Chronic inflammatory demyelinating polyradiculoneuropathy (CIDP)

CIDP is one of the few peripheral neuropathies that can be treated, and one of the few where the prognosis depends directly on the date the name was given. Between two neurology appointments three months apart, the physiotherapist is often the only professional to see the patient every week: it is the physiotherapist who notices that the patient can no longer rise from a chair without using their arms, that the foot catches on the stair carpet, that the handshake has weakened. This article sets out what needs to be recognised, what the EAN/PNS 2021 criteria changed, and what rehabilitation actually contributes alongside immunomodulatory treatment. One caveat from the outset: the rehabilitation literature specific to CIDP amounts to a single trial of 18 patients. We will say so rather than fill the gap with data borrowed from Guillain-Barré syndrome.

• Updated 14 August 2026
• Level Clinical summary
• Sources 31 verified references
• ICD-11 8C01.3

What is CIDP, and why did its name change in 2021?

An autoimmune disease of the peripheral nerve, whose nomenclature was overhauled by the 2021 European guidelines. Understanding what changed avoids reading a neurology report with the vocabulary of the previous decade.

A sheath under attack, not a diseased muscle

Chronic inflammatory demyelinating polyradiculoneuropathy is a dysimmune neuropathy: the immune system attacks the myelin surrounding the fibres of the peripheral nerve and, in certain forms, the proteins that structure the node of Ranvier. The nerve conducts more slowly, sometimes not at all across a segment. The muscle itself is healthy: it simply no longer receives the command properly.

This distinction is not academic for a physiotherapist. Denervation weakness is not deconditioning weakness, and it does not respond to the same reasoning. The deficit varies with the rhythm of inflammation and treatment, not with the rhythm of training. A session that appears to make a patient worse may simply coincide with a relapse. A session that appears to make them better may coincide with an immunoglobulin infusion. We will return to this at length: it is the methodological crux of this whole chapter.

The term "polyradiculoneuropathy" says something important: the disease affects the roots as much as the nerve trunks. That is precisely what explains the proximal weakness, atypical for a neuropathy, and it is where recognition of the picture is won or lost.

How many people are affected?

CIDP is rare, and published figures vary by a factor of four depending on the method. An analysis of United States claims data covering 2016 to 2023 puts the prevalence at 23.3 per 100,000 (95 % CI 23.1 to 23.5) and the incidence at 2.8 per 100,000 per year (95 % CI 2.7 to 2.9), that is around 77,000 people living with the disease in the United States3. The authors themselves flag the limitation of their source: a billing code is not a validated diagnosis, and the overdiagnosis documented elsewhere is mechanically reflected in the figure.

A population study conducted in Birmingham, in the United Kingdom, on patients actually under follow-up gives much lower values: prevalence of 6.18 per 100,000 (95 % CI 4.66 to 8.05) and mean incidence of 0.54 per 100,000 per year (95 % CI 0.404 to 0.713)4. That same study reports a lower prevalence in people of Indian, Pakistani or Bangladeshi origin after the age of 50 (2.64 against 10.15 per 100,000), with no difference between 18 and 49 years.

The gap between those two figures is not a methodological detail: it says that we do not know precisely how many people have CIDP, because we do not know precisely to whom the diagnosis is correctly assigned. For practice, keep the order of magnitude in mind: a community practice will not see several CIDP cases a year. It may see one, and perhaps before the neurologist does.

What 2021 changed in the vocabulary

The joint guidelines of the European Academy of Neurology and the Peripheral Nerve Society, published in 2021, revised the 2010 consensus and introduced two changes of vocabulary that can be read directly in neurology reports1.

The first: the phrase "atypical CIDP" has gone in favour of "CIDP variants". The reason is explicit in the text: these forms are well-characterised entities, not irregularities. There are five: multifocal, focal, distal, motor, sensory.

The second: the levels of diagnostic certainty have gone from three to two. Where definite, probable and possible CIDP used to be distinguished, only "CIDP" and "possible CIDP" now are. The justification is methodological and worth knowing: the diagnostic accuracy of the criteria for "probable" and "definite" did not differ significantly, so the distinction carried no information.

These variants are not rare. In a consecutive Japanese series of 33 patients diagnosed between 2020 and 2025 using the 2021 criteria, 36.4 % were typical CIDP and 63.6 % variants : distal in 39.4 % of cases, multifocal in 15.2 %, sensory in 6.0 %, motor in 3.0 %5. This is a referral centre, so recruitment is biased towards difficult cases, and the authors say as much. But the lesson holds: a picture that does not look like the textbook portrait does not rule out CIDP.

How do you recognise the clinical picture in the clinic?

The physiotherapist does not make the diagnosis of CIDP. But they see the patient every week, they see them stand up, walk, grasp. This chapter describes what, in what they observe, should prompt a letter.

Proximal weakness, the signature

Most of the polyneuropathies met in practice are length-dependent: they begin with the longest fibres, so with the toes, and slowly work upwards. The patient describes tingling in a stocking distribution, then a glove one. Strength, when it is affected, is affected distally: the foot dorsiflexors, the intrinsic muscles of the hand.

CIDP does not work that way. Because it affects the roots, it strikes the proximal at the same time as the distal. The patient can no longer rise from a low chair without using their hands, struggles to climb stairs, can no longer reach an object above their head. And they also have numb toes and weak dorsiflexors. It is this coexistence that should raise the alarm, because no ordinary polyneuropathy produces it.

Proximal weakness is not one sign among others: it is the sign whose absence led nearly half of the false positives to be wrongly labelled, and whose presence was constant in every missed true positive.

This is not a clinical hunch, it is a numerical result. In the retrospective study conducted at the Erasmus centre in Rotterdam between 2011 and 2017, among patients referred with a CIDP label that turned out to be wrong, 48 % had no proximal weakness and 29 % had distal weakness only. Conversely, among patients referred with another diagnosis and in whom CIDP was ultimately retained, 100 % had proximal weakness7.

Symmetry, areflexia, the sensory side

Three further features complete the picture of the typical form.

Symmetry. Weakness is broadly symmetrical. In the same Rotterdam series, 20 % of the weaknesses in overdiagnosed patients were asymmetrical7. One caution, though: asymmetry does not rule out CIDP, it points towards the multifocal variant, which is precisely the one that takes longest to recognise.

Areflexia. Tendon reflexes are reduced or absent, and not only at the ankles: in all four limbs. A brisk knee jerk in a patient suspected of CIDP is an inconsistency worth flagging.

Sensory involvement. It affects at least two limbs in the typical form. It is often in the background of the patient's account, which starts with strength, but it has to be sought: vibration sense at the malleolus, position sense of the big toe, discrimination. Proprioceptive ataxia may dominate the picture, and the patient then says they "lose their balance in the dark" or in the shower.

What the physiotherapist sees and the consultation does not

A patient is seen in neurology every three to six months. They are seen in physiotherapy two to three times a week. That difference in frequency is the physiotherapist's main asset in this condition, and it is underused.

What is visible in a session and not in a consultation: the speed of a sit-to-stand, the number of steps before stopping, the fact that a patient now takes the handrail with both hands, an object dropped from the right hand that they put down to tiredness. It is these observations, if they are dated and quantified, that make it possible to tell the neurologist not "he seems less well to me" but "his MRC sum score has gone from 56 to 51 in six weeks, and he no longer does the five sit-to-stands he was doing in March".

Fatigue deserves a mention of its own. It is common in CIDP and is not only motor. A study comparing 47 patients with 31 controls finds total, motor and cognitive fatigue significantly higher in the patients, the motor component being the most marked. The only predictors of both components are functional disability and depressive symptoms, with no link to the progression biomarker tested26. A recent review broadens this finding to a whole set of non-sensorimotor symptoms: dysautonomia, circadian rhythm disturbance, mood, subtle cognition27. An exhausted patient is therefore not necessarily a patient who is deconditioning.

Why does the 8-week threshold separate CIDP from Guillain-Barré?

The two diseases share a mechanism, dysimmune demyelination, and much of the clinical picture. What separates them is a curve over time. And that is exactly what weekly follow-up allows you to plot.

The nadir, not the onset

Guillain-Barré syndrome is an acute, monophasic disease: worsening is rapid, maximum severity (the nadir) is reached in under 4 weeks, and recovery then begins. CIDP, in its typical form, progresses or relapses over at least 8 weeks : that is the temporal criterion of the European guidelines16.

So it is not the speed of onset that settles it, it is the date at which the patient stops getting worse. A picture that sets in over ten days may be either. A picture still worsening in the third month is not a Guillain-Barré.

The grey zone: acute-onset CIDP

There is a form of CIDP that begins like a Guillain-Barré, with a nadir reached in under 8 weeks: acute-onset CIDP. A recent review reports a proportion of the order of 16 % of CIDP cases in one of the series it cites6, while a European survey reports, citing the earlier literature, an acute onset in under 4 weeks in up to 13 % of patients8. The two figures come from different sources and do not measure quite the same thing; what they agree on is that this is not anecdotal.

This form reveals itself after the initial treatment. The patient receives immunoglobulin for their presumed Guillain-Barré, improves, then deteriorates again. Two readings are then possible: a treatment-related fluctuation, where the effect of the immunoglobulin wears off before the disease burns out, or a CIDP resuming its course. A literature review devoted to this distinction concludes that it remains difficult, that the electrophysiological, ultrasound and immunological criteria proposed have not settled it, and that the very concept of recurrent Guillain-Barré is disputed9.

The stakes are not nosological. Guillain-Barré is treated with a short course; CIDP requires long-term immunosuppression. Getting it wrong means letting a patient deteriorate for want of maintenance treatment. Seen from the other side of that border, the figure is of the same order: the European guideline devoted to Guillain-Barré syndrome asks that the diagnosis be reconsidered in favour of acute-onset CIDP when progression continues beyond eight weeks, which happens in about 5 % of patients initially labelled Guillain-Barré. Our article on that disease covers its recovery phase, its residual fatigue and the two deteriorations a physiotherapist may observe in it.

A patient who has come through a Guillain-Barré, referred for rehabilitation, and who starts declining again past the eighth week, does not need a programme adjustment. He needs a letter.

The last line of this table deserves a word. Diabetic polyneuropathy is the context in which a CIDP most easily goes unnoticed, because the patient already has an explanation for their tingling. Yet diabetes is also the most frequent comorbidity in patients with CIDP in the American cohort3. Having diabetes does not immunise you against CIDP: it merely makes it more tempting to stop looking. Our article on peripheral neuropathy, notably diabetic sets out the length-dependent picture against which the comparison has to be made.

What do the EAN/PNS 2021 criteria say, and what did they change?

These criteria are the ones that appear in the neurologist's report. Reading them correctly means understanding why a patient is classed as "possible CIDP" and what that implies for what follows.

How the guidelines were built

The 2021 text is the work of a joint task force: seventeen experts, a patient representative and two Cochrane methodologists, who framed twelve questions in PICO format and then graded the evidence using the GRADE method1. The same text was published simultaneously in two journals, the Journal of the Peripheral Nervous System and European Journal of Neurology2 : a reference to either points to the same document.

Diagnosis rests on three pillars

A compatible clinical picture (the one described above), electrophysiological abnormalities of demyelination in at least two motor nerves, and supportive findings when the first two are not enough: raised CSF protein, hypertrophy of the roots or trunks on imaging or ultrasound, sensory abnormalities on nerve conduction studies, and above all the objective response to treatment.

This last point is central and often misunderstood. Improvement after immunoglobulin is not proof if it is not objectively documented : the 2025 report on unmet needs in CIDP stresses that there is no consensus definition of what constitutes a "response", a "refractory" patient, a "remission" or a "relapse", and that this lack of shared vocabulary is detrimental to treatment decisions11. That is one more reason for the physiotherapist to bring figures: they are missing from the file.

What these criteria are worth, in figures

A study published in 2026 compared the diagnostic accuracy of the EFNS/PNS 2010 and EAN/PNS 2021 criteria in 339 patients with CIDP and 339 matched controls referred for suspected chronic dysimmune neuropathy who received another diagnosis10.

The results, including the "possible" category: sensitivity of 92 % for the 2010 criteria and 94 % for those of 2021, specificities of 67 % and 66 % respectively. Restricting to the full category (probable or definite in 2010, "CIDP" in 2021): sensitivity of 91 % against 83 %, specificity of 78 % against 86 %.

In other words, the 2021 revision traded sensitivity for specificity. It labels fewer patients wrongly, and misses more of them. Given the massive overdiagnosis documented elsewhere, that choice is defensible. It does, however, have a practical consequence: a "possible CIDP" patient is not a patient without CIDP, and a patient who does not meet the criteria but responds objectively to treatment remains a patient to be treated.

One last result from this study bears directly on the care pathway: a unilateral nerve conduction study would have missed 77 diagnoses of possible CIDP10. How far the testing extends matters as much as the grid used to read it.

Why is CIDP so often misdiagnosed?

Overdiagnosis and underdiagnosis coexist, and they do not have the same causes. Knowing the five recorded pitfalls tells you, in the session, which of the two is happening.

A third of labels are wrong, in both directions

The reference study on this question went back over the records of patients referred to the Erasmus centre in Rotterdam between 2011 and 2017, either with a diagnosis of CIDP or with another diagnosis later revised to CIDP, and reclassified them using the EFNS/PNS 2010 criteria7.

The diagnosis of CIDP was overturned in 32 % of patients who arrived with that label, that is 31 out of 96. In the other direction, of 81 patients who really did have CIDP, 16, that is 20 %, were referred with another diagnosis.

The two errors do not cost the same. Overdiagnosis exposes the patient to expensive immunotherapy that is not without risk, for a disease they do not have, and it delays the real diagnosis. Underdiagnosis lets a treatable disease progress towards axonal loss, which does not recover.

The five pitfalls, named

The authors of that study set out the mechanisms of the error explicitly. They are worth knowing because three of them can be spotted without any additional investigation.

1. Paying no attention to proximal weakness, when it is the hallmark of the disease.
2. Failing to recognise clinically atypical phenotypes : half the underdiagnosed patients had a variant.
3. Overrating CSF protein. A raised protein in the cerebrospinal fluid is not specific: 74 % of overdiagnosed patients had one.
4. Misinterpreting nerve conduction studies.
5. Not applying the electrodiagnostic criteria : 65 % of the overdiagnosed did not meet them.

To which an implicit sixth can be added: 97 % of overdiagnosed patients did indeed have a neuropathy, of another kind. The problem was therefore not seeing a neuropathy, but stopping there.

Seven months, nineteen investigations

A survey conducted between September 2022 and April 2023 among 542 patients and their neurologists in the United Kingdom, France, Germany, Italy and Spain gives the measure of the pathway8. Mean age 54.0 years (standard deviation 12.4), 62 % men. Mean number of diagnostic procedures per patient: 19.6 (standard deviation 9.4). Median time between first symptoms and diagnosis: 7.0 months, with an interquartile range of 3.2 to 13.0 months.

37 % of patients had received at least one wrong diagnosis along the way, and the most frequent of those wrong diagnoses was Guillain-Barré syndrome, in 37 % of cases. Three factors lengthened the delay: mild symptoms at the outset, a variant form, and having been misdiagnosed.

That last point deserves emphasis for a physiotherapist. A patient referred with subtle symptoms, an atypical form and a first label that does not fit is precisely the patient who will spend months in rehabilitation without anyone questioning the diagnosis. That is where the physiotherapist's place matters most, and where it is least often taken.

What do disease-modifying treatments do, and what do they not replace?

This chapter is not here so that the physiotherapist may prescribe anything. It is here so that they know what the treatment does, at what rhythm, and why a session held the day after an infusion does not mean the same thing as a session held the day before.

Intravenous immunoglobulin, the foundation

The Cochrane review updated in 2024 retained nine randomised trials totalling 372 participants12. Compared with placebo, intravenous immunoglobulin increases the probability of a significant improvement in disability within six weeks of starting treatment: relative risk 2.40 (95 % CI 1.72 to 3.36), that is 4 patients to treat for one additional benefit (95 % CI 3 to 5). The level of evidence is high.

On the Rankin scale, the improvement is 0.26 points (95 % CI 0.05 to 0.48), also high-quality evidence. On the INCAT scale at 24 weeks, it reaches 0.80 points (95 % CI 0.23 to 1.37), moderate evidence. The authors conclude that further placebo-controlled trials would probably not change these conclusions.

Compared with oral prednisolone, immunoglobulin probably makes little difference to the probability of improvement at four weeks (relative risk 0.91, 95 % CI 0.50 to 1.68). Compared with intravenous methylprednisolone, it probably increases the probability of improvement at two weeks, with a wide confidence interval.

Corticosteroids, widely used and barely proven

This is the most striking paradox in the file. The Cochrane review devoted to corticosteroids found only one unblinded randomised trial comparing prednisone with no treatment, in 35 eligible patients, which did not even measure the review's primary outcome13. The impairment score improved in 12 of the 19 patients on prednisone against 5 of the 16 with no treatment, a relative risk of 2.02 (95 % CI 0.90 to 4.52), with a very low.

The authors write that corticosteroids are commonly used in practice on the strength of very low quality observational data, and that those same observational data document serious adverse effects in the long term. A second trial comparing monthly high-dose oral dexamethasone with standard-dose daily prednisolone in 40 participants finds little difference in efficacy.

Despite that evidence base, the EAN/PNS 2021 guidelines place immunoglobulin or corticosteroids as initial treatment, with a strong recommendation for either1. This is not an inconsistency: the GRADE method allows a strong recommendation backed by weak evidence when the expected benefits clearly outweigh the harms, which is the case for a disabling disease with limited alternatives.

Plasma exchange, effective and short-lived

The Cochrane review devoted to plasma exchange retains two small trials. The first, a cross-over trial in 18 participants, shows at four weeks an improvement of 2 points (95 % CI 0.8 to 3.0) more than sham exchange on an 11-point disability scale. But 8 of the 12 patients who improved then deteriorated rapidly14. Adverse effects occur, according to non-randomised data, in 3 to 17 % of procedures, sometimes serious ones.

The 2021 guidelines logically reserve plasma exchange for situations where immunoglobulin and corticosteroids have failed, again with a strong recommendation1.

A point that defeats intuition: some patients can stop

A significant share of patients go into remission, whether spontaneous or treatment-related. A London team offered supervised withdrawal to clinically stable patients on immunoglobulin: of 125 people, 45 met the stability criteria, 36 were eligible for the withdrawal trial, 12 agreed, and 8 of them, that is 66.7 %, stayed stable for at least 12 months without treatment. Every patient who deteriorated returned to their previous level once treatment was restarted15.

For a physiotherapist, this means two things. First, a patient who is offered withdrawal is not being abandoned: it is a documented, reversible process. Second, it is precisely during these periods off treatment that close clinical monitoring is most valuable, and the physiotherapist is the professional who sees the patient most often.

What rehabilitation does not do

No rehabilitation modality has ever been compared with an immunomodulatory treatment in CIDP, and the only available exercise trial involved patients who were all on subcutaneous immunoglobulin. Rehabilitation is added to the treatment; at no point does it take its place.

This has to be said plainly because the opposite is sometimes said. Autoimmune demyelination is not corrected by exercise. What exercise addresses is the consequences: deconditioning, loss of strength in the muscles still innervated, apprehension when walking, cardiorespiratory deconditioning. That is useful, measurable, and it is not the same thing.

What rehabilitation for CIDP, and at what level of evidence?

The short answer is: very little evidence, and it needs saying. The long answer consists in examining exactly what we have, what can be drawn from it, and where the boundary with extrapolation lies.

The only exercise trial conducted in CIDP

A Danish team enrolled 18 patients with CIDP, all treated with subcutaneous immunoglobulin, in a protocol of three 12-week periods16 :

• 12 weeks without exercise (reference period);
• 12 weeks of aerobic training, three times a week on a cycle ergometer;
• 12 weeks of resistance training, three times a week, unilateral work in knee and elbow flexion and extension.

Results. During the exercise-free period, maximal oxygen uptake and strength did not move significantly (respectively 4.9 % ± 10.3, p = 0.80 and 3.7 % ± 10.1, p = 0.17, downwards). Aerobic training increased VO₂ max by 11.0 % ± 14.7 (p = 0.02). Resistance training increased combined isokinetic strength by 13.8 % ± 16.0 (p = 0.0004).

The exercise-free reference period is what makes this trial interpretable. It shows that the disease was neither improving nor deteriorating spontaneously in these patients over the preceding twelve weeks: the gain that follows is therefore not the natural course. It is the best methodological protection available in the absence of a control group, and it is also its limitation: without randomisation, order effects and expectation effects are not controlled for.

What is lost when you stop

A year after the end of the trial, 10 of the 17 eligible patients agreed to be reassessed17. Six had stopped training: their combined isokinetic strength had fallen by 13.0 % (range 25.8 to 2.9, p = 0.03) and their VO₂ max by 16.6 % (range 18.8 to 12.6, p = 0.06). Four had carried on, three with aerobic training and one with resistance training: in them, strength and aerobic capacity were preserved.

The numbers are tiny and the comparison is not randomised: those who carry on are not comparable to those who stop. But the message is consistent with what we know about training in general, and it has a direct practical consequence: what is not maintained is lost, and a CIDP programme has to be conceived as something lasting, not as a course of thirty sessions.

What the broader literature says, and why it is not enough

For want of trials in CIDP, one is tempted to look around. Three sources widen the field, each with its limitations.

Strengthening in polyneuropathies in general. A scoping review surveyed studies of strength training at a load of at least 70 % of the repetition maximum in patients with polyneuropathy: of 362 articles screened and 101 read in full, eight studies were retained, covering Charcot-Marie-Tooth disease, CIDP and diabetic polyneuropathy. Five randomised trials, two case series, one cross-over trial. Methodological quality was judged fair to poor for seven of them. The conclusion is explicit: strengthening may improve strength, but no recommendation can be made18.

Promoting physical activity in neuromuscular diseases. A Cochrane review retained 13 studies and 795 randomised participants, covering nine inherited or acquired neuromuscular diseases. No meta-analysis could be carried out. Ten of the thirteen studies were at high or unclear risk of bias in the collection of the physical activity outcome. The authors' conclusion is that they remain uncertain about the effectiveness of these interventions. One point deserves to be noted separately: they found no evidence that supervised activity increases the risk of a serious adverse event, while noting that the collection of such events was incomplete19.

Balance. A systematic review of six studies and 133 participants with non-diabetic peripheral neuropathies reports improvements in static and dynamic balance and a reduction in fall risk with stabilometry platforms combined with virtual reality and feedback, with good adherence. The authors stress the variability of the protocols and the small number of studies21. These populations include a range of neuropathies and not specifically CIDP.

Modalities and levels of evidence

The table below summarises what each modality can claim in CIDP. The GRADE levels given are our own appraisal from the sources cited, applying the usual downgrading criteria: small numbers, absence of randomisation, indirectness of the population.

How hard can you load without overdoing it?

This is the question every physiotherapist asks in front of a demyelinating neuropathy, and it is the one on which the literature is poorest. This chapter sets out what is documented, what is not, and an approach that takes account of both.

What do we mean when we say "overwork"?

The notion of overwork weakness, or overwork weakness, is an old one in neuromuscular rehabilitation. It describes a lasting loss of strength said to follow overly intense exercise in a partially denervated or dystrophic muscle, as opposed to the transient fatigue of normal exercise. The term is still used in the recent case literature: a Japanese team describes a patient with Lambert-Eaton myasthenic syndrome in whom independence in activities of daily living was achieved through low-intensity, high-repetition endurance work, with monitoring of subjective intensity over time, with no overwork weakness appearing31.

Two honest caveats are needed about that reference. First, it is a single case: one case demonstrates nothing about how often a phenomenon occurs. Second, Lambert-Eaton syndrome is a disease of the neuromuscular junction, not a demyelinating neuropathy: the transposition is not direct. We cite it because it illustrates a method of monitoring, not because it constitutes evidence in CIDP.

What the data allow you to say

Three findings are firmly established.

First finding: at the doses tested, no worsening has been reported. In the Danish trial, three weekly resistance sessions over twelve weeks in patients with CIDP increased strength, and the authors report no deterioration16. That is an observation, not a predefined safety outcome.

Second finding: no signal of excess risk in neuromuscular diseases in general. The Cochrane review on promoting physical activity concludes that it found no evidence that supervised activity increases the risk of a serious adverse event, while noting that the collection of such events was incomplete in the included studies19. An absence of signal in an incomplete collection is not proof of safety.

Third finding: a signal of harm does exist, but elsewhere. A systematic review of 60 publications on exercise and electrical stimulation in nerve repair concludes that an early active programme broadly favours axonal regeneration, while reporting that harmful effects of exercise or stimulation given too early after nerve repair have been recorded. The authors specify that human research is too limited to conclude22. That body of work is mostly animal and concerns traumatic nerve injury, not autoimmune demyelination.

What the data do not allow you to say

We have to be explicit, because this is the point where it is most tempting to fill the gap with conviction.

• No study conducted in CIDP has compared a high load with a moderate one. There is therefore no threshold beyond which we would know not to go.
• No study has measured overwork weakness as an outcome in CIDP. We know neither its frequency, nor its reversibility, nor its risk factors in this disease.
• The available trial included only treated, stable patients. Nothing is known about exercise during a relapse, or in an untreated patient.

We do not have evidence that loading heavily is safe. We have an absence of any signal of harm at one specific dose, in eighteen treated, stable patients. That is not the same sentence, and the difference is exactly what separates cautious practice from extrapolation.

An approach that takes account of the uncertainty

What follows is not evidence but clinical reasoning applied to a context where evidence is lacking. We flag it as such.

1. Start from the documented dose rather than inventing one. Three sessions a week, segmental work in knee and elbow flexion and extension, a cycle ergometer for the aerobic component: that is the protocol that produced the only published results in this disease16. Departing from it upwards is a personal choice, not an application of the data.
2. Make monitoring measurable instead of leaving it subjective. Weakness that is getting worse shows up on a repeated MRC sum score, not on an end-of-session impression. The useful reference point is the next day and the day after, not the minute that follows.
3. Distinguish fatigue from loss of strength. Fatigue is intrinsic to the disease and affects a majority of patients26. A patient who is tired after a session is not necessarily an overworked patient; a patient who has lost two MRC points on a muscle group is.
4. Do not credit the session with what belongs to the treatment cycle. Deterioration that always occurs at the end of the interval between two infusions is telling you about maintenance, not about training. Record the date of the last infusion at every assessment.
5. Progress slowly and one variable at a time. Increase the load or the volume or the frequency, never all three together, so that any deterioration can be attributed.
6. Set yourself a stopping rule written down in advance. For example: progression suspended and reassessment if the MRC sum score falls by more than two points, or if a benchmark functional capacity deteriorates over two consecutive assessments.

How do you measure, so as to see deterioration before the neurologist does?

This is probably the physiotherapist's most concrete contribution in this disease. It does, however, require using the tools that the CIDP literature itself uses, and knowing which ones have been tested and set aside.

The two measures to keep

The MRC sum score. It is the sum of the manual muscle testing grades over a standardised set of muscle groups. It appears as an outcome in CIDP trials and in the Cochrane review on immunoglobulin12, and it serves as a benchmark of clinical stability in treatment withdrawal protocols15. It requires no equipment.

The I-RODS scale. It is a patient-reported disability questionnaire built on the Rasch model, specific to inflammatory neuropathies. It is used to characterise clinical stability in CIDP cohorts15, and it is among the disability scales for which a recent review considers that the minimal clinically important difference has demonstrated its validity and clinical relevance25.

That same review adds an important and counter-intuitive qualification: the minimal clinically important differences established for strength scores, walking tests, sensory scales and electrophysiological measures lack reliability and direct clinical relevance in relation to the notion of clinically meaningful benefit, whereas those of the INCAT, ONLS and I-RODS disability scales have proved themselves25. In other words: measuring strength remains useful for following a trend, but the scale that tells you whether the patient is better is the disability one.

What has been tested and set aside

Posturography on a force platform looks attractive for objectifying a balance disorder. It has been evaluated in dysimmune neuropathies, in 52 patients with CIDP and 13 patients with neuropathy associated with a monoclonal IgM. Centre-of-pressure path length was 25 % greater in patients complaining of a balance problem (p = 0.03), and reliability between successive measurements was excellent. But correlations with disability and quality of life scales were weak, and changes over time were not consistently associated with changes in the other outcomes. The authors' conclusion is unambiguous: despite excellent reliability, validity and responsiveness are poor, and its use in clinical practice or in trials cannot be recommended24.

That is a lesson useful beyond CIDP: a reliable instrument, one that gives you the same figure twice running, may be measuring nothing that matters.

Grip strength

Grip dynamometry is widely used in the follow-up of CIDP. A 2025 review sets out its value for quantifying weakness, following its course and assessing the response to treatment, while stressing the need to standardise measurement protocols and the value of pairing it with clinical scales23. It is a quick, reproducible and sensitive measure, provided you always use the same dynamometer, the same position and the same number of attempts.

What do published clinical cases teach us?

Three real, published and referenced observations. None involves a rehabilitation programme: that kind of publication does not exist in CIDP, and we will not fabricate a case to fill the gap. What they do illuminate precisely is what the physiotherapist is in a position to see.

How do you apply this in the clinic?

A five-stage approach, from the first assessment to the letter to the neurologist. It owns the distinction between what is grounded in data and what is clinical reasoning on undocumented ground.

Stage 1: establish a numerical baseline

Before anything else, set down the markers that will make it possible to say, in six weeks' time, whether anything has changed. MRC sum score, grip strength by dynamometer, 30-second sit-to-stand, I-RODS, date of the last infusion. Five measures, some twenty minutes, and a dated table. Without that baseline, everything that follows is impression.

Stage 2: set the rhythm to the treatment cycle

Intravenous immunoglobulin acts in courses spaced out over time, with an effect that wanes before the next infusion. A patient assessed three days after their infusion and reassessed the day before the next one will appear to have deteriorated when nothing will have changed underneath. Always assess at the same point in the cycle, and record it. Subcutaneous immunoglobulin, given at home and more frequently, smooths this out.

Stage 3: build the programme on the documented dose

The only protocol to have produced a published result in CIDP rests on three sessions a week, with segmental resistance work for the knee and elbow flexors and extensors, and aerobic reconditioning on a cycle ergometer16. Two additions are matters of clinical reasoning rather than evidence: balance work, consistent with the proprioceptive deficit and documented in non-diabetic neuropathies21, and fitting an ankle-foot orthosis for foot drop, which has never been evaluated in CIDP but whose indication arises on the same grounds as in other neuropathies.

Stage 4: monitor with a rule written in advance

Decide before you start what will trigger a suspension of progression and a reassessment. A simple rule, written in the notes and explained to the patient, beats a judgement made after the event: a fall of more than two points in the MRC sum score, or deterioration of a benchmark functional capacity over two consecutive assessments, or the appearance of one of the red flags listed above.

Stage 5: write to the neurologist with figures

This is the act with the greatest value and the lowest cost. A useful letter does not say "the patient seems less well to me". It says:

That paragraph is what the consultation cannot produce, because it sees the patient only once. It supplies exactly the kind of material whose absence the report on unmet needs in CIDP flags: objective data for deciding whether the treatment is working, whether the schedule needs adjusting, or whether this is a relapse11.

Frequently asked questions