	15	12	8	a. Plasma exchange b. Prednisolone c. IVIg d. Cyclophosphamide + plasma exchange	a. 10/12? b. 1/2 c. 2/6 d. 2/2	Unclear
Rudnicki 1998	Case report	1	1	1	a. Fludarabine + plasma exchange b. Bone marrow transplant		a. 0/1 b. 0/1
Latov 1999	Case reports	2	2	2	Rituximab	‐	1/2
Levine 1999	Probably included in Pestronk 2003 ‐ unclear	‐	‐	‐	Rituximab	‐	‐
Wilson 1999	Open prospective	4	4	4	Fludarabine	‐	4/4
Nobile‐Orazio 2000	Retrospective case series	25	25	25	a. Plasma exchange b. Plasma exchange + chlorambucil c. Prednisolone d. IVIg e. Cyclophosphamide		a. 2/5 b. 2/2 c. 0/6 d. 0/2 e. 1/5
Gorson 2001	Retrospective case series (NB some cases in Gorson 1997)	24	22	24	a. Plasma exchange b. IVIg c. Prednisolone d. Cyclophosphamide e. Plasma exchange + cyclophosphamide f. Interferon alfa‐2a g. Chlorambucil h. Azathioprine	‐	a. 8/20 b. 3/19 c. 0/8 d. 3/6 e. 2/8 f. 1/8 g. 0/2 h. 0/2
Tagawa 2001	Case report	1	1	1	Plasmapheresis + cyclophosphamide	‐	1/1
Canavan 2002	Retrospective case reports (abstract)	3	3	2	Rituximab	‐	1/2
Ghosh 2002	Case report	1	1	1	Cladribine	‐	1/1
Pestronk 2003	Prospective open study	21	21	7	Rituximab	‐	Unclear ‐ improvement
Renaud 2003	Prospective open phase II study	9	9	9	Rituximab	‐	6/9
Barohn 2005	Uncontrolled case series	5	5	3	Rituximab	0/5	0/3
Benedetti 2005	Uncontrolled ?prospective (abstract)	13	13	13	Rituximab	‐	8/13
Broglio 2005	Case report	1	1	1	Rituximab	‐	Worse
Hamidou 2005	Uncontrolled case series	9	9	‐	Cyclophosphamide	7/9 improved, 2 stable	‐
Kilidireas 2006	Uncontrolled case series	4	4	2	Rituximab	3/4 improved
Niermeijer 2006	Open prospective	16	16	6	Fludarabine	3/10 improved, all others stabilised	2/6 improved all others stabilised
Renaud 2006	Prospective follow‐up study to 2003	8	8	8	High‐dose rituximab	‐	4/7 improved (1 death)
Gironi 2006	Case report	1	1	1	Rituximab	‐	Worse
Noronha 2006	Case report	1	1		Rituximab	‐	Worse
Niermeijer 2009	Case series	17	17	6	Rituximab	Improvement in: ODSS 2/17, MRC sum score 11/17, Sensory sum score 10/17	3/6
Delmont 2010	Case report	1	1	0	Rituximab (after IVIg, corticosteroids, chlorambucil)	1	N/A
Gruson 2011	Case series	5	5 (2 MGUS, 3 WM)	5	Rituximab and fludarabine	4/5 improved clinically, electrophysiologically, IgM level and anti‐MAG titre	All MAG
Smith 2011	Phase II non‐randomised case series	21	21	N/K	Rituximab	NIS significant (>10 pts) improvement at 6 months 13/21, and 24 months 8/21	N/A
Abs: antibodies; anti‐MAG; anti‐myelin‐associated glycoprotein; IgM: immunoglobulin M; IVIg; intravenous immunoglobulin; pts: participants; Rx: treatment; WM: Waldenstrom's macroglobulinemia

Table 3. Details of non‐randomised studies

Comparison 1. IVIg versus interferon alfa‐2a

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Change in Clinical Neuropathy Disability Score (CNDS) at six months Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

2 Subjective score at six months Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

3 Number of participants improved by at least 20% on NIS at 6 months Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

Comparison 1. IVIg versus interferon alfa‐2a

Comparison 2. Interferon alfa‐2a versus placebo

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Change in the CNDS at 6 months (maximum 93) Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

2 Number improved in CNDS by 20% at 6 months Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

Comparison 2. Interferon alfa‐2a versus placebo

Comparison 3. Rituximab versus placebo

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Number of participants improved on INCAT score (see text) at 8‐12 months Show forest plot	2	73	Risk Ratio (M‐H, Fixed, 95% CI)	3.51 [1.30, 9.45]

2 Mean improvement in INCAT score (see text) at 8 ‐ 12 months Show forest plot	2	73	Mean Difference (IV, Fixed, 95% CI)	‐0.45 [‐0.85, ‐0.05]

3 Rituximab vs placebo. Mean improvement in INCAT score at 8 ‐ 9 months (post hoc data) Show forest plot	2	70	Mean Difference (IV, Fixed, 95% CI)	‐0.33 [‐0.73, 0.07]

4 Mean improvement in NIS at 12 months Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

5 Improvement in 10‐metre walk time at 8 ‐ 12 months Show forest plot	2	68	Mean Difference (IV, Fixed, 95% CI)	‐0.35 [‐1.89, 1.19]

6 Number improved in 10‐metre walk at 6 months Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

7 Change in IgM level 8 months after treatment Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

8 Change in IgM anti‐MAG titre at 8 ‐ 12 months Show forest plot	2	71	Mean Difference (IV, Fixed, 95% CI)	‐17.79 [‐33.33, ‐2.25]

9 Participant subjective impression of change stable or improved at 8 ‐ 12 months Show forest plot	2	70	Risk Ratio (M‐H, Fixed, 95% CI)	1.86 [1.27, 2.71]

10 Participant subjective impression of change improved at 8 ‐ 12 months Show forest plot	2	70	Risk Ratio (M‐H, Fixed, 95% CI)	9.67 [1.84, 50.85]

11 Mean change in SF36 physical subscores at 12 months Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

12 Mean change in SF36 mental health subscores at 12 months Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

13 Any adverse event Show forest plot	2	80	Risk Ratio (M‐H, Fixed, 95% CI)	1.18 [0.84, 1.66]

14 Severe adverse event Show forest plot	2	80	Risk Ratio (M‐H, Fixed, 95% CI)	3.11 [0.34, 28.54]

Comparison 3. Rituximab versus placebo

Comparison 4. Cyclophosphamide and steroids versus placebo

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Improvement in modified Rankin Scale at 6 months Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

2 Number of participants improved on modified Rankin Scale at 6 months Show forest plot	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

3 Improvement in Rivermead Mobility Index (subjective participant score) at 6 months Show forest plot	1		Mean Difference (IV, Fixed, 95% CI)	Totals not selected

Comparison 4. Cyclophosphamide and steroids versus placebo