Article Text

Download PDFPDF

The dorsal root ganglion under attack: the acquired sensory ganglionopathies
  1. Sarah I Sheikh1,
  2. Anthony A Amato2
  1. 1Neuromuscular Fellow, Department of Neurology, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts, USA
  2. 2Chief, Neuromuscular Division and Vice Chair, Department of Neurology, Brigham and Women's Hospital and Harvard Medical School, Boston, Massachusetts, USA
  1. Correspondence to Dr S I Sheikh, Department of Neurology, Brigham and Women's Hospital and Harvard Medical School, 75 Francis St, Boston, MA 02115, USA; sisheikh{at}


Acquired sensory ganglionopathies—or sensory neuronopathies—are a rare type of peripheral neuropathy characterised by damage to the sensory nerve cell bodies in the dorsal root ganglia. Subacute or chronic in onset, sensory ganglionopathies typically present with a non-length dependent pattern of large fibre sensory loss. The causes of this distinct clinical picture include paraneoplastic syndromes, immune mediated diseases, infections, as well as drug, toxin and excess vitamin exposure. Here we discuss the clinical and pathological features of acquired sensory ganglionopathies and focus on a practical approach to their diagnosis and management.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Sensory ganglionopathies or sensory neuronopathies (these two terms are interchangeable but one does have to be careful not to confuse neuronopathy with neuropathy) are a rare but distinct subgroup of the peripheral neuropathies. They are characterised by damage to the sensory nerve cell bodies of dorsal root and trigeminal ganglia, leading to degeneration of their central and peripheral sensory projections.1 2 The diagnosis rests on the clinical and electrodiagnostic picture of peripheral sensory impairment. Once this is established, further testing is required to determine the underlying cause.

The first description of a purely sensory neuronopathy was published in 1948 in two patients with bronchial carcinoma.3 They had presented with severe sensory loss to all modalities, beginning in the periphery of the limbs and advancing to the trunk, in one case even involving the face. At autopsy there was severe loss of nerve cells in the dorsal root ganglia without any obvious pathological changes in the ventral roots.

The dorsal root ganglia are also implicated as a selective target in genetic diseases such as hereditary sensory and autonomic neuropathy type 1 as well as a number of acquired diseases, including sensory ganglionopathies associated with cancer (paraneoplastic sensory ganglionopathies), autoimmune diseases, viral infections, toxins, medications and vitamins. Here we will review the clinical and pathological features of the acquired sensory ganglionopathies and focus on a practical approach to their diagnosis and management.


Dorsal root ganglion cells are T-shaped entities which lie in the dorsal roots in the intraforaminal spaces of the vertebral column. They are supplied by fenestrated capillaries and thus have no tight blood–nerve barrier which is why they are thought to be vulnerable to autoimmune attack and infection.2 They contain two types of cell bodies: those of large fibre myelinated neurons and unmyelinated C fibres which constitute the majority of the dorsal root ganglion cells. Large myelinating afferents for proprioceptive and tactile sensation project through the posterior columns of the spinal cord to the nucleus gracilis and nucleus cuneatus of the brainstem. Unmyelinated primary pain and thermal fibres enter the spinal cord, and ascend one or two levels before synapsing in the substantia gelatinosa. Second order neurons then cross in the spinal cord and ascend as the spinothalamic tracts to synapse with tertiary neurons in the ventro-posterolateral nucleus of the thalamus.

Early in the disease, sensory signs and symptoms are patchy, asymmetrical and involve proximal as well as distal regions. Only later may they develop into a more symmetrical sensory neuropathy. This makes the diagnosis of a chronic sensory ganglionopathy particularly difficult. In contrast, most other peripheral neuropathies are caused by damage to the axon and/or its myelin sheath, giving rise to a very different clinical and electrophysiological entity, which usually manifests with distal signs and symptoms.

Clinical features

Damage to the dorsal root ganglion cells leads to simultaneous degeneration of short (non-length dependent) as well as long (length dependent) axons and it is this feature that is the key to understanding the clinical presentation. Patients show early gait and limb ataxia, likely due to the denervation of proximal muscle spindles and joints. Severe loss of joint position and vibration sense is common. Pseudoathetosis and a severe sensory ataxia are often seen, again reflecting the loss of sensory input. There is also widespread loss of tendon reflexes, arguably the most objective clinical finding. Both positive (tingling, burning, pain) and negative (numbness) sensory phenomena have been described, often in a patchy, asymmetrical distribution. Pain is described as aching, lancinating and burning and can be quite severe. Paraesthesias are less common. Frequently, the arms are more severely affected than the legs. More proximal sites, including the face and tongue, may also be involved. This patchy distribution, caused by the simultaneous damage to short and long length axons, has given rise to the notion of a non-length dependent pattern of denervation. In contrast, the more typical length dependent, dying back axonal neuropathies, present with a distal worse than proximal gradient of sensory loss.

Despite the severe loss of sensory modalities, it is striking to see near normal strength. However, strength is difficult to test when there is sensory deafferentation because there is such poor coordination that patients need to visually focus on the movement they are trying to perform. With visual cues, movements can usually be performed with almost normal power.3 4

Differential diagnosis of ‘ataxic neuropathy’

Many different disease processes can present with ataxia and so it is important to recognise the different types to narrow down the differential diagnosis (table 1). It should be fairly straightforward to distinguish between sensory and cerebellar ataxia on the basis of the history and examination.

Table 1

Differential diagnosis of ‘ataxic neuropathies’

It becomes more difficult to differentiate patients with a sensory ganglionopathy (who may well have absent reflexes) from other chronic neuropathies with ataxic features, particularly some of the demyelinating neuropathies such as CANOMAD (chronic ataxic neuropathy, ophthalmoplegia, IgM paraprotein, cold agglutinin and diasialosyl antibodies) and anti-MAG (myelin associated glycoprotein) neuropathy. Further investigations, in particular demyelinating features on electrodiagnostic testing, should help point in the direction of a demyelinating neuropathy.

In patients presenting with subacute sensory ataxia, the main differential diagnoses are two variants of Guillian–Barré syndrome (GBS): pure sensory GBS and Miller Fisher syndrome. Pure sensory GBS is extremely rare. It is a monophasic disease in which features of a demyelinating neuropathy are seen (often in both sensory and motor nerves) along with the typical GBS CSF findings of a high protein but normal cell count. Miller Fisher syndrome is characterised by the triad of sensory ataxia, external ophthalmoplegia and areflexia. Often the same CSF findings are observed as in GBS in addition to high serum titres of GQ1b antibodies. The electrodiagnostic findings may be normal or just show small sensory nerve action potentials (SNAPs).

Chronic immune sensory polyradiculopathy is another very rare condition with ataxic neuropathic features. It is a restricted from of chronic inflammatory demyelinating polyneuropathy. The diagnostic findings are a raised CSF protein, normal standard motor and sensory nerve conduction studies and electromyography (because the disease process spares the sensory nerve cell bodies) but abnormal somatosensory evoked potentials and H reflexes reflecting slowing of conduction in the proximal segments of the nerves. MRI of the spine may show thickening and enhancement of the posterior nerve roots.

Differential diagnosis of the sensory ganglionopathies

There is a fairly limited differential diagnosis of the sensory ganglionopathies (table 2).

Table 2

Differential diagnosis of the acquired sensory neuronopathies

Sensory ganglionopathies associated with neoplastic conditions

Small cell lung cancer (SCLC) is the tumour which has been associated most often with paraneoplastic sensory ganglionopathies although other malignancies, including breast and ovarian cancer, lymphoma, colon, stomach and neuroendocrine tumours, have also been implicated.5 6 Patients may present with a rapidly progressive, painful, patchy neuropathy with loss of all sensory modalities, often starting in the arms. There can be marked gait ataxia and pseudoathetosis. Areflexia in the face of preserved muscle strength is invariable. The sensory ganglionopathy often precedes cancer symptoms and diagnosis by several months. The median time from onset of sensory ganglionopathy to diagnosis of cancer is 5 months but can be from a few weeks to up to 5 years. It is, however, rare to find a malignancy beyond 3 years.7 This means that— especially early on—there has to be a thorough search for an underlying malignancy, including total body fluorodeoxyglucose–positron emission tomography if required.

Two antibodies have been associated with paraneoplastic sensory ganglionopathies and are most commonly found in SCLC: anti-Hu antibodies, also known as antineuronal nuclear autoantibodies type 1 (ANNA-1), and antibodies to the collapsin response mediator protein-5 (CRMP-5).7 8 With the latter, there appears to be more widespread damage to peripheral nerves in addition to the sensory cell bodies in the dorsal root ganglion.8

Hu belongs to a family of mRNA binding proteins which are expressed during embryogenesis but are also found in the nuclei of neurons, SCLC, neuroblastomas and neuroendocrine cells. Anti-Hu antibodies are closely related to but do not seem to be directly involved in the pathogenesis of sensory ganglionopathies. In the setting of a sensory ganglionopathy, these antibodies have a high specificity (99%) but low sensitivity (82%) and thus the absence of anti-Hu antibodies in a sensory ganglionopathy does not exclude an underlying malignancy. Furthermore, there is no correlation of antibody titre with disease progression or outcome.1 5 7 Hu proteins are expressed widely throughout the nervous system which may explain why there are other paraneoplastic syndromes, such as limbic encephalitis, encephalomyelitis and gastrointestinal dysmotility, which are found in about 20% of patients with a sensory ganglionopathy. Indeed, the first description of sensory ganglionopathies associated with bronchial carcinoma included gastrointestinal dysmotility in one patient.3 Figure 1 shows histological findings in the dorsal root and cervical cord at autopsy in a patient with lung cancer and a paraneoplastic syndrome related to anti-Hu antibodies.

Figure 1

(A) Dorsal root ganglion of the cervical cord, showing marked parenchymal and perivascular inflammation, loss of ganglion cells and fibrosis (haematoxylin and eosin, ×100). (B) Section of cervical spinal cord showing marked pallor of the dorsal columns (arrows) (Luxol Fast Blue–haematoxylin and eosin, ×5). (With permission from Amato AA, Sanelli PC, Anderson MP. A 51-year-old woman with lung cancer and neuropsychiatric abnormalities. Case 38-2001. N Engl J Med 2001;354:1758–65, figs 5 and 7, p1763, 1764.)

Sural nerve biopsies have shown depletion of large myelinated fibres. Chronic lymphocytic cell infiltration has been observed in the dorsal root ganglia along with degeneration of the ascending tracts in the posterior columns at autopsy.7

Very rarely, subacute sensory ganglionopathies have been reported as paraneoplastic syndromes associated with lymphoma, particularly Hodgkin's disease.9

Inflammatory sensory ganglionopathies

The association between sensory ganglionopathy and inflammatory or immune mediated disorders was first described in Sjögren's syndrome but has since been found in other autoimmune diseases, such as rheumatoid arthritis and systemic lupus erythematosus and autoimmune hepatitis.

Sjögren's syndrome is a chronic inflammatory disorder associated with reduced tear production (keratoconjuctivitis sicca) and dry mouth (xerostomia), the so-called sicca syndrome. A patient with sensory ganglionopathy may well not volunteer this history and so it is worthwhile asking whether the patient uses artificial tears (dry eyes) or chews gum (dry mouth). A wide variety of other neurological complications of Sjögren's syndrome have been described as well as sensory ganglionopathy, including peripheral neuropathy, mononeuropathy and multiple mononeuropathies, trigeminal neuropathy, radiculoneuropathy, autonomic neuropathy and cranial neuropathies.10 11

In a study of 92 patients with primary Sjögren's syndrome associated neuropathy, the majority were diagnosed with Sjögren's syndrome only after neuropathic symptoms (likely the most troubling symptoms) appeared. Approximately a third of patients had a sensory ganglionopathy.10 Electrodiagnostic testing of the patients with sensory ganglionopathy showed low amplitude or absent SNAPs and MRI revealed T2 hyperintensities in the dorsal columns in some patients. The pathological underpinnings were dorsal root ganglion cell destruction associated with lymphocytic infiltration (as detected on biopsy, and at autopsy in one study patient). In addition, in those patients who underwent sural nerve biopsies, there was no axonal sprouting even with large axon loss, supporting the view of a sensory ganglionopathy (an intact cell body is required for sprouting to occur).10

A second type of sensory ganglionopathy which was painful and seemed to be small fibre predominant was also seen. The patients tended to have well preserved motor function, relative sparing of large fibre sensory modalities and features of a small fibre neuropathy. MRI of the spine again showed T2 hyperintensities but of a lesser extent than in the large fibre neuronopathy patients. Again, sural nerve biopsies in these patients lacked axonal sprouting, arguing for a more central site of attack, such as the dorsal root ganglion cells.

With time, some of the patients who had a large fibre neuronopathy developed painful dysaesthesias suggestive of small fibre involvement; these two types of neuronopathy may therefore be part of the same spectrum with similar pathological processes.10

Interestingly, in a retrospective case review series of 23 patients with a non-length dependent small fibre neuropathy, the pattern of non-length dependent neuropathic pain with facial and truncal involvement was seen in several different clinical settings, including three patients with Sjögren's syndrome, six with abnormal glucose tolerance, and one patient each with monoclonal gammopathy, sprue and hepatitis C. The other 11 patients were idiopathic.12

Testing for Sjögren's syndrome requires demonstrating decreased tear production with Schirmer's test (a small strip of sterile filter paper placed in the lower eyelid measures the degree of wetting over 5 min—less than 5 mm of wetting is abnormal). Other tests include the Rose Bengal test or the tear break-up time which are mostly administered by ophthalmologists. There is no simple bedside test to quantify xerostomia. Of note, patients with a sensory ganglionopathy may have neither a dry mouth nor dry eyes despite having Sjögren's syndrome. Lack of these clinical findings should not dissuade one from looking for further laboratory evidence to clinch the diagnosis.

Other investigations should include serum antibodies against nuclear antigens SSA/SSB, also known as Ro and La. These antibodies, however, may be negative and the diagnosis of Sjögren's syndrome may only then be made on the basis of a lip biopsy which typically shows inflammatory infiltrates in small salivary glands.

Sensory ganglionopathies due to infections

HIV has been associated with a large number of different neurological presentations in the central and peripheral nervous systems. The neuropathies can manifest in many ways, including distal sensory neuropathy (the most common presentation), polyneuropathy, chronic inflammatory demyelinating polyneuropathy, polyradiculopathy, plexopathy, mononeuropathy and sensory ganglionopathy.13 Other viruses associated with sensory ganglionopathies include Epstein–Barr, varicella zoster, measles and human T cell lymphotropic virus type 1.

Medication related sensory ganglionopathies

Chemotherapeutic agents

In general, drug induced peripheral nerve toxicity depends on the total cumulative dose of drug. The first descriptions of drug related toxicity to sensory nerve cell bodies came from chemotherapeutic agents, mainly platinum based drugs, including cisplatin, carboplatin and oxaliplatin.14 15

Cisplatin is used in the treatment of a number of solid tumours and the limiting factor is its cumulative dose dependent neurotoxicity. Many patients experience tingling in the limbs after cumulative doses of 300 mg/m2 and most adults who receive more than 400–500 mg/m2 experience peripheral neurotoxicity about 3–6 months into treatment. The neuropathy may start during treatment but can also present up to 3–6 weeks after the treatment is completed (the so-called ‘coasting’ effect).15 Large fibre sensory impairment is prominent which may progress to severe sensory ataxia. Lhermitte's phenomenon may also be seen, likely due to irritation of the dorsal columns.

Neurotoxicity is also the most common dose limiting factor for oxaliplatin. The acute neurotoxicity is usually reversible, occurs within 30–60 min of drug administration and involves paraesthesias in the throat, mouth, face and hands. About 30% of patients develop a fixed sensory deficit similar to the one seen with cisplatin.15

Carboplatin alone is less neurotoxic but in combination with paclitaxel can cause a severe sensory ganglionopathy.15

Platinum concentrations in peripheral nerve tissue are similar to those in tumour tissue whereas concentrations in the brain are much lower. This difference may be due to the lack of a tight blood–nerve barrier in peripheral nerves. Platinum compounds are thought to induce aberrant re-entry into the cell cycle and apoptosis in nerve cells, thus leading to drug induced apoptosis of dorsal root ganglion cells. A possible mechanism for delayed neuronal death weeks after the drug has been discontinued—the coasting phenomenon—may be due to binding of platinum to mitochondrial DNA.15

The main differential diagnosis in this setting is between a chemotherapy induced sensory ganglionopathy and a paraneoplastic sensory ganglionopathy; not surprisingly, it can be difficult to tell the two apart. Laboratory testing (including paraneoplastic antibody and autoimmune panels) and CSF are usually normal in chemotherapy induced sensory ganglionopathy whereas in a paraneoplastic ganglionopathy the CSF may show increased protein and cells. Anti-Hu antibodies do not help to distinguish one from the other because although they are found in the paraneoplastic ganglionopathies, they can also occur in cancer patients who have no neurological involvement.

If platinum chemotherapy drug neurotoxicity is suspected, discontinuation of the drug is warranted immediately. Despite the clearance of drug from the circulation in less than a week, the neuropathy may progress for up to several weeks. Once a plateau is reached, there may be some recovery but it is incomplete.14 Unfortunately, clinical trials to date have not identified any prophylactic strategy to prevent sensory ganglionopathies with chemotherapeutic drugs.15


There is just one report of three patients who developed a sensory ganglionopathy within 4–12 days after treatment with antibiotics for a febrile illness. The sensory symptoms remained static over a 5 year period of follow-up without any other disease process appearing in the interim. Given this close temporal relationship, it was postulated that antibiotics may be implicated in the pathogenesis of sensory ganglionopathies. However, this association has not been found subsequently. Large numbers or patients are treated with antibiotics, and if such a cause–effect relationship were really to exist, one would have expected larger numbers of sensory ganglionopathies. Perhaps the underlying infection treated with the antibiotics caused the neuropathy.16


Vitamin associated sensory ganglionopathy has been best described with vitamin B6 intoxication. Vitamin B6 (or pyridoxine) is an essential dietary vitamin which is a coenzyme in many decarboxylation and transamination reactions, and it is also used to treat peripheral neuropathy associated with isoniazid or hydralazine. The daily requirement is approximately 2–4 mg/day. Initially thought to be one of the ‘safest substances known’ and thus used in extremely high doses of up to 3000 mg/day for body building regimens, as well as for the treatment of various conditions ranging from premenstrual syndrome to schizophrenia, vitamin B6 intoxication was first described in 1983 in seven patients using large amounts of up to 6000 mg/day for several months at a time.17 They developed a progressive sensory ataxia with impairment of all sensory modalities, loss of tendon reflexes but no loss in strength. Mega doses of pyridoxine appear to cause a dose dependent loss of dorsal root ganglion cell bodies at doses of 250 mg/day (about 50–100 times the recommended daily dose) or more. This dose dependence has been illustrated in dogs and rats where lower doses cause a reversible sensory axonopathy and higher doses an irreversible sensory ganglionopathy with necrosis of dorsal root ganglion neurons.17 18 The exact mechanism of damage is unknown but interference with neuronal metabolism has been postulated.


Despite extensive investigations, no cause is found in a significant number of patients with sensory ganglionopathy, even nowadays, as was the case in a paper published in the 1980s.19 In this paper, 15 patients with a chronic sensory large fibre neuropathy were followed for an average period of 17 years. Nine had a serum monoclonal or polyclonal gammopathy but no circulating antibodies were found and treatment with immunosuppressants or plasmapheresis was unsuccessful. A toxic aetiology was suspected but could not be proven.19


Investigations can be divided into two parts: those to confirm the presence of a sensory ganglionopathy, discussed in more detail below, and those helpful in determining the underlying cause of the sensory ganglionopathy, which are shown in table 3.

Table 3

Investigations to determine the aetiology of sensory ganglionopathies

Electrodiagnostic testing

  • This is important to confirm the diagnosis.

  • Sensory nerve conduction studies typically demonstrate a non-length dependent sensory neuropathy with diffusely decreased and eventually absent SNAPs. This pattern of non-length dependent axonal degeneration allows one to differentiate ganglionopathy from the more common dying back axonal polyneuropathies in which a distal–proximal gradient is usually seen, such that the sural SNAPs are absent or decreased before any sensory SNAPs in the hands are affected.

  • Motor nerve conduction studies are often normal but sometimes there are subclinical reductions in the compound muscle action potential, particularly in patients with paraneoplastic disease. There are no demyelinating features, such as temporal dispersion, increased distal motor latencies or slowing of conduction velocities across sites usually spared from compression.20 21

  • Needle electromyography is usually normal.

  • Abnormalities of the blink response, which is mediated by afferent trigeminal sensory nerve fibres and efferent facial motor fibres with interposed interneurons in the pons and lateral medulla, have been described in patients with a non-paraneoplastic aetiology. The underlying pathophysiology is unknown and we do not do a blink response in these patients as a routine.22

  • The value of somatosensory evoked potentials in assessing central sensory pathway impairment is hindered by the often low or absent SNAPs (central involvement is most easily seen with cervical spine MRI, as discussed below).20


When the typical electrodiagnostic findings of a non-length dependent sensory neuropathy are seen, cervical spine MRI is useful to demonstrate any central involvement. Typically, it shows T2 hyperintensities in the dorsal columns of the spinal cord which is the neuroradiological correlate of degeneration of large myelinated sensory fibres due to damage to the dorsal root ganglion cells, as has been confirmed in pathological studies. T2 hyperintensities therefore possibly reflect gliosis. These MRI findings take some time to develop and may not be seen in the subacute setting (figure 2).

Figure 2

Sagittal T2 weighted MRI scan of the cervical spine showing a disc osteophyte complex with mild central canal stenosis at the C4–5 level. Associated linear hyperintensity involving the spinal cord is probably related to gliosis (arrows). (With permission from Amato AA, Sanelli PC, Anderson MP. A 51-year-old woman with lung cancer and neuropsychiatric abnormalities. Case 38-2001. N Engl J Med 2001;354:1758–65, fig 4, p1760).

There is, of course, a differential diagnosis of dorsal cord T2 hyperintensity that needs to be considered. Subacute combined degeneration of the cord due to vitamin B12 deficiency is the main differential but the patients tend to present with both clinical and radiological signs of pyramidal involvement. The other main differentials are copper deficiency myelopathy with similar features to B12 deficiency and Friedreich's ataxia, an inherited disease with early onset cerebellar ataxia, scoliosis, cardiomyopathy and diabetes.22,,24

Dorsal root ganglion biopsy

This provides the only definitive proof of pathology in life but is an invasive and traumatic procedure which cannot be recommended for routine investigation. However, microsurgical dorsal root ganglionectomy can be safely performed by experienced surgeons and may yield vital diagnostic information.26

Skin biopsy

Skin biopsies have been obtained to demonstrate a non-length dependant distribution of sensory loss in the diagnosis of sensory neuronopathy and have also been used in the investigation of patients with a purely small fibre sensory ganglionopathy. Although not in routine use, they may add additional information in that they may show lack of axonal sprouting and loss of large with or without small fibre sensory axons.26


Management depends on the underlying cause (table 4).

Table 4

Management of sensory ganglionopathy

Some forms of sensory ganglionopathy are eminently treatable and so it is worth looking carefully for an underlying cause. The earlier this is discovered the higher the chances of some improvement. For example, in vitamin B6 overdose, stopping the offending agent often leads to significant improvement. In platinum induced nerve toxicity, it is absolutely paramount to stop chemotherapy as soon as a neuronopathy is suspected because of the coasting phenomenon, which means that nerves can be damaged long after the drug is discontinued.14 17 In patients with HIV, anti-Hu antibodies or Sjögren's syndrome, treating the underlying condition may stop worsening of the sensory ganglionopathy. In some patients with Sjögren's syndrome, steroids or intravenous immunoglobulins may be helpful but there are no randomised controlled trials to guide treatment.2 5

Practice points

  • Sensory ganglionopathies are rare but should be considered in the setting of a non-length dependent sensory neuropathy.

  • Sensory non-length dependence manifests as proximal and distal sensory symptoms and signs, with decreased or absent reflexes in the face of relatively preserved muscle strength.

  • The causes include paraneoplastic disease, Sjögren's syndrome, infections and various toxins.

  • The diagnosis of a sensory ganglionopathy is established with nerve conduction studies showing low or absent SNAPs with preserved motor responses; cervical cord MRI may be useful.

  • Once the diagnosis has been established, it is essential to exclude treatable causes, such as vitamin B6 toxicity, and search thoroughly for an underlying malignancy, autoimmune condition or infectious cause such as HIV.

  • Treat any underlying cause; trial of high dose steroid and/or intravenous immunoglobulins are warranted, particularly early on in the disease process.


This article was reviewed by Jane Pritchard, London, UK.



  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Other content recommended for you