The Striking Gender Bias of Hepatic Lesions

For unfathomable reasons, most hepatic lesions have striking gender bias.

Amoebic liver abscess is 9 times more common in men, and almost always present in the right lobe of the liver. It is almost always single. Coexistent amoebic colitis is rare and occurs in only 10% of cases. Serology is an excellent test.

Pyogenic abscesses are usually multiple. No gender bias.

Hepatocellular carcinoma occurs in a shrunken, cirrhotic liver, usually in the right lobe. It is 3 times more common in men, occurs in older subjects (usually >60) and is associated with a rise in serum alpha-fetoprotein in 70% cases.

Metastatic liver cancers are 30-times more common than primary hepatocellular cancer.

Fibrolamellar hepatocellular carcinoma occurs in much younger subjects- mean age 25. It accounts for 10% of primary liver cancers, and is not associated with a rise in alpha-fetoprotein. Distribution is equal between men and women, and between right and left lobes of liver. Prognosis is good, and if resected before metastasising , this cancer is associated with more than 60% 5-year survival. It occurs in a normal appearing liver (not cirrhotic) and has a central scar. Lesions are usually >10 cm, and can present with subcostal pain, intermittent pyrexia and weight loss over years.

Also with a central scar is focal nodular hyperplasia. It is usually solitary, discovered incidentally, is <5 cm in size, much more common in females and occurs in 3rd to 5th decades of life. Hepatic adenomas are almost exclusively seen in young women on oral contraceptives. When they occur in men, a history of anabolic steroid abuse must be sought. Cholangiocarcinomas are only slightly more common in men, but much more so when associated with sclerosing cholangitis, which is 9-times more common in men. Cholangiocarcinomas are often associated with a rise in serum Ca-19-9. However, this antigen is only present in subjects who have the Duffy blood group antigen. Gallbladder cancer is 2-6 times more common in women. Choledochal cysts are 3-4 times more common in women.

Random X Chromosome Inactivation as a Marker of Clonality

A female subject inherits two X-chromosomes, one from each parent. Shortly after conception, each somatic cell randomly inactivates (turns off) one of the X chromosomes. All daughter cells arising subsequently will therefore carry the same pair of inactive and active X chromosome. Since the process is random, the likelihood is that there will be a roughly equal split between inactivation of the paternal or maternal X chromosome.

The process by which this inactivation occurs is by methylation of the genes present on the inactivated X chromosome. Methylation silences the gene, an example of epigenetic modification.

The phenomenon of random X-inactivation can be used to determine the clonality of haematopoetic cells, when these cells are suspected to be neoplastic.

Determining clonality is easy for B cells and T cells. Each B cell experiences an isotype switch from IgM to IgG, IgA, or IgE during its development. B cells then rearrange their immunoglobulin heavy chains and express only the kappa or lambda light chain. Thus a lymphoma arising from a B cell clone will carry a rearranged mu, gamma, alpha or epsilon heavy chain and either kappa or lambda light chain.

T cells don't express cell surface immunoglobulins, but do carry T cell receptors. Just like B cells, they rearrange these receptors upon exposure to their cognate antigen. T cells carry alpha/beta and gamma/delta receptors, although only one of these pairs is expressed. It is easier to test for rearrangement of gamma receptors, because they have less polymorphisms.

However, such unique receptors or cell surface immunoglobulins do not exist for other haematopoetic cells, such as natural killer cells, macrophages or eosinophils. (Yes, they do express generic receptors, but these are not unique for each cell, unlike in T or B cells, which have to recognise specific antigens)

In females, as long there is heterozygosity, inactivation of a gene present on the X chromosome can be used to determine clonality in these haematopoetic cells. Such genes include phosphoglycerate kinase (PGK), hypoxanthine guanine phosphoribosyl transferase and human androgen receptor. Such analysis cannot be done if a both alleles of the gene are the same, i.e. with a state of homozygosity.

PGK has been used, but carries the drawback that there is only 40% heterozygosity in the population for this gene. Most such analyses are now done with the Human Androgen Receptor gene (HUMARA), as the prevalence of heterozygosity is very high- around 90%.

Restriction endonucleases which are methylation sensitive, are used to digest the DNA on each allele. The methylated (inactivated) allele will not be broken down, while the unmethylated (active) allele will be. All daughter cells arising from a single (clonal) cell will carry either the unmethylated (cleaved) or methylated (uncleaved) DNA, but not both. In homozygotes, since the paternally and maternally inherited alleles would be identical, it would not be possible to tell whether the paternal or maternal allele of the gene was present in the cells under consideration.

Males carry only one X-chromosome, hence the phenomenon of random X inactivation cannot be used to determine clonality in men.

Natural Killer Cell Alloreactivity in Haplo-identical Stem Cell Transplantation

Only 30% of subjects needing haematopoetic stem cell transplantation (HSCT) will find HLA identical donors. In the case of related donors, the requirement is to match all 6 alleles at HLA A, B and DRB1, while, for unrelated donors, matching needs to be 8/8, i.e. at HLA A, B, C and DRB1.

An alternative to fully matched donors, is using a haplo-identical donor. This could be a parent, an offspring, a sibling or half-sibling. It is estimated that there is a mean of 2.7 haplo-identical donors available for every subject needing HSCT. These donors are naturally highly motivated, being related donors. Unlike cord blood donations, haplo-identical donors will be available for further infusions of stem cells or lymphocytes at a later date for graft failure or infections respectively if needed.

However, samples of bone marrow or peripheral blood obtained from haplo-identical donors have large numbers of T-cells, giving a much higher risk of GVHD. There is severe bidirectional alloreactivity, also increasing the risk of graft failure. This can be mitigated by depleting the T-lymphocytes in the donor sample with pre-treatment with anti-thymocyte globulin and drugs such as cyclophosphamide or busulfan, and by reducing the risk of GVHD with medthotrexate and cyclosporin or tacrolimus.

Children tend to do better when they receive a haplo-identical graft from the mother rather than the father. Similarly, when receiving the graft from a sibling, they do better if the haplotype mismatch applies to the maternal antigens rather than paternal ones. This is due to the effect of non-inherited maternal antigens (NIMA). It is thought that exposure to maternal antigens in utero, even if those antigens have not been inherited, tolerizes the child's immune system to NIMA. The same does not apply to paternal antigens. Mismatch for paternal antigens carries a worse outcome.

In 2002, Ruggeri and colleagues from Perugia, Italy published a seminal work of research in Science that was to unleash a decade long quest to investigate the role of Natural Killer Cells (NK cells) in transplantation. The role of NK cells, as a part of innate immunity, is to find and destroy cancer cells and infected cells before the adaptive immune system is called into play. They do this by recognising the absence of cognate HLA Class I receptors on the surface of their target cells. If such receptors are detected, the NK cells do not kill them, while if the "self" receptors are absent or not recognised, cytolysis occurs.

To this end, NK cells carry two types of receptors, the NKG2D/CD94 hetero-dimeric receptor which is inhibitory, and a much larger group of receptors called Killer Cell Immunoglobulin-like receptor (KIR), which is comprised of some 15 genes. KIR receptors are broadly of 2 types- A & B. Type A KIR receptors do not have a great deal of antigenic heterogeneity, and are mostly inhibitory. Type B KIR receptors are much more diverse and include both inhibitory and activating receptors. The inhibitory receptors have the suffix L (e.g. KIR2DL. KIR3DL) while activating receptors end with a S (e.g. KIR2DS).

After HSCT, NK cells are usually the first to be repopulated, before T-cells have had time to recover. It is probably stating the obvious to mention that T-cells account for the bulk of graft versus leukaemia (GVL) effect, but increase the risk of GVHD.

Ruggeri and his colleagues found that in T-cell depleted stem cell transplants, when there was mismatch between the recepient's and donor's KIR ligands (the receptors for KIR), GVL effect was better, leukaemia free progression (LFP) was longer and GVHD occurred less often. It is easy to understand why NK cells would attack leukemia cells, but how do you explain a reduced incidence of acute GVHD? It is thought that NK cells do this through allo-reactivity against the recipient's antigen presenting cells (mismatched for KIR ligands).

This effect was not replicated in T-cell replete grafts, presumably because T-cells would increase the risk of GVHD and thus nullify the effect of NK cells.

It is now clear that it is not just the host characteristics that influence graft survival. Unrelated donors who have at least one KIR-B haplotype (KIR B/x, as opposed to KIR A/A), display greater GVL effect, even in T-cell replete grafts. There may therefore be a case or deliberately selecting donors whose NK cells carry KIR-B. KIR-B alleles lying in the centromeric rather than telomeric position appear to provide greater GVL effect. There is also a dose response effect, with greater number of KIR B alleles providing more efficacy.

Further, recipients who are homozygous for HLA-C2 benefit less than those heterozygous or homozygous for HLA-C1. It appears that the activating receptor KIR2DS1, whose cognate receptor is the HLA Class C molecule, will only destroy leukaemia cells that carry HLA-C1, but is tolerized by HLA-C2. Fortunately only 15% of subjects with acute leukaemia are homozygous for HLA-C2.

Finally, mismatch for HLA-Class C antigen between donor and recipient increases the benefits offered by NK cells in terms of GVL effect and LFP through alloreactivity.

It is worth mentioning that these salutary effects of NK cells in haplo-identical and unrelated-donor HSCT are only seen for recipients with AML, but not with ALL.

Hence, if multiple haplo-identical or unrelated, otherwise matched donors are under consideration for a recipient with AML, it is worth selecting the donor with KIR B alleles over a donor who is homozygous for KIR A. There is less agreement over whether to select donors who are deliberately mismatched for KIR ligands to increase NK cell alloreactivity in haploidentical transplants, as some studies have shown a higher risk of GVHD.

References:

1. Ruggeri L, Capanni M, Urbani E, Perruccio K, Shlomchik WD, Tosti A, Posati S, Rogaia D, Frassoni F, Aversa F, Martelli MF, Velardi A: Effectiveness of donor natural killer cell alloreactivity in mismatched hematopoietic transplants. Science 2002; 295: 2097–2100.

2. Cooley S, Weisdorf DJ, Guethlein LA, Klein JP, Wang T, Le CT, Marsh SG, Geraghty D, Spellman S, Haagenson MD, Ladner M,Trachtenberg E, Parham P, Miller JS: Donor selection for natural killer cell receptor genes leads to superior survival after unrelated transplantation for acute myelogenous leukemia. Blood 2010; 116: 2411–2419.

Differentiating Neuropsychiatric Lupus from Anti-Phospholipid syndrome

We've all been there. An elderly woman with obvious cognitive dysfunction, otherwise well and apyrexial, but with high ESR/CRP, is screened for "autoimmunity". The screen comes back as strongly positive. The patient is positive for ANA, dsDNA, C3, C4, lupus anticoagulant and IgG antibodies to beta-2 glycoprotein I. A question is asked. Does this patient have "cerebral lupus?" Do we give steroids, anticoagulation, or both?

Even experts find it difficult to make such determinations. There is too much soft research, nonspecific tests and grey areas passed off as "evidence" in the literature that confuses rather than resolves the issue. However, there are some hard facts that can be relied on. One must take care to ask the right questions.

1. Lupus vasculitis as a cause of neuro-psychiatric lupus (NP lupus) is uncommon~3% of all cases of NP lupus. Most patients with lupus vasculitis present with fever, severe headache and confusion and may progress to coma. A true emergency. The focus is on ruling out infection and TTP, and immunosuppressing ASAP with high dose steroids and cyclophosphamide.

2. Most cases of dementia related to lupus have stable, non-progressive disease and do not need to be immunosuppressed. Aspirin and statins are the order of the day.

3. Two-thirds of neuro-psychatric manifestations in lupus are due to non-lupus causes. Consider infections, medications (e.g. aseptic meningitis caused by NSAIDs and azathioprine, CVA caused by steroids), posterior reversible encephalopathy syndrome in those with very high BP and visual disturbances, TTP, atherosclerosis (risk factors diabetes, BP, hyperlipidaemia and smoking- all treatable causes), obstructive sleep apnoea (causing headache and fatigue).

4. Lumbar puncture is one of the most useful tests in NP lupus. Neutrophilia and high CSF protein indicates vasculitis (if bacterial meningitis can be ruled out). Anti-phospholipid (APL) syndrome is associated with raised protein but no pleocytosis. Presence of raised protein and mild lymphocytic pleocytosis indicates non-vasculitic active NP lupus (raised protein more common than pleocytosis). Hypoglycorrhachia indicates infection, with the exception of transverse myelitis, where CSF sugar is often low.

5. CSF IgG levels, IgG index, and CSF oligoclonal bands are really useful tests. CSF IgG levels are raised in over 2/3 of patients with NP lupus and a CSF IgG level> 6mg/dl is virtually pathognomonic of active NP lupus, although present in only 40% of cases. A raised CSF IgG index and CSF oligoclonal bands are seen in ~80% of cases with active NP lupus, particularly those with diffuse manifestations such as encephalopathy and psychosis. IgG index and oligoclonal bands will be normal/absent in those with focal symptoms such as hemiparesis caused by APL syndrome or chorea. CSF Q-albumin is not an useful test as it'd be raised in any aetiology that causes breakdown of the blood brain barrier such as stroke.

(IgG index is the ratio of two ratios- CSF IgG/CSF albumin divided by serum IgG/serum albumin. Normal IgG index is less than 0.66)

(Q-albumin is simply the ratio between CSF albumin and serum albumin).

6. Demographics are really important. Lupus patients with chorea and transverse myelitis tend to be young women. Always check ANA in a young woman who develops chorea. A positive ANA makes it more likely than not that she has lupus. APL is overrepresented in these two conditions, and in those with seizures.

7. Always check NMO-IgG in a lupus or Sjogren's patient who develops optic neuritis or transverse myelitis. The antibody is present in 75% of subjects with neuromyelitis optica, and carries an adverse prognosis, with multiple recurrences. Anti-Ro antibodies are often associated.

8. Large vessel stroke in SLE is bad news. Can be caused by APL, hypertension or TTP, the typical patient is 35 years old, 86% have active SLE at the time of stroke, and 40% will die in the short term. Thirteen percent will have recurrent stroke.

9. Recurrent TIAs with small infarcts is more characteristic of APL.

10. MRI is overrated. Although it is the preferred imaging procedure in picking up lesions of NP lupus, these typically appear as small white matter lesions on T2 weighted MRI. These are non-specific and can be found in many subjects with lupus who do not have NP manifestations.

11. Studies on anti-neuronal antibodies, antibodies to NMDAR and anti-ribosomal P antibodies are simply too discordant for these tests to be useful, although the latter is reasonably specific for lupus with severe psychosis or depression.

12. The old adage about ESR and CRP in SLE is true. ESR tends to be raised in subjects with lupus, while CRP rises with infection or vasculitis. A subject with NP lupus with raised CRP should be suspected to have infection or vasculitis.

13. NP lupus typically occurs in subjects with active lupus, but less commonly, can be the first presentation of the condition.

14. Scan the heart in all patients with NP lupus. Non infective vegetations (Libman sacks endocarditis) on left sided valves can be a source of emboli for subjects with stroke or multi-infarct phenotypes. APL is overrepresented in such subjects.

15. Don't ignore investigations you'd otherwise perform in non-lupus subjects with stroke, young or otherwise. Thus, carotid dopplers in all patients, and serum homocysteine in young patients should be checked. Young patients with recurrent stroke should be screened for patent foramen ovale through bubble contrast echo.

16. Presence of livedo reticularis (Sneddon's syndrome), thrombocytopenia, adverse obstetric history, or a history of clots makes it more likely that APL is the cause. Subjects with arterial clots related to APL tend to have recurrent arterial clots, while those with venous clots tend to have recurrent VTE, albeit at different sites.

Drug Side Effects can be Predictors of Better Outcomes in Oncology

Drug side effects can limit the usefulness of even the most powerful drugs in medicine. However, there are certain situations in Oncology where the occurrence of certain drug side effects predicts better outcomes for the patient...eventually, if the side effect is recognised and treated.

An example of this can be found in the treatment of metastatic colorectal carcinoma with monoclonal antobodies to EGFR receptors, namely cetuximab and panatumimab. These two agents commonly cause a generalised papulopustular rash resembling acne. However, the rash doesn't occur in all patients, and when it does occur, can be of varying severity. Subjects who develop the rash, or those who have a more severe form of the rash, are likely to have longer progression free survival from cancer.

There is one caveat however. The rash only predicts longer survival in subjects who do not have mutations in the KRAS oncogene, i.e. have wild type KRAS. In subjects with mutated KRAS, the acneiform rash caused by monoclonal antibodies to EGFR does not portend a better outcome. Therefore, in subjects with wild type KRAS, increasing the dose of cetuximab or panitumumab to the point that the rash appears can be a powerful strategy to improve progression free survival.

Another example of this slightly non-intuitive phenomenon of a serious side effect being the harbinger of an eventually better outcome can be observed in patients with malignant melanoma treated with the ipilimumab. Ipilimumab is a monoclonal antibody to CTLA-4 (also called CD152), a receptor found on activated T-cells. CTLA-4 is a "negative" receptor. When stimulated by its ligands, namely B7-1 (also called CD80) or B7-2 (also called CD86) it inhibits the function of the activated T cell. Thus CTLA-4 exists to provide a counterbalance to CD28, the "positive" receptor on T cells, stimulated by the very same ligands that bind to CTLA-4, and thus stop the T cells from running amok.

By inhibiting the inhibitory receptor CTLA-4, ipilimumab therefore stimulates the T cells to seek out and destroy the cancer cells. This strategy works spectacularly well for melanoma, but unfortunately not so well for most other tumours. However, activated T-cells are like a loaded gun. They will attack cancerous cells, but might also harm innocent bystanders and thus cause colitis, dermatitis, hepatitis, neuropathies, or endocrinopathies, such as hypophysitis.

Hypophysitis can be troublesome, particularly by affecting the adrenocortical and thyroid axes. Severe fatigue and headache culminating in life threatening adrenocortical insufficiency may occur in approximately 1.5% of treated patients. However, this autoimmune side effect is now well recognised and can be managed effectively with replacement doses of hydrocortisone (and thyroxine and sex hormones, where applicable, although these other anterior pituitary hormone deficiencies tend to be self limited). Paradoxically, the development of hypophysitis predicts improved survival following ipilimumab therapy. Nowhere is the old adage, "if it doesn't kill you, it makes you stronger" better epitomised than with these agents used by oncologists.

An Approach to Penicillin Allergy

Around 10% of subjects report that they are allergic to penicillin. More than 90% of them are found not to be allergic to penicillin on skin testing. It is estimated that only 0.02%-0.04% of the population has severe (anaphylactic) penicillin allergy.

Strictly speaking, the term "allergy" connotes an IgE mediated process, but in practice, intolerance, idiosyncracy and pseudo-allergy are all lumped under the sobriquet of "allergy." This is unhelpful for a lifesaving medication such as penicillin.

It would be useful to recall the Gel-Coombs classification of hypersensitivity reactions and recollect specific examples.

Gel-Coombs Classification

Type I- IgE mediated, e.g. urticaria, angioedema, bronchoconstriction, hypotension, syncope. i.e. anaphylaxis. Occurs within minutes to an hour (sometimes up to 2 hours, if only oral exposure)
Type II- Antibody mediated cytotoxicity. e.g. Haemolytic anaemia. Occurs within hours to days.
Type III- Immune Complex reaction, e.g. serum sickness- maculopapular rash, fever, joint pains, low C3, C4, low CH50. Occurs from 7-21 days after exposure
Type IV- Delayed hypersensitivity. e.g. maculopapular rash without pruritus, SJS, TEN, DRESS, occurs after days to weeks

It is important to differentiate between these various processes as it is only the Type I (IgE mediated) reaction that lends itself to skin testing, graded dose challenge and desensitisation. These procedures should not be attempted in subjects with non-IgE medited reactions, that is Types II, III or IV hypersensitivity.

Further, it is important to remember that hypersensitivity to skin testing declines with time @ 10% per year. Therefore, 10 years after an index episode of penicillin allergy, 80-100% of subjects would have lost IgE mediated hypersensitivity to penicillin.

The likelihood of developing penicillin allergy is higher after intravenous dosing, frequent dosing, as opposed to oral or infrequent dosing. Atopy itself does not increase risk of penicillin allergy, but atopic subjects are likely to have more severe anaphylactic reactions if they do develop it.

Subjects with EBV developing a maculopapular rash with an aminopenicillin are not penicillin allergic. They can receive penicillin and beta lactams if indicated.

The cross reaction between penicillin and other beta lactam antibiotics is low. There is estimated to be <2% cross-reaction with amino-peicillins, 3% cross-reaction with cephalosporins (less with 3rd and 4th generation than with 1st and 2nd), 0% with monobactams such as aztreonam and <1% with carbapenems. Aztreonam and ceftazidime cross react with each other. There is 10-40% cross reaction between aminopenicillins such as amoxycillin or ampicillin and cephalosporins. Cross reaction occurs due to shared side chains and is non-IgE in character. It is best not to use one if there is a history of insensitivity to the other. It is important to quiz the patient regarding the type of reaction they had with penicillin. The most important arbiter is the speed with which the adverse reaction occurred. If the hypersensitivity was within minutes to an hour, it was likely to have been IgE mediated. Reactions occurring after days are non-IgE medated. For severe delayed reactions such as SJS, TEN, AIN, DRESS, haemolytic anaemia or serum sickness, repeat use of penicillin and other beta-lactams is best avoided. Penicillin is the only antibiotic for which skin prick testing is reliable. There are two antigens that are used to test for penicillin allergy- a major determinant- penicilloyl polylysine, also called Prepen, and a minor determinant- penicillin G itself. Both must be injected. Intradermal administration is necessary only if there is no reaction to skin prick testing. A negative control (normal saline) and positive control (histamine) is used simultaneously. Results are read at 15 minutes and a wheal 3mm or bigger compared with the negative control is considered to be evidence of Type I penicillin hypersensitivity. Other beta lactams are not used for skin testing. The results are read at 15 minutes. The PPV of skin prick test for penicillin allergy is 50% while the NPV is 97%. Thus, a subject with negative skin prick test is very unlikely to have penicillin allergy, but half of those who test positive may not have allergy. Skin prick testing is best delayed for 4-6 weeks after a severe allergic (anaphylactic) reaction.

If skin test is positive, it is best to use an alternative antibiotic. If penicillin or a related beta lactam must absolutely be used as there are no effective alternative, then densitisation may be attempted. This is attempted by giving progressively increasing doses of penicillin every 15-20 minutes S/C or IV or every 30 minutes, orally. It is customary to begin with 1/10,000 to 1/1000 of the therapeutic dose. Success rates are high. For example, between 75% to 100% of subjects with cystic fibrosis, who generally have high rates of drug intolerance, are rendered penicillin tolerant.

If the skin prick test is negative in a subject with a history of an adverse reaction to penicillin, a graded dose challenge is used as a precursor to giving the therapeutic dose. This is done using 1/100 to 1/10 of the therapeutic dose to start with, increasing the dose every 30 minutes.

It is important to mention again, that neither skin prick testing, nor graded dose challenge or desensitisation should be used for non-IgE mediated hypersensitivity reactions to penicillin. In such hypersensitivity reactions,if severe, penicillin is best avoided altogether.


Reference:

Gozalez-Estrada A, Radojicic C. Penicillin allergy: A practical guide for clinicians. CCJM 2015;82:295-300.

Fractional Excretion of Nitric Oxide in Assessment of Inflammatory Asthma

The National Institute of Clinical Excellence (NICE) in the United Kingdom recently recommended measurement of fractional excretion of Nitric Oxide (FeNO) in exhaled air as a surrogate for inflammatory, i.e. eosinophil driven asthma, which, by definition, would be corticosteroid responsive.(1)

Nitric Oxide is produced in response to Th2 lymphocyte driven inflammation. While a Cochrane review in 2009 could not find sufficient evidence to recommend this approach, subsequent studies have shown that the fraction of NO in exhaled air predicts response to anti-inflammatory therapy such as inhaled corticosteroids even in subjects with FEV1 >80% of predicted. Conversely, low FeNO levels indicate well controlled asthma, allowing reduction or discontinuation of oral/inhaled steroids.

Levels of NO in exhaled air of less than 25 parts per billion is characterised as "low", levels of 25-50 ppb as "intermediate" and >50 ppb as "high".

FeNo should be used in addition to standard PFTs such as FEV1, FEV1/FVC ratio and PEF. However, experience shows that these tests are imperfect predictors of response to inhaled steroid treatment. Observational studies show that 60% of subjects with putative asthma referred to secondary care and 30% of subjects diagnosed with asthma in primary care, have no evidence of airway dysfunction, and that in a majority of such subjects, inhaled steroid therapy can be stepped down without adverse consequences. (2)

Measurement of FeNO is therefore a welcome addition to the current diagnostic/monitoring modalities for inflammatory asthma. Following widespread adoption in the UK, I expect that this will now find acceptance elsewhere.

References:

1. Povard ID, Bush A, Holgate S. Asthma diagnosis: addressing the challenges. Lancet Respir Med 2015; 3: 184

2. Hawkins G, McMahon AD, Twaddle S, Wood SF, Ford I, Thomson NC.
Stepping down inhaled corticosteroids in asthma: randomised controlled
trial. BMJ 2003; 326: 1115.