Lupus, TB, and IgG4

Motivation: Recently, I have heard increasing whispers about IgG4 related disease.  In the  list of diseases that are hard to diagnose with widely varying symptoms, long time residents like TB and lupus are being joined by IgG4.  Of course, I have yet to see a single patient with IgG4 related disease personally.  But, when mentioned, I do not know how to fend off the possibility.  So, how do patients with IgG4 typically present?

As way of background, IgG4 related disease has become the topic of much discussion only since 2001, when autoimmune pancreatitis related to IgG4 was first described.  Pathologically, IgG4 related disease is characterized by extensive IgG4+ plasma cell infiltration.

Paper: Zen, Y and Nakanuma, Y. "IgG4-Related Disease: A Cross-Sectional Study of 114 Cases" Am J. Surg. Pathol. (2010); 34: 1812.

Methods: Between 1990 and 2009, pathological samples with diffuse IgG4 plasma cell infiltration were retrieved from case files at Kanazawa University Hospital and affiliated hospitals in Japan.  Clinical features of patients meeting criteria were retrospectively reviewed from medical records.

Results: 
Location: Total of 114 patients with IgG4 related disorder were identified.  Location of disease is as follows:

• 23 patients (20.1%) with head and neck lesions - primarily salivary gland and lacrimal gland
• 16 patients (14.0%) with thoracic lesions found in lung, pleura, and breast
• 27 patients (23.6%) with hepatobiliary lesions in pancreas, bile duct, gallbladder, and liver
• 13 patients (11.4%) with retroperitoneal disease presenting with retroperitoneal fibrosis and aortitis
• 35 patients (30.7%) with diffuse systemic manifestations

Demographics: Median age of presentation is 64 with youngest aged 42.  Male and females were almost evenly distributed for the head and neck group.  On the other hand, >70% of the patients were male with thoracic lesions and >80% were male with hepatobiliary, retroperitoneal, and systemic presentations.

Associated Autoimmune Diseases: In the cohort, 22 patients had history of allergy while only 2 had a clinical history of autoimmune disease - one with sarcoidosis and one with rheumatoid arthritis.  Enlarged lymph nodes were found in 47 patients (41%) by physical and radiologic examination.

Serum Concentration of IgG4: IgG subset information was available in 58 patients.  Serum IgG4 was elevated in 50 of 58 patients (86%).  Serum levels of IgG4 were highest in patients with systemic and head & neck disease.

Lesion Features: Of the biopsied lesions, all except one lesion was macroscopic - could be identified by physical exam or by radiology.  When IgG4 related disease affected a solid organ, the pattern of infiltration was either diffuse or as mass-like lesion.

Malignancy: Of 114 patients, 3 patients in the systemic group presented with malignancy during follow-up: small cell carcinoma (1 year after IgG4 diagnosis), adenocarcinoma (2 years later), and B cell lymphoma (4 years later).

Discussion: What I take away from this cross-sectional description is that this oft invoked but little understood disease is very different from traditional auto-immune diseases.  IgG4 related disease can occur in just about any part of the body, but the likely presentation is a macroscopically visible diffusely infiltrative or mass-like lesion.  Besides testing for IgG subsets, there is little to distinguish it (at least in this survey) at first glance from neoplastic disease.  Importantly, the disease does not appear correlated with other auto-immune phenomenon such as lupus or rheumatoid arthritis or paraneoplastic disease.  I think that in the future, we will learn more about historical features, but for now, IgG4 related disease remains hard to diagnose with protean manifestations - just like lupus and TB.

MRI for C-spine

Motivation: An elderly man trips on ice and falls.  He complains of neck pain in the ED.  He gets a cervical spine CT, which is normal. He still complains of neck pain.  What do you do now?  I think that many doctors would follow different strategies.  Some would take off the cervical collar and provide ibuprofen.  Others would continue imaging with MRI.  What is the right approach?

As way of background, for cervical spine injury, potentially dangerous ligamentous injuries remain on the differential even after a negative CT.  What is less clear is the incidence of these injuries and the clinical significance of these findings.

Paper: Schoenfeld, A.J., Bono, C.M., McGuire, K.J. et. al. "Computed Tomography Alone versus Computed Tomography and Magnetic Resonance Imaging in the Identification of Occult Injuries to the Cervical Spine: A Meta-Analysis" J. Trauma. (2010); 68(1): 109.

Methods: Meta-analysis of prospective and retrospective studies of MRI after negative CT.  True positive MRI was defined as change in clinical management per clinician judgement.

Results:
Studies: Eleven studies were identified evaluating 1,550 blunt trauma patients evaluated initially by CT and then by MRI.  None of the trials were randomized controlled trials.

Abnormalities: Of the entire cohort, 194 MRI abnormalities were detected in 182 patients (12% of cohort) with negative CT scans.  Of these abnormalities, majority (86, 47% of abnormalities) were ligamentous injuries.  Other significant injuries detected were: degenerative changes (47, 24% of abnormalities),  cord contusion (16, 8.2% of changes), fracture (3, 1.5% of changes), and cord contusions (16, 8.2% of changes).

Clinical Outcomes: As a result of MRI, 84 patients (5% of cohort) required prolonged use of collar, and 12 patients (1% of cohort) required cervical stabilization.  In 86 patients (5.5% of patients), MRI findings were judged clinically insignificant.  For MRI, the pooled sensitivity was 100%, and the pooled specificity was 94%.

Discussion: This paper surprised me with the number of clinically significant injuries missed on CT scanning. About 1% of patients required operative intervention after an apparently normal CT scan.  Similarly, about 1% of patients had cervical cord contusions despite a negative CT scan.  What is not apparent from this paper though are the factors predictive of positive MRI findings with a negative CT scan.  In the absence of clear prospective data, I would think that persistent neck pain or any neurological findings calls for an MRI scan despite a negative CT scan.  So, in summary, if an elderly man continues to complain of neck pain even with a negative CT, I would recommend a cervical MRI.

Back Brace

Motivation: I have never worn body armor, but a thoracolumbar brace looks close to it.  I often see the full body brace clipped onto elderly women after a compression fracture of the spine from a fall, and I wonder if it actually does anything besides making for a closeted experience.  Although there are many practices in surgery without evidence, turns out that this idea has actually been tested in a single randomized trial.

Paper: Stadhouder, A., Buskens, E., Vergroeen, DA, et. al. "Nonoperative treatment of thoracic and lumba spine fractures: a prospective randomized study of different treatment options." J Orthop. Trauma (2009); 23: 588.

Methods: Randomized trial carried out in Amsterdam.  Patients younger than 80 with traumatic thoracic or lumbar compresssion fractures ( less than 50% disc height loss or 30% spinal canal stenosis) without neurological impairment were randomized to (1) physical therapy for 6 weeks, (2) removable brace for 6 weeks, or (3) plaster of paris cast for 6 or 12 weeks.  No single primary outcome pre-specified.  Parameters followed included radiographic improvement, pain measure, and disability index meaurd at 6 weeks, 12 weeks, 6 months, and 12 months.

Results:
Cohort: Total of 108 patient were randomized with 54.1% women with mean age of 45.  Baseline measurements such as gender distribution, age, and admission time were similar among groups.  Majority of patients had high-energy trauma.

Tolerability: For compression fractures, physical therapy was better tolerated than plaster of paris but not significantly different than brace therapy.  Brace therapy was also better tolerated than plaster of paris therapy.

Residual Pain: At long term follow-up (1 year later), visual analogue scale residual pain was significantly less for brace compared to plaster of paris therapy for 12 weeks (mean difference 19.0, CI: 1.87-36.2).  There was no difference between physical therapy and brace therapy or between brace therapy and cast for 6 weeks.

Disability Index: At long term follow-up, measure by Owestry Disability Index, brace therapy for 6 weeks was superior to physical therapy (mean difference 14.9, CI: 2.7-27.1) and cast for 12 weeks (mean difference 10.1, CI: 0.25-25.0) or 6 weeks.

Discussion: I think that this paper represents one of the many cases in medicine where an exploratory trial of limited power has been generalized to a wide population.  The patients in the trial were primarily in their 40s presenting after a high-impact accident.  These patients are clearly different than the elderly patients with osteoporotic compression fractures encountered in clinical practice.  Moreover, the trial itself suffers from several notable flaws with perhaps the lack of a pre-specified primary outcome being the most significant one.  Even for the significant outcomes, the confidence intervals are very wide.  While a brace makes pathophyiological sense for treating fractures, the treatment choice for compression fractures remains an area of uncertainty.

CNS Tuberculosis

Motivation: As one of my attendings put it after morning rounds filled with arcane diagnoses, "By the time you end up at MGH, you the definition of atypical."  One of the atypical diagnoses we often consider is central nervous system tuberculosis - entering into the differential for both persistent meningitis and brain masses.  Every time you mention CNS TB, some smirk while others edge towards a face mask.  I am torn between the two largely because I do not know what the typical presentation is or even how to diagnose it prior to a culture that takes weeks to mature.  So, how do you diagnose CNS TB?

As way of background, central nervous system tuberculosis is thought to occur from hematogenous spread of TB from a primary pulmonary focus.  There is TB deposition and formation of subpial and subependymal foci - called Rich foci.  If these foci rupture, TB meningitis results.  Otherwise, growth of these foci results in tuberculomas.

Paper: Systematic review of diagnoses of CNS TB by British Infectious Society. "British Infectious Society guidelines for the diagnosis and treatement of tuberculosis of the central nervous system in adults and children" Journal of Infection (2009); 59: 167-187.

Methods: Systematic clinical evidence review of papers published in Medline and Pubmed between 1966 and 2008 dealing with CNS TB.  The working group consisted of an interdisciplinary group of specialists.

Results:
Basic Clinical and Laboratory Features of CNS TB:
TB Meningitis: The top three clinical features were fever (60-95%), headache (50-80%), and anorexia/weight loss (60-80%).  Top three clinical signs were neck stiffness (40-80%), focal cranial nerve palsy (30-50%), and coma (30-60%).  On CSF examination, majority have clear appearance (80-90%) with 50% having opening pressure > 25 cm H2O.  CSF glucose:blood glucose ratio was < 0.5 in 95% of patients with TB meningitis.  The leucocyte count varies between 5-1000 cells with variable neutrophilic and lymphocytic predominance.  Protein was also variable ranging from 45 to 300 mg/dL.
CNS Tuberculoma: Few large case series but most common presenting feature is seizure with complaints of headache, fever, and weight loss.  CSF pleocytosis of 10-100 cells is present in only 50%.

Microbiology:
TB Meningitis: Sensitivity of AFB smear up to 80% provided that multiple large volume (>6 mL) taps are submitted for analysis.  Culture media usually takes > 2 weeks to be positive and not helpful with decision making.
Tuberculoma: CSF AFB smear usually not positive and require stereotactic biopsy, which is diagnostic in about 94% of patients.

Nucleic Acid Amplification:
TB Meningitis: Currently available nuceic acid amplification tests about 56% sensitive (95% CI 46-66%) and 98% specific (95% CI 97-99%).  After start of anti-TB treatment, PCR test may retain sensitivity even when AFB smear turns negative.  PCR methods of CSF have not been well studies with tuberculomas.

PPD: Rates of positive PPD with CNS TB vary widely with reported ranges around 10-50%.
CSF adenosine deaminase activity: Unclear sensitivity (estimated around 57%) with CNS TB but lack of specificity with false positives from lymphoma, malaria, crytococcoal meningitis, and other etiologies.

Interferon-gamma release assay: Estimated sensitivity of about 50% since CSF lymphocytes may die more rapidly than peripheral lymphocytes.

Discussion: Analysis of the literature supports why TB occupies such a murky place in the differential for brain masses or meningitis.  There are no very sensitive ways to prove TB meningitis.  In some sense, the history of subacute onset of symptoms with a suggestive CSF profile is the best clue for TB meningitis.  The rest of diagnostic tests are not sensitive enough to rule out TB meningitis over clinical suspicion alone.  One point that I was especially surprised to learn is that although TB is acquired via pulmonary source and spread hematogenously, the PPD is positive in only 50%.  Other assays like CSF adenosine deaminase or even interferon-gamma release assay are relatively insensitive.  So, one has to be careful to over-rule clinical suspicion!

Tackling Nerve Pain

Motivation: Pain is the bane of internists - particularly neuropathic pain. The gnawing, burning pain keeps people (patients and internists) up all night debating whether to use IV opioids.  And, gabapentin hardly seems up to the job.  There were several instances last year when even mega-doses of gabapentin failed to lessen the pain.  I often wondered at that point, what next?  To start a new agent, should I stop gabapentin? (And, is this pain real?)

During a clinic this year, one of the attendings pointed out a helpful trial that attempted to deal with failed control of neuropathic pain.  The trial tested combination of tricylcic agent nortriptyline with gabapentin.  As way of background, gabapentin is an alkylated analogue of GABA that is actually thought to work through modulating calcium channel subunits.  Nortriptyline is derived from amitriptyline with multiple effects including blockade of norepinephrine and serotonin re-uptake, blockade of sodium channels, anti-cholinergic actions, and antagonism of NMDA receptors.

Paper: Gilron, I., Bailey, J.M., Tu, D. et. al. "Nortriptyline and gabapentin, alone and in combination for neuropathic pain: a double-blind, randomised controlled crossover trial" Lancet (2009); 374: 1252.

Methods: Double-blind crossover trial in patients with diabetic polyneuropathy and postherpetic neuralgia with persistent neuropathic pain. Randomized trial of three treatments - gabapentin, nortriptyline, and combined gabapentin and nortriptyline.  In crossover design, each of three groups spent 6 weeks in each of the three treatment groups (received placebo pill for second agent when getting just gabapentin or nortriptyline).  After 6 weeks, each of the groups were switched to one of the other three treatment assignements.  During the trial period, drug doses were titrated to maximum tolerated dose.  Primary outcome was mean daily pain (0-10 rating scale).

Results:
Subjects: The trial enrolled total of 56 participants.  Of these patients, 40 had diabetic polyneuropathy and 16 had post-herpetic neuralgia.  All of the patients were white with men comprising 65% of diabetic groups and 56% of post-herpetic neuralgia group.  In the diabetic group, 20% were taking opioids with about 33% having been exposed to prior tricyclic and 20% having been exposed to gabapentin/pregabalin.  Similarly, in the post-herpetic neuralgia group, 25% were taking opioids with about 31% having been exposed to prior tricyclic and 44% having been exposed to gabapentin/pregabalin.

Gabapentin vs. Nortriptyline: As monotherapy, both gabapentin and nortriptyline achieved lower daily pain intensity compared to baseline (5.4 in baseline vs. 3.2 for gabapentin; 5.4 vs. 2.9 for nortriptyline). The pain rating is significantly lower from baseline but is not significantly different between drugs.  Similarly, neither drug was superior to each other in terms of worst pain in 24 hours, least pain in 24 hours, or percentage pain relief on treatment.

Combination Therapy: Combined nortriptyline and gabapentin was significantly superior to gabapentin and nortriptyline monotherapy in terms of daily pain intensity (2.3 in combination vs 3.2 for gabapentin alone, p=0.001; 2.3 vs 2.9 for nortriptyline, p=0.02).  Combination therapy also provided better percentage pain relief on treatment (63.4% vs. 48.1% for gabapentin, p=0.007; vs. 45.7% for nortriptyline, p=0.002).  In secondary analysis, Beck depression score was significantly better for combination therapy compared to nortriptyline (5.4 vs. 6.8 for nortriptyline, p=0.01).  Self-reprorted vitality was also higher with gabapentin and combination therapy compared to nortriptyline alone. Depression scores were not different between combination therapy and gabapentin.

Dosage: As monotherapy, average maximum doses of gabapentin was 2433 mg/day while that of nortriptyline was 61.6 mg/day.  In combination, average gabapentin dose was 2180 mg/day and nortriptyline dose was 50.1 mg/day.

Adverse Effects: No serious side effects were noted.  Dry mouth was more frequent with nortriptyline while inability to concentrate was more frequent with gabapentin.

Discussion: For patients with poorly controlled neuropathic pain, combination therapy with nortriptyline and gabapentin was superior to monotherapy.  While both drugs appeared to be equivalent in terms of controlling pain, I was surprised at the worse depression scores and decreased vitality with nortriptyline monotherapy compared to gabapentin.  Given the alternate use as anti-depressant, I would have expected nortriptyline to be at least as good as gabapentin in terms of mood.  This analysis is, of course, secondary and limited by the small number of subjects (56 total patients), but given the limited data, I would start with gabapentin monotherapy and then add nortriptyline as an adjunctive agent.  Another cautionary note is that the tolerated doses of gabapentin in the trial was more than 2000 mg/day.  With the usual starting dose of 300 mg thrice daily, the drug needs to be significantly uptitrated prior to pronouncing therapeutic failure.

Preserve your wisdom

Motivation: I dislike going to the dentist.  I do not mind the whining drill, but after finishing poking through my teeth, every dentist tells me to take out all my wisdom teeth.  And, I have resisted for more than a decade.  In few other fields of medicine are prophylactic painful procedures so prevalent for a less than life-threatening outcome - a cavity in a wisdom tooth.  To further solidify my argument for my next visit, I wondered about the data backing this recommendation.

As way of background on dentistry, current American guidelines recommend extracting wisdom teeth that fail to erupt completely ("impacted wisdom teeth" in dentistry parlance) preferably during adolescence.  Since no randomized trials exist in adults, reviewed below is a paper which examines the strength of existing evidence.

Paper:  Shallu, B. and Rajesh, S. "Is it Wisdom to Remove a Wisdom Tooth? - Extraction versus nonextraction Management of Impacted Tooth" Ind. J. of Den. Sci. (2010); 2: 4-6

Results:
Incidence of Pathology in Wisdom Teeth: When followed for four years, about 10% of non-erupted wisdom teeth result in adults develop surrounding gingival inflammation or pericoronitis.  There is low incidence (less than 1%) of root resorption in surrounding molar teeth as a result of non-erupted wisdom teeth.  Incidence of inflammation spreading from wisdom teeth to surrounding teeth (ie. causing dental caries) is estimated at 1 to 4.5%.

Early Removal of Wisdom Teeth vs. Later Removal: In post-intervention series, persons aged 35 to 83 years of age were less likely to experience dry socket, secondary infection, and parasthesias compared to those aged 12-24.  The highest risk of complications occurred in people aged 25 to 34 years.

Erupting Wisdom Teeth Causing Crowding of Anterior Teeth: Primarily theoretical prediction without substantial longitudinal clinical data.  (From a Cochrane review), a single trial comparing outcomes at 5 years did not find substantial differences between tooth extraction and observation.

Discussion: I think that I will take the one out of ten chance that keeping wisdom teeth may result in pathology in the next four years.  There are no data to support benefits of early asymptomatic extraction.  In fact, in England, the governmental health organization, NICE, recommended against prophylactic removal of asymptomatic wisdom teeth.  A similar Cochrane review on this topic also found insufficient evidence.  I was personally surprised at the lack of data given the estimated mutli-billion dollar industry among oral surgeons dedicated to wisdom tooth removal.  I think that like other similar problems in medicine such as with gallstones, preventive intervention may not be justified on a population basis. 

Is High Fructose Corn Syrup Unhealthy?

Motivation: Somehow we all know it - high fructose corn syrup makes you gain weight. I cannot remember when I first heard it or when I last read something about it, but this notion is spreading from the health community outwards.  I even had a patient ask me about avoiding foods with high fructose corn syrup.  Not knowing much beyond rumors, I felt that it was time to investigate the merits of different kinds of sugars.  What is the data linking high fructose corn syrup to the epidemic of obesity and diabetes?

As way of background, regular sugar consists of sucrose, which is a disaccharide of fructose and glucose.  High fructose corn syrup (HFCS), in contrast, consists of monosaccharides with varying concentrations of fructose and glucose (from 42% to 90% fructose).

Paper: Forshee, R.A., Storey, M.L., Allison, D.B. et. al. "A Critical Examination of the Evidence Relating High Fructose Corn Syrup and Weight Gain" Crit. Rev. in Food Sci. and Nut. (2007); 47:561-582.

Method: Systematic review conducted by expert panel convened by Center for Food, Nutrition, and Agriculture Policy to examine the evidence linking HFCS and changes in BMI/body weight.

Results:
The strength of evidence from different classes of evidence are summarized below. Given the breadth of evidence, only significant studies are presented:

Ecological Studies: These are large scale studies linking trends between food consumption and health measures.  Between 1909 to 1997, incidence of type 2 diabetes was positively and significantly correlated with corn syrup intake along with intake of fat, total carbohydrate, protein, and total energy.  Using food-consumption data to estimate sweetener consumption from 1962 to 2000, the authors found a 74 kcal/person increase in per capita sweeter availability in this time period.  No causality with disease is implicated in this study.

Cross-Sectional Studies: Four studies were identified examining relationship between soft or sweetened drink consumption (proxy for HFCS) and obesity.  Three out of four studies found that sweetened regular soft drinks were not significantly associated with BMI. One study among European American children showed positive association between overweight and sweetened beverages along with association with consumption of increased total food and beverages.

Longitudinal Studies: Four studies examined growing children and sweetened drink consumption.  Three out of these four studies did not find any association in excess of total calories consumed.  One study showed that over 19 months, BMI increased by 0.24 units for every additional serving of sweetened drink.

Among adults, when 38,480 women health professionals were followed, total caloric sweetener (fructose or glucose) consumption did not correlate to risk of type 2 diabetes.  Another study also examined sweetened beverage consumption among women 24-44 for 15 years and did not find excess weight gain among high consumers of sweetened beverages.  The study did find that women who moved from low consumption to high consumption were at higher risk of weight gain.

Controlled Trial: Only one RCT has been performed to test this question among children 7-11 years of age.  The intervention group received education emphasizing healthy diet and discouraged sweetened and carbonated beverages.  Mean percentage of overweight and obese children decreased in the intervention arm.

Discussion: In summary, the evidence linking high fructose corn syrup and obesity is circumstantial at best.  The overall upward trend in obesity and diabetes can also be explained by increased caloric intake - which has paralleled the weight gain the past few decades.  Of course, the evidence also does not conclusively refute the hypothesis that HFCS is in part responsible for accelerating obesity.  Given the complex interactions between food habits and general health habits, it is hard to see how anything besides a randomized trial can provide more definitive evidence.  For now, though, to avoid obesity and diabetes, limiting caloric intake is likely a better idea than avoiding high fructose corn syrup.

Filtering out Clots

Motivation: Anti-coagulation for acute DVT and PE does not make complete sense to me.  I understand that anti-coagulation stops clot extension, but what about the big DVT stranded in the vein?  Nothing really prevents it from chipping off in the next few days to the pulmonary vasculature.  Seems to me that a temporary IVC filter while beginning anti-coagulation may prevent PE while buying time for the DVT to be naturally absorbed.  Two days ago, while contemplating this apparently novel strategy, I was surprised to see that this idea had not only been thought about but also been tested through a randomized trial.

Paper: Decousus, H. et. al. "A Clinical Trial of Vena Caval Filters in the Prevention of Pulmonary Embolism in Patients with Proximal Deep-Vein Thrombosis." NEJM (1998); 338: 409-416.

Methods: Multicenter, open label randomized trial in France.  Two-by-two factorial design randomizing patients with proximal DVT to permanent IVC filter vs no filter and to low molecular weight heparin (LMWH) vs unfractionated heparin.  Patients on LMWH and UFH were transitioned to coumadin treatment starting on day 4.  Inclusion criteria were adults with proximal DVT with or without concomitant PE.  Important exclusion criteria were contraindication to anti-coagulation and hereditary thrombophilia.  Primary outcome was occurrence of pulmonary embolism within first 12 days after randomization.  All patients were evaluated first by V/Q scan and then by pulmonary angiography if necessary.  Patients were followed for two years.

Results:

Cohort: In total, 400 patients were recruited and randomized to IVC filter (200 to filter, 200 to no filter) and anti-coagulation (195 to LMWH and 205 to UFH).  Mean age was 72 years and 47.5% were male.  49% of patients presented with PE in addition to DVT.  About 35% had prior thromboembolism and 14% had known cancer.  About 11% had surgery in the past 60 days.

Primary Outcome: Within the first 12 days, there were 2 PE in group with filter compared to 9 in the group with no filter (OR: 0.22, CI: 0.05-0.90), p=0.03.  There were 5 deaths in each group.  Within the first 12 days, there were no differences in PE incidence, major bleeding, or death between the LMWH or unfractionated heparin.

2 Year Follow-Up: At two years post-randomization, patients with IVC filter had significantly more DVT compared to no filter (20.8% vs. 11.6%, OR: 1.87, CI: 1.10-1.45), p = 0.02.  There were no differences in symptomatic PE or mortality.  For patients treated initially with UFH or LMWH, there were no differences in outcomes after two years.

Discussion: This trial shows that while IVC filter may prevent PE in the short term, presence of a filter increases risk of DVT in the long term.  The paper also shows that on average, there is no difference between LMWH and UFH for acute treatment of DVT.  The trial still leaves open the possibility though that an IVC filter inserted for the short tem may decrease the incidence of PE.  While there was no difference in mortality between the two groups, a larger cohort may have shown mortality benefit in addition to the short-term decrease in PE.  This question, however, is probably going to be hard to address since this trial is one of the very few randomized trials evaluating IVC filter.  Perhaps, the makes of the IVC filter may be interested in supporting such a trial!

A Plavix a Day Keeps MI Away?

Motivation: Last week, I was trying to coax a patient into taking aspirin every day, and he refused to give way.  He had stomach pain after every dose of aspirin.  To me, it seemed reasonable to refuse aspirin, but then again, I did not want his next stroke on my conscience either.  I wondered if there was a way out or even a better way.  How good is clopidogrel compared to aspirin for prevention of vascular events in high risk populations?

Paper: CAPRIE Steering Committee, "A randomised, blinded trial of clopidogrel versus aspirin in patients at risk of ischaemic events (CAPRIE)" Lancet (1996); 348: 1329-39.

Methods: Randomised blinded trial of 19,185 adults in USA, Canada, and Europe with symptomatic atherosclerotic peripheral arterial disease, recent MI (within 5 weeks), or recent ischemic stroke (within 6 months).  Patients with bleeding disorders were excluded.  Patients were randomized to clopidogrel 75 mg daily or aspirin 325 mg daily.  Primary outcome was ischemic stroke, MI, or vascular death.  Secondary outcomes included all-cause mortality, intracranial hemorrhage, fatal bleeds, and other clusters of outcomes.  Mean duration of follow-up was 1.91 years.

Results: 
Cohort: Among 19,185 patients, 9,586 were randomized to aspirin while 9,599 were randomized to clopidogrel.  Mean age of cohort was 62.5 years, and the cohort was 72% male and 95% white.  Baseline characteristics of the patients in terms of comorbidities were evenly balanced.  Prior to end of study, 21.3% in clopidogrel group and 21.1% in aspirin group discontinued drug.  Most common cause for stopping was adverse drug effect.

Primary Outcome: For the primary outcome of ischemic stroke, MI, or vascular death, the average event rate per patient year was 5.32% in clopidogrel group and 5.83% in aspirin, yielding a relative risk reduction of 8.7% (95% CI 0.3-16.5%, p = 0.043), in favor of clopidogrel.

Secondary Outcome: For the pre-defined secondary outcome measures of vascular death, death from any cause, or cluster of ischemic/hemorrhagic stroke, MI, and death, there was no difference between aspirin and clopidogrel.

Post-Hoc Analysis: In post-hoc subgroup analysis, patients with peripheral arterial disease benefited the most from clopidogrel over aspirin.  The relative risk reduction for clopidogrel over aspirin for primary outcome was 23.8% (CI: 8.9 to 36.2), p = 0.0028.  For patients with MI or stroke, the effect of clopidogrel over aspirin was non-significant.

Adverse Effects: Frequency of rash was higher in clopidogrel over aspirin whereas GI bleed was more frequent in aspirin users.

Discussion: For secondary prevention of vascular events, clopidogrel is at least as good as aspirin.  What was surprising in the post-hoc analysis was the apparent benefit of clopidogrel over aspirin in patients with PAD but not in post-MI and post-stroke patients.  I think that this finding is a note of caution that atherosclerosis in all vascular territories is not equivalent, and benefit of a drug in a type of vascular disease does not necessarily generalize.  The trial was generally well-done, but the 21% rate of drug discontinuation in both groups remains a concern and may have masked beneficial effects in intention-to-treat analysis.  In summary, if a patient balks against aspirin for secondary prevention, daily clopidogrel is an equivalent option and may even be better if patient has PAD.

Keppra for Subdurals

Motivation: As a medicine intern last year, I often scoffed at the algorithmic recommendation by neurosurgery of Keppra for seven days after every small subdural hemorrhage.  Two nights ago, I self-consciously found myself writing the same recommendation while consulting as neurology for a patient after a fall.  What is the data?

Turns out that there is no high quality randomized trial testing anti-epileptics in acute sub-dural hematomas.  The practice is drawn from literature in traumatic brain injury.

Paper: Chang, B.S. and Lowenstein, D.H.  "Practice Parameter: Antiepileptic drug prophylaxis in severe traumatic brain injury." Neurology (2003); 60: 10-16.

Method: A meta-analysis of prospective placebo controlled trials testing anti-epileptic drugs in traumatic brain injury.  Literature search revealed that data existed only for severe TBI - typically with loss of consciousness or amnesia for 12 to 24 hours, depressed skull fracture, or brain contusion on CT scan.  No constraint was placed on the choice of AED.  Authors separated analysis by prevention of early or late seizures.

Results:
Prevention of Early Seizures: Four randomized trials examined early (less than 7 days) prevention of seizure from TBI.  Three studies used phenytoin and one used carbamazepine.   The pooled RR for seizure on AED compared to placebo was 0.37 (CI: 0.18-0.74).  Of the four studies, one did not find any benefit with AED though that study was hindered by low event rate.  There was no significant increase in frequency of adverse effects in group taking AED.

Prevention of Late Seizures: Five high quality trials tested long-term AED for prevention of seizures.  The treatment duration was 6 months to 2 years (most trials were 1-2 years) with median follow-up of 2 years (five trials) with range of 18 months to 2 years.  Trials used phenytoin, carbamazepine, or valproic acid.  The pooled RR for seizure on AED compared to placebo was 1.05 (CI: 0.82-1.35) demonstrating no benefit - none of the individual studies showed benefit.  Some unblinded studies (not included in pooled analysis) showed benefit.  Adverse effects (particularly rash) was significantly higher in the intervention group than placebo group.

Discussion: In severe TBI, anti-epileptic drugs appear beneficial in preventing seizures in the short term (7 days) but ineffective in the long-term.  This landmark meta-analysis, though, has been commonly extended to justify prescribing anti-epileptics for many minor to mild TBI without changes in consciousness or brain contusions.  Although probably beneficial, the practice is still more opinion based rather than evidence based.    Also, the trials used mostly phenytoin as the anti-epileptic of choice while current practice is more heavily biased towards levetiracetam (Keppra).  While I do not plan to rebel against the seven day course of Keppra for subdurals, there is need for data in this common clinical situation.

New Way for Hypertension

Motivation: Last weekend, I was flipping through an old issue of TIME magazine when I noticed an article with large picture of pink kidneys.  Wondering why TIME magazine would be concerned with urine production, I read a little more and learnt about a "revolutionary" new treatment for hypertension discovered in England in 2010.  The article refers to amazing results published in a trial in Lancet in 2010 from renal sympathetic ablation.  Throughout intern year, we had never seriously considered this therapy.  Why? How good is the data?

Paper: Symplicity HTN-2 Investigators. "Renal sympathetic denervation in patients with treatment-resistant hypertension (The Symplicity HTN-2 Trial): a randomised controlled trial" Lancet (2010); 376: 1903-09.

Methods: Multicenter randomized unblinded trial of renal denervation with treatment resistant hypertension, defined as SBP > 160 mmHg (SBP > 150 with DM2) despite compliance on 3 or more anti-hypertensives.  Important exclusion criteria were GFR < 45 and history of MI, unstable angina, or stroke.  Study conducted at Europe, Australia, and New Zealand.  Patients randomized to renal denervation had endovascular catheter directed radioablation through renal artery access.  Primary endpoint was change in average office-based SBP from baseline to 6 months.

Results:
Cohort: In total, 52 patients were randomized to renal denervation while 54 were in the control group.  The two groups were overall similar except for differences in gender (35% female in intervention vs. 50% in control group) and difference in baseline GFR (77 ml/min in intervention vs. 86 in control group). Three patients were lost to follow-up in each group.

Efficacy: In per-protocol analysis, at six-months, renal denervation resulted in mean decrease in BP of 33/11 mmHg compared to no changed in the control group.  The baseline was 178/96 in the intervention group and the same in the control group.  The result is statistically significant, p < 0.0001.  Similar changes were noted in BP measurements at home-based and 24 hour ambulatory measurements.  At least 84% of patients in intervention group had >10 mmHg decrease in SBP at 6 months.

Secondary End Points: There were no differences in serum creatinine, eGFR, and cystatin C between baseline and 6 months in both groups.  There were no differences in composite cardiovascular between the two groups.

Safety: No serious procedure-related or device related complications noted in the six month follow-up.

Discussion: In this preliminary phase trial of renal denervation, the procedure appears remarkably efficacious.  The degree of BP measurement is clearly better than any other current medicinal or lifestyle approach.  Yet, before adopting this intervention, the paper has multiple limitations that need to be clarified.  First, the control and intervention groups were different some important baseline parameters such as gender distribution and GFR.  Second, there were no hard endpoints used in the paper such as MI or long-term cardiovascular outcomes.  It is unclear whether renal denervation has any long-term risks that may outweigh short-term benefits in BP reduction.  Finally, this entire study was funded, designed, and analyzed by Ardian, the maker of the denervation catheter.  The data would gain more integrity if the trial was designed and analyzed independently.  Thus, this therapy is promising but needs a larger follow-up trial before widespread use. 

Color of the Eyes

Motivation: Part of the checklist on the physical exam is looking at patients' eyes and deciding on paleness. I would love to be able to gauge degree of anemia without Q6 blood draws, but are the eyes a good way to decide on degree of anemia?

For background, while looking at the conjunctiva, pallor is determined by comparing the posterior edge of the palpebra (just where it intersects with the conjunctiva) with the anterior edge of the palpebra bordering the eyelashes.  Normally, the anterior edge is redder than the posterior edge.  In anemia, the anterior and posterior edge are the same color.  Pasted below are pictures of a normal and anemia conjunctival exam:

Normal Conjunctiva

Conjunctival Pallor

 Paper: Sheth, T.N., Choudhry, N.K., Bowes, M. et. al. "The Relation of Conjunctival Pallor to the Presence of Anemia." J. Gen. Intern. Med. (1997); 12: 102-106.





















Methods: 302 medical and surgical inpatients at The Toronto Hospital were prospectively assessed for conjunctival pallor (present/borderline/absent) and compared to the patient's hemoglobin.  The observers (medical students and general internists) were initially trained on 25 patients but were blinded to the hemoglobin levels.  While no primary endpoint was stated, the overall goal was the utility of conjunctival pallor in ruling in or ruling out moderate anemia (hemoglogin < 9 g/dL).

Results:
Cohort: Of the 302 patients examined, 171 were male and 131 were female.  55 had hemoglobin less than 9 g/dL while 247 had higher values.

Conjunctival Pallor:  With hemoglobin cutoff of 9 g/dL, the performance of pallor is as below:
- Sensitivity (pallor present):  14.5%
- Sensitivity (pallor present/borderline): 54.5%
- Specificity (pallor absent): 74.4%.


Discussion: In the final analysis, conjunctival pallor is just not a reliable test for anemia even for trained observers.  A definitely positive test only has sensitivity of about 15%.  Even stretching the boundary to positive or borderline positive only yields a sensitivity of about 54%.  So, anemia cannot be ruled out by conjunctival  pallor.  On the other hand, specificity is somewhat higher though 25% of patients with hemoglobin > 9 g/dL remarkably also had conjunctival pallor.  Thus, if pallor is observed, the patient likely has anemia though even that is not a sure thing (in the study, positive likelihood ratio of 4.49)!  I guess CBC are here to stay.

Mood Enhancement

Motivation: Recently, one of my friends in psychiatry commented that the blog almost never touches upon psychiatry.  And, that is unfortunately true despite the amount of psychiatric diseases in the medical wards.  So, today, I wanted to examine the question of what happens when the SSRI just does not work after a good six weeks.  Often, I have seen psychiatrists "augment" the treatment with a variety of anti-psychotics.  Is there data behind the practice, and which anti-psychotic is the best one?

Paper: Nelson, J.C. and Papakostas, G.I.  "Atypical Antipsychotic Augmentation in Major Depressive Disorder: A Meta-Analysis of Placebo-Controlled Randomized Trials" Am. J. Psychiatry (2009) 166: 980-991.

Methods: Meta-analysis of double blind randomized trials comparing addition of atypical anti-psychotic versus placebo in patients with non-responsive major depression.  Response was defined as >50% improvement in baseline scores of a depression rating scale (Montgomery-Asberg Depression Scale or Hamilton Depression Scale).

Results:
Trials: In total, 16 trials were analyzed with trial duration ranging from 4-12 weeks (9 trials lasted 8 weeks, 5 lasted 6 weeks, 1 lasted 12 weeks, and 1 lasted 4 weeks).  The agents tested were olanzapine, risperidone, quetiapine, and aripiprazole.  These anti-psychotics were added to anti-depressants which were typically a SSRI or SNRI.

Olanzapine: Overall, meta-analysis of 586 in treatment arm and 414 in control group.  Odds ratio as listed below (higher OR with favorable treatment):
- Response: 1.39 (95% CI: 1.05-1.84), Remission: 1.83 (1.30-2.56)
- Adverse Effect: 3.85 (2.03-7.29)

Risperidone: Analyzed 211 in treatment arm and 175 in control group.  Odds ratios as below:
- Response: 1.83 (1.18-2.82), Remission: 2.63 (1.51-4.57)
- Adverse Effect: 2.84 (0.91-8.91)

Quetiapine:Analyzed 677 in treatment arm and 352 in control group. Odds ratios as below:
- Response: 1.60 (1.24-2.08), Remission: 1.89 (1.41-2.54)
- Adverse Effect: 5.52 (2.71-11.24)

Aripiprazole: Analyzed 540 in treatment arm and 525 in control group. Odds ratios as below:
- Response: 2.07 (1.58-2.72), Remission: 2.09 (1.55-2.81)
- Adverse Effect: 2.68 (1.23-5.81)

Duration of Trial: No effect seen in response to drug based on duration ranging from 4 to 12 weeks.

Discussion:  This meta-analysis both demonstrates the utility of anti-psychotics in treatment resistant depression and the equivalence for the four examined anti-psychotics (despite aripiprazole being the only FDA approved augmentation anti-psychotic).  Interestingly, addition of anti-psychotics also resulted in increased adverse effects pretty uniformly.  The meta-analysis did not describe the most common adverse effects (such as mild or severe or suicidal).  Another limitation of this analysis is that only the acute phase of treatment (4-12 weeks) was examined.  No data about maintenance phase was available although treatment is often extended to the maintenance phase as well.  In summary, when faced with a depressed patient unresponsive to usual first line SSRI/SNRI agents, addition of an accessible anti-psychotic (whether risperidone or aripiprazole) is reasonable though patients should be monitored closely for severity of adverse effects.

Is this MDS?

Motivation: In the past few months, I have come across multiple patients with chronic unexplained macrocytosis.  I am often tempted to diagnose MDS and request a bone marrow biopsy.  But, is there a better non-invasive scoring system to triage the likelihood of MDS?  Recently, while searching this topic, I came across this paper that estimates a pre-test likelihood of MDS.

Paper: Rauw, J., Wells, R.A., Chesney, A., et. al.  Validation of a scoring system to establish the probability of myelodysplastic syndrome in patients with unexplained cytopenias or macrocytosis.  Leukemia Research (2011) 35: 1335-38.

Methods: Retrospective review of bone marrow biopsies (2006-2008) at an academic hospital in Toronto, Canada.  Inclusion criteria were anemia (hemoglobin <12 g/dL), thrombocytopenia (platelet < 150), leukopenia (WBC < 4), neutropenia, or macrocytosis (MCV > 96).  Patients with known or suspected hemato-lymphoid diseases (such as lymphoma or widespread lymphadenopathy) were excluded.  BM biopsies were classified as MDS, suspected MDS (met most but not all pathological criteria), or not MDS.

Results:
Scoring System: The authors designed a scoring system with four parameters: (1) age >= 65, (2) RDW > 14.5%, (3) MCV > 96, and (4) LDH  > 250.

Bone Marrow Diagnoses: Of the 313 biopsies reviewed, 100 had MDS, 27 had suspected MDS, and 186 did not have MDS.  Among those not having MDS, 34% were normal, 10% had AML, and the rest had other diagnoses.

Sensitivity and Specificity of Confirmed/Suspected MDS Based on Risk Factors:
Risk Factors              Sensitivity              Specificity
>=1 factor                  95%                       18%
>=2 factors                74%                        46%
>=3 factors                39%                        79%
    4 factors                6%                          96%

Discussion: I liked this paper because it provided a risk assessment of MDS based on parameters that are easily and routinely measured.  Having multiple risk factors should lower the threshold for referral to hematology and having a bone marrow biopsy.  As in other retrospective reviews from tertiary medical centers, the numbers in this paper are affected significantly by referral bias.  In the patients with anemia and macrocytosis getting BM biopsy, probably all had B12 and folate checked and had MCV measured when abstaining alcohol.  In the regular inpatient setting, alcoholism as well as B12 deficiency probably accounts for a significant portion of macrocytosis.  For patients without these other more straightforward explanations, stratifying patients for MDS by these risk factors is a quick clinical tool.

G-CSF for Acute Liver Injury

Motivation: When patients do not respond to steroids for acute alcoholic hepatitis, there is little else you can do - besides watching them wither away.  While observing another such tragedy, I wondered if there is anything else to do.  Most patients with acute alcohol use do not qualify for a liver transplant.  I came across this paper published recently evaluating the remarkable potency of G-CSF for acute on chronic liver injury.

For background, the theory is that G-CSF mobilizes bone marrow derived stem cells (CD 34+), and one hopes that with more stem cells floating around, some will engraft in the liver and lead to regeneration.

Paper: Garg, V., Garg, H., Khan, A. et. al. "Granulocyte Colony-Stimulating Factor Mobilizes CD34+ Cells and Improves Survival of Patients with Acute-on-Chronic Liver Failure" Gastroenterology (2012); 142: 505-512.

Methods: Blinded randomized trial in a hospital in New Delhi, India.   Between December 2008 and August 2010, consecutive patients with acute on chronic liver failure (defined as acute rise in bilirubin, INR, ascites, or encephalopathy in patients with known liver disease) were randomized to placebo or 5 ug/kg G-CSF (12 doses over month).  Exclusion criteria included hepatocellular carcinoma, portal vein thrombosis, multi-organ failure, sepsis, or grade 3 or 4 encephalopathy.  Primary end point was survival at day 60.

Results:
Cohort: In total, 47 patients were randomized with 23 patients to G-CSF and 24 patients to placebo.  Majority of patients were male in each group (3 females in each group).  Mean age was 40 in either arm.  Both groups were balanced in terms of MELD score, grade of varices, CTP score, liver enzyme levels, and encephalopathy.  Most patients had acute alcoholic hepatitis.  The placebo group had more HBV reactivation liver injury.

CD 34 Cells: To verify whether G-CSF actually led to stem cell engrafment, authors compared CD34 cells in liver at baseline and at day 30.  Compared to baseline value of 27.5% CD34 cells in sinusoids (based on cell counting in immunohistochemistry), G-CSF treatment resulted in 45% CD34 cells in sinusoids.  No increase was seen in the placebo group (CD34 cells decreased from 30% to 20%).

Survival: The actuarial probability of survival was 66% with G-CSF versus 26% with placebo (p=0.001).  In total, 7 patients died in G-CSF group versus 17 deaths in placebo arm (majority from progressive multi-organ failure, mostly HRS)

MELD Scores: .Median change in MELD score at days 7, 30, and 60 was -7.4%, -18.23%, and -15.34% in G-CSF group versus +3.33%, +6.25%, and +11.76% in placebo group.

Adverse Effects: G-CSF treatment resulted in three adverse events - one had transient rash, one had herpes zoster, and one had high fevers.  Other patients completed therapy without reported adverse effects.

Discussion: This small randomized trial suggests that G-CSF has significant mortality benefit in acute on chronic liver failure!  In the two month trial period, there were 10 fewer deaths in the G-CSF group suggesting an amazingly large benefit.  There are, of course, a few caveats to this trial.  With the number of recruited patients numbering in the 20s in each group, there is no guarantee that the groups were actually well-balanced.  Also, with two month follow-up, it is unclear whether the effects of G-CSF are persistent or transient.  Clearly, this smaller trial calls for a larger, better designed randomized trial.  In the meantime, if someone with acute on chronic liver failure is perishing on standard therapy, I would strongly consider G-CSF therapy.

CSF Cortisol in Meningitis

Motivation: This year, I have seen two patients with community acquired meningitis - both had viral meningitis.   When we stopped antibiotics on both, I had slight trepidation about what if we were wrong.  After all, microbiologic data on CSF can be misleading.  For example, in about 10% of patients, bacterial meningitis can present with lymphocytic predominance.  Low plasma glucose is present in only 50-60% of patients with bacterial meningitis.  Recently, I came across this article evaluating CSF cortisol as a specific marker of acute bacterial meningitis.

Paper:  Holub, M., Beran, O., Dzupova, O., et. al. "Cortisol levels in cerebrospinal fluid correlate with severity and bacterial origin of meningitis." Critical Care (2007) 11:R41

Methods: Study conducted in an academic hospital in Prague.  Inclusion criteria were symptoms of menigitis (fever, headache, meningismus) for less than 72 hours and lumbar puncture performed on admission prior to administration of steroids as part of meningitis treatment.  Bacterial meningitis was diagnosed by positive bacterial CSF culture or detection of bacterial DNA in CSF using PCR.  These patients were compared retrospectively to data from 37 patients with asceptic meningitis as well as data from CSF of 13 control patients who had received LP as part of headache workup.

Results:
Cohort: In total, 47 patients were diagnosed with bacterial meningitis (mean age of 42) with mean APACHE II score of 12.3.  At day 28, there was 15% mortality.  For the asceptic meningitis group, 37 patients were included with mean age of 38 and average APACHE II score of 3.  There were no deaths at 28 days. 

CSF Cortisol: Mean CSF cortisol level was 8.45 ug/dL (interquartile range: 2.14-10.08 ug/dL) in patients with bacterial meningitis compared to mean CSF cortisol level of 0.62 ug/dL (interquartile range: 0.47-1.02 ug/dL), p = 0.001.  Control patients had mean CSF cortisol level of 0.36 ug/dL (interquartile range: 0.29 - 0.44 ug/dL). 

Correlation: CSF cortisol level correlated with APACHE II score - a measure of severity of sickness ( r = 0.763, p < 0.001).  CSF cortisol also correlated with serum cortisol (r = 0.587, p < 0.001). 

Sensitivity/Specificity: After analyzing receptor operating curves, the best sensitivity/specificity for discriminating bacterial and asceptic meningitis are obtained by setting a threshold of 1.67 ug/dL, which resulted in sensitivity of 82% and specificity of 100%.  When comparing bacterial meningitis and control patients, a threshold value of 0.47 ug/dL results in sensitivity and specificity of 100%. 

Discussion: This paper adds cortisol to one of the panel of factors in CSF chemistry that can aid in discriminating bacterial and asceptic meningitis.  While neutrophilia has higher sensitivity than cortisol, the cortisol level is more sensitive than CSF glucose in detecting bacterial meningitis.  Perhaps, more importantly, the very high specificity makes elevated cortisol a very strong indicator of bacterial meningitis.  The etiology of elevated CSF cortisol appears to be directly related to the overall systemic inflammatory insult in bacterial infection (elevated APACHE II score and serum cortisol).  While this study is a good start, some of the weakness of the design itself are the retrospective nature and generally different clinical condition of bacterial and asceptic meningitis patients (vastly different APACHE scores and 28 day mortality).  If patients are very sick with viral meningitis, do they also have non-specific cortisol elevation?  This paper does not address and was not powered to evaluate this comparable subgroup of patients.  Nonetheless, next time, I do a lumbar puncture to evaluate for meningitis, I will add on a CSF cortisol.

Garlic for Hepatopulmonary Syndrome

Motivation: At some unfortunate moments, you see a person struggling to breathe, and there is not much you can do - even with a tank of oxygen.  Hepatopulmonary syndrome (HPS) falls into this category.  Short of a liver transplant, no effective medical therapies exist.  Multiple vasoconstrictive agents such as somatostatin analogues have been tried without success.  After meeting yet another patient with hepatopulmonary syndrome, I was searching for something out of the ordinary when I came across this paper testing the efficacy of garlic (Allium sativum).  Seems like it might work.

Paper: De, B.K., Dutta, D., Pal, S.K. et. al.  "The role of garlic in hepatopulmonary syndrome: A randomized controlled trial."  Can. J. Gastroenterol.  (2010), 24: 183-88.

Methods: Randomized controlled trial in a single center (Calcutta Medical College in Kolkata, India).  Patients were screened for enrollment if they had portal hypertension.  From this cohort, patients were screened for presence of intrapulmonary vascular dilation with elevated A-a gradient.  Exclusion criteria included intrinsic cardiopulmonary disease, massive ascites, sepsis, or other severe comorbid condions.  The selected patients  were randomized to placebo or garlic (at a dose between 1 to 2 g/m2/day based on multiples of 250 mg pills, mean dose 1.55 g/day).  Follow-up was 9-18 months.  Treatment of primary disease was continued during the trial (antivirals for HBV and HCV and abstinence from alcohol).

Results:
Cohort: Overall, 42 patients were randomized, and 21 patients received either garlic or placebo.  In followup, one patient in the placebo group was lost and not analyzed.  The mean age of the cohort was about 40 years of age, and majority had alcoholic liver disease followed by chronic Hepatitis B.  Majority (at least 85% in both groups) had clubbing on physical exam.  Mean PaO2 was 66 in the garlic group and 64 in the placebo group.  Mean MELD score was about 15 in both groups.

Efficacy:  After nine-months of follow-up, the mean arterial oxygen was higher in the garlic group compared to placebo (83.05 versus 68.75 mmHg, p<0.001).  The mean A-a gradient was lower in the garlic group compared to placebo (21.35 vs. 29.11 mmHg, p<0.001).  In the garlic group, there was a 24.7% increase in arterial oxygen levels compared to baseline (83.05 versus 66.62 mmHg, p<0.001).  Interestingly, in the placebo group, there was also a 7.37% increase in the A-a gradient (68.75 vs. 64.05 mmHg, p = 0.02).

Mortality:  In the garlic group, two patients died on follow-up (GI bleed and sepsis).  In the placebo group, there were seven deaths (mostly sepsis followed by GI bleed).  The mortality, while higher in placebo, did not reach significance (p = 0.052).

Adverse Effects: While no data were presented, authors mentioned that no complications occurred except for "occasional bad breath."

Discussion: This small single center randomized trial suggests that garlic may be efficacious in treating hepatopulmonary syndrome.  For a complex mixture like garlic, it is unclear how it acts, but garlic has been speculated to change vascular tone (particularly by NO signaling).  Since the patients continued to be treated for primary disease (such as by abstinence from alcohol or by antivirals for HBV or HCV), garlic may alternatively act by potentiating therapy for the primary conditions rather than by treating HPS.  In fact, even in the placebo group, the mean PaO2 increased over nine month follow-up indicating that patients getting effective treatment for primary liver condition also had mild subsequent improvement.  Given the surprising findings in this small trial (and some smaller previous trials), I was surprised that there has not been a larger trial yet.  Part of the difficulty likely lies in the fact that garlic is natural, cheap, and hence unlikely to generate large revenue for a drug manufacturer.  Still, a larger trial is very much needed since this current trial suffers from being small, single-center, and likely unblinded to the investigators.  For the next patient I meet with HPS, I will think about suggesting garlic!

Serial ANCA for Wegener

Motivation: Any time a  patient with Wegener's vasculitis (granulomatosis with polyangiitis) is admitted, the same question pops up on rounds.  Should an ANCA level be sent?  And, how does disease activity change with ANCA?  This question is a controversial one, and even briefly asking around, strong feelings abound about this topic among rheumatologists.  In 2007, one of the largest studies to examine this question was published.

Paper: Finkielman, J.D., Merkel, P.A., Schroeder, D., et. al. "Antiproteinase 3 Antineutrophil Cytoplasmic Antibodies and Disease Activity in Wegener Granulomatosis." Ann. Intern. Med. (2007), 147: 611-619.

Methods: This paper analyzes data from WGET trial, which was a radomized trial evaluating etanercept for maintenance of remission.  Patients who tested positive for antibodies against myeloperoxidase (p-ANCA) were excluded from further analysis both because of small number of patients and because of difference in disease course/phenotype in patients with p-ANCA Wegener's.  In this trial, quantitative measurement of anti-Proteinase 3 (anti-PR3) titers was used.  Serum samples were collected every 3 months.

Results:
Cohort: From the WGET trial, the paper analysed data from 156 patients with median age of 51 and median follow-up of 34 months.  The majority had history of disease activity in ear, nose, and throat (79%), kidneys (54%), and lungs (59%).  At baseline, 85.8% of the cohort had positive ANCA with antibodies against Proteinase 3.  Among the 156 patients, 76 received etanercept while 80 received placebo.

ANCA and Disease Severity: When evaluated across patients in cross-section, there was no statistically significant association between ANCA level titers (anti-PR3) and disease severity (measured by the Birmingham Vasculitis Activity Score), p = 0.07.  When measured longitudinally within the same patients, changes in ANCA levels explained less than 10% for variability in disease activity.

ANCA and Remission/Relapse:  In patients who were positive for ANCA at baseline, after adjusting for age, sex, disease severity, and treatment group, decrease in ANCA levels did not correlate with shorter time to remission (hazard ratio, 1.6 (0.97-2.49)).  Among patients who experienced relapse after remission (54 patients), increase in ANCA levels did not correlate with risk of relapse, adjusted hazard ratio of 1.0 (CI, 0.5-1.9).

Discussion:  Serial ANCA testing is unlikely to be helpful in predicting disease flare/remission or explaining disease activity.  Although ANCA titers taken longitudinally weakly correlated with disease activity, variance in ANCA levels explained less than 10% of disease activity hardly making it the first test to follow for diagnostic or prognostic purposes.  Also, while following patients, if ANCA turns positive, disease activity does not necessarily follow and may not require pre-emptive treatment.  Interestingly, decrease in ANCA was almost statistically significant in predicting shorter time to remission (CI: 0.97-2.49).  While this study may have been underpowered and may be hinting at a significant finding, the association is not strong enough to be visible even with a cohort above 100 patients making it likely a weak correlation.  In summary, while majority of patients with Wegener's have positive PR3 antibody or c-ANCA (about 85% in this group), serial follow-up of ANCA levels after diagnosis is unlikely to be helpful.

Steroids for Dizziness Too

Motivation:  During my final rotation in the ED, I met surprisingly many patients with sudden onset of severe unrelenting dizziness - chalked up to vestibular neuritis after clean MRI.  Besides sympathetic listening and symptomatic control with meclizine, is there anything else to do?  I had thought not.  I recently found out, however, that steroids can help vestibular neuritis too.  How good is the data?

For background, the cause of vestibular neuritis is not definitely known but based on associational data, appears related to viral infection or reactivation of latent HSV1.  Symptoms of dizziness usually last three to seven days but recovery is often incomplete with many patients having persistent vestibular deficits on exam.

Paper: Strupp, M., Zingler, V.C., Arbusow, V. et. al.  "Methylprednisolone, Valacyclovir, or the Combination for Vestibular Neuritis" N. Engl. J. Med. (2004) 351: 354-61.

Methods: A prospective, randomized, double-blind, 2x2 factorial trial testing methylprednisolone and valacyclovir in patients with acute vestibular neuritis.  Adult patients were recruited through ED in academic hospitals in Munich and Mainz, Germany.  Diagnosis of vestibular neuritis was based on history of acute onset of severe prolonged vertigo with exam showing unidirectional horizontal nystagmus with a rotational component.  Patients were primarily excluded if they had additional cochlear symptoms (tinnitus or hearing loss), other brainstem deficits, central lesions on MRI, or contraindications to steroids or valacyclovir (peptic ulcer disease, osteoporosis, renal failure, liver injury, etc.).  Methylprednisolone was dosed at 100 mg on first three days with a prolonged taper ending at 10 mg on day 22.  Valacyclovir was dosed at 1000 mg three times daily for seven days.  Primary outcome was vestibular paresis at 12 months.

Results:
Cohort: In total, 141 patients were randomized - 38 to placebo, 35 to methylprednisolone, 33 to valacyclovir, and 35 to steroid plus valacyclovir.  Baseline characteristics were balanced, and average age was in late 40s for most groups.  Treatment was started on average 1.6 to 1.8 days after onset of symptoms.  Follow-up data present for 30/38 in placebo, 29/35 in steroids, 27/33 for valacyclovir, and 28/35 for steroid plus valacycolvir group.  Most common reason for withdrawing was unwillingness to follow-up or non-compliance.

Efficacy Measurement: Change in vestibular function over time was measured by assessing nystagmus in response to caloric stimulation (warm water) in the ears.  Intact vestibular response leads to appropriate nystagmus while damaged vestibular system leads to relative paresis.  Analysis was by as treated and not by intention to treat population.

Steroid: At 12 months, patients receiving methylprednisolone had 62.4% improvement in vestibular function compared to 39.6% improvement with placebo (p<0.001).

Valacyclovir: At 12 months, patient receiving valacyclovir had 36% improvement in vestibular function compared to 39.6% improvement with placebo (not significant).  Combination of steroid and valacyclovir resulted in 59.2% improvement in vestibular function, but the improvement was not significantly different from that achieved by steroids alone.

Discussion:  This trial demonstrates that acute treatment of vestibular neuritis with corticosteroids may help prevent long-term vestibular damage.  Interestingly, the authors did not track resolution of vertigo - the primary symptom that brought the patients in.  The authors intentionally did not track vertigo because they assumed that the brain would compensate centrally for the imbalanced input from the two ears after a few days.  Symptom improvement would not correlate with actual recovery of the vestibular system, but it would have been nice to have the data to compare.  The trial also shows that valacyclovir treatment really did not affect the long-term outcome.  The reason could either be that vestibular neuritis is usually not HSV mediated or that the actual reactivation and replication occurred prior to symptoms and initiation of treatment with valacyclovir.  While convincing, this trial has some serious limitations including small number of subjects, no intention to treat analysis, and significant number of patients lost to follow-up in each group (17% dropout rate in steroid group).  This data needs to be backed up with a larger trial.  For now, though, for patients with acute vestibular neuritis with no contraindications to steroids, starting treatment early may likely have long-term beneficial effects.

Renal Artery Stenosis - Ultrasound or MRI?

Motivation: During this year, I have presumed renal artery stenosis in many people, tested some of them, and found none.  Among so many with vascular disease just about everywhere, I have not discovered a single new case of renal artery stenosis.  After the last patient in whom I was very sure of finding renal artery stenosis and found myself fooled once more when reading the radiology report, I wondered whether I am using the right test.  The first-line test used - at least as far as I can tell from my experience - has been renal ultrasound with doppler.  I wondered whether other imaging ways such as MRI might be better.

Paper: Solar, M., Zizka, J., Krajina, A., et. al. "Comparison of Duplex Ultrasonography and Magnetic Resonance Imaging in the Detection of Significant Renal Artery Stenosis." Acta Medica (2011) 54: 9-12.

Methods: In an academic hospital in Prague, 94 patients with high clinical suspicion of renovascular hypertension were examined by duplex ultrasound, MRA with contrast, and angiography.  Angiography was performed last.  The gold standard was angiography.  For ultrasound, significant renal artery stenosis was defined by peak systolic velocity above or equal to 180 cm/sec (usual standard).  For MRA and angiography, a cutoff of 60% stenosis was used.

Results:
Cohort: Of the 94 patients, 41 were women.  Average age was 60 years with patients taking on average 3.8 anti-hypertensives.  The mean systolic pressure was 153 mmHg (standard deviation of 31).  About a third had diabetes.  Based on angiography, renal artery stenosis was found in 61 kidneys among 186 kidneys examined (atherosclerosis in 58 and additional four had fibromuscular dysplasia - unexplained whether one case of fibromuscular dysplasia was overlap with atherosclerosis or did not cause significant stenosis).

Duplex Ultrasonography: Ultrasonography could be performed successfully in 81 patients (86%).  The most common reason for inability was poor imaging quality.  For ultrasound (among imaged kidneys), sensitivity: 85%, specificity: 84%.  Positive predictive value: 70%, negative predictive value: 92%.

MRA: By MRA, adequate images could be obtained in 93 of 94 patients.  Sensitivity: 93%, specificity: 93%.  Positive predictive value: 86%, negative predictive value: 97%.

Discussion:  Based on this direct comparison of ultrasound and MRA, testing for someone with high pre-test probability of renal artery stenosis should probably proceed by MRA given higher sensitivities.  In cases with low to moderate pre-test probability, renal ultrasound may sufficiently decrease the probability.  Neither method is, of course, perfect.  While this study was done in Prague, it is likely that a regular American population might be more obese and have higher BMI - particularly in patients with sustained hypertension.  The adequacy of an ultrasound examination may be further decreased in this group.  A misconception that I had was that if an ultrasound examination could be performed well, it is just as good as MRA.  But, even among kidneys that could be examined well, the sensitivity of ultrasound was lower than MRA.  Some caveats of this study are that this was a single center study and that there is no verification that radiologists examining the MRA were blinded to the results of the ultrasound (although with MRA, more lesions were detected compared to US).  So, from now, if I am really suspicious of renal artery stenosis, I will go ahead with the MRA.  Maybe, I will find someone with renal artery stenosis.

Aspirin After Stenting

Motivation: Reviewing a patient's home medication list yesterday, I found my favorite offender on the list - aspirin at 325 mg dosage.  The man had a drug eluting stent placed two years ago.  I grumbled out loud something about unnecessary risks to patients.  One of the seniors sitting within earshot struck back that in fact every patient is kept on a full dose aspirin plus clopidogrel after drug eluting stent placement.  Turns out that this fact had escaped me nine months into internship.  But, is there data to support the practice?

Paper: Mehta, S.R., Tanguay, J.-F., Eikelbloom, J.W., et. al. "Double-dose versus standard-dose clopidogrel and high-dose versus low-dose aspirin in individuals undergoing percutaneous coronary intervention for acute coronary syndromes (CURRENT-OASIS7): a randomized factorial trial." Lancet (2010) 376: 1233-43.

Methods: Multinational randomized factorial trial.  Inclusion criteria were symptoms of acute coronary syndrome and either EKG evidence or raised biomarkers.  All patients underwent coronary angiography with intent to undergo PCI within 72 hours.  Exclusion criteria was increased risk of bleeding or active bleeding.  All patients initially got full dose aspirin.  From day two, patients were randomized to either high dose (300-325 mg) or low dose (75-100 mg) aspirin.  In the clopidogrel arm, patients either received 600 mg loading dose followed by 150 mg from days 2-7 or received 300 mg loading dosed followed by 75 mg on days 2-7.  With the factorial design, there were four groups receiving varying combinations of clopidogrel and aspirin.  Primary outcome was composite of CV death, MI, or stroke within 30 days.

Results:
When the trial was first released, the clopidogrel trial results attracted much attention and controversy.  Multiple editorials comment on the analysis of the clopidogrel piece of the trial.  I will just describe the aspirin portion of the trial.

Cohort: There were 8624 patients randomized to high dose aspirin and 8639 randomized to low dose aspirin. The two treatment arms were very similar in baseline characteristics.  Average age was about 61 years with 25% female.  About 63% of the patients had NSTEMI/UA while the remaining patients presented with STEMI.  About 37% of patients took regular aspirin, and another 20% took regular clopidogrel prior to randomization.

Angiography: After initial angiography, 8160 of the 8624 (94.6%) in the high dose ASA group received a stent (58% BMS and 42% DES).    In the low dose ASA group, 8163 of the 8639 (94.5%) received a stent (58% BMS and 42% DES).  Patient did not undergo PCI because of (1) no coronary disease - most common, (2) no suitable area for revascularization, or (3) CABG was better alternative.

Follow-up: Primary outcome of cardiovascular death, stroke, or MI did not differ between the high dose and low dose group (HR: 0.98, 95% CI 0.84-1.13, p = 0.76).  Between high and low dose aspirin, there was no difference in definite or probable stent thrombosis (0.9% vs 1.0%, CI: 0.68-1.26).  In secondary analysis, recurrent ischemia was slightly higher in low dose aspirin (0.4%) compared to high dose aspirin (0.2%) though event numbers were small to make meaningful post-hoc secondary comparisons.  In the overall trial analysis (including those who got PCI and those who did not), high dose aspirin resulted in significant increase in minor bleeding (5.0% vs 4.2%, p = 0.04).

Discussion: This trial shows that even in the acute post-PCI setting, use of a lower dose aspirin does not increase likelihood of stent thrombosis.  Some of the strengths of this trial are the randomized trial design with very large number of subjects (over 8,000 in both groups).  While this trial looks only at initial 30 day events, low dose aspirin is expected to be adequate in long-term setting as well given its equivalence to high dose aspirin in the high-risk post-procedure period.  There are no randomized trials in the long-term setting, but there have been retrospective observational studies which show that there is no increased risk of stent thrombosis with low dose aspirin.  On the other hand, there is more definite evidence showing increased risk of bleeding with high dose aspirin compared to low dose aspirin.  Current practice guidelines, however, continue to recommend prolonged administration of high dose aspirin after stenting!

Diuresis on steroids

Motivation: Ever been stuck with diuresis?  Every morning you are drawing even with fluid balance, and metolazone plus bumex is making no progress on the crackly lungs.  Short of sending the patient to the ICU for ionotropes, is there anything else?  Recently one of my fellow interns pointed out that prednisone may actually help with diuresis.  I was at first skeptical.  After all, with the mineralocorticoid activity of prednisone, I would expect prednisone to worsen diuresis.  But, turns out that there is more to steroids.  Prednisone also has renal vasodilatory effects and increases renal plasma flow and GFR.  So, some data?

Paper: Zhang, H., Liu, C., Ji, Z. et. al.  Prednisone Adding to Usual Care Treatment for Refractory Decompensated Congestive Heart Failure.  Int. Heart J. (2008) 49: 587-95

Methods: The authors conducted an observational study in which patients were recruited who had been been hospitalized for at least one week for decompensated heart failure and had failed to respond to IV diuretics.  Exclusion criteria included infection, cardiogenic shock, SBP<80 or >140, HFpEF, or myocarditis.  Patients could receive IV ionotropes so long as they were maintaining blood pressure.  Prednisone (1 mg/kg/day with max dose of 60 mg) was added to therapy for nine days.  Primary end-points were urine volume, patient assessed dyspnea, changes in renal function, and physician assessed clinical status.

Results:
Cohort: The authors recruited 35 patients with median age of 52.3 years and gender distribution of 48.6% male.  All patients were in class IV heart failure with all patients having LVEF less than 30%.  Besides IV diuretics, 74% of patients (26/35) were also receiving IV ionotropes with the majority also getting IV nitroglycerine (91.4%, 32/35) and digoxin (82.9%, 29/35).

Outcomes: After nine days of prednisone therapy, mean urine output increased from 1400 mL/day to 2400 mL/day (statistically significant).  Interestingly, the main jump in urine output occurred at day 3 when urine output jumped to about 2100 ml/day.  Patient assessed dyspnea improved in 80% of patients (p<0.01).  All but one patient were transitioned entirely to PO medications (off IV diuretics, ionotropes, or nitroglycerin) by day nine.  GFR increased from baseline of 63.4 mL/min to 74.1 mL/min (p<0.05).  Average weight loss was 3.17 kg. 

Safety: There were no deaths.  Prednisone therapy increased fasting glucose levels in diabetic patients but did not affect fasting levels in non-diabetic patients.  There were no new infections noted in the cohort.

Discussion:  This study could really benefit from a control group!  The patients improved remarkably with prednisone administration, but one could also argue that if you stick with diuresis long enough, you will eventually make some headway.  The argument made by the authors is that the patients had received conventional therapy for the week prior without improvement in clinical status.  A concurrent control group or even a historical control group in the second week of therapy for CHF would make the case more convincing.  On the flip side, 34 out of 35 patients progressed from decompensated CHF dependant on IV therapy to PO therapy.  The progress was quite dramatic and may suggest benefit from prednisone.  So, how does this change therapy?  I would still not add prednisone to usual therapy for CHF, but if a patient presents with dyspnea along with the usual CHF/COPD combination, I would be less hesitant to add on the steroids.  Also, this topic is a great mini research project for a retrospective outcomes analysis for CHF patients who got steroids (likely for COPD) compared to CHF patients who did not get steroids.

Drinking Stool

Motivation: Pouring mashed up liquid stool down NG tube? Sounds disgusting.  Imagine the smell from the next burp.  I heard about his idea from a friend last night post-dinner, and I almost could not believe that it works.  But, "fecal transplant" or fecal bacteriotherapy exists and, in some niche corners of the country, is considered the salvage therapy of choice for severe C. diff infections.  What about the data?  There have not been RCTs but plenty of convincing case series. Reviewed is one of the larger case series.

Paper: Aas, J., Gessert, C.E., and Bakken, J.S.  Recurrent Clostridium difficile Colitis: Case Series Involving 18 Patients Treated with Donor Stool Administered via a Nasogastric Tube. CID (2003) 36: 580-585.

Methods: Study was conducted in a single centre in Minnesota.  Inclusion criteria for stool receipients were documented C. difficile infections and two or more relapses despite adequate treatment.  Stool donors were usually spouses or other close family members or donors if family members were not present.  Collected stool was screened by culture and O+P screen to rule out pathogenic infection.  Collected stool was mixed with 50-70 mL of normal saline and then homogenized using a household blender.  The blended mixture was filtered through a paper coffee filter.  For four days prior to transfer, receipients were treated with oral vancomycin and then treated with omeprazole on the day of transplantation.  An NG tube was placed with instillation of 25 mL of liquid stool followed by 25 mLof NS through the tube.  Then, the NG tube was removed, and patients were discharged or returned to their wards.

Results:
Cohort: The study recruited 18 participants.  Mean age was 73 years (range, 51-88) with 72% female (13/18 patients).  Five were inpatients while 13 were treated at GI clinic.  Patients in general had received a mean of 3.6 courses of anti-microbial therapy (metronidazole/vancomycin) prior to consideration for stool transplant.

Follow-up Results:  Of the 18 patients treated, 15 experienced complete resolution defined as cessation of diarrhea and negative C. diff stool testing at ninety day follow-up! Two inpatients treated with stool transplant died (one had ESRD undergoing peritoneal dialysis who developed peritonitis and died, other patient had COPD complicated by pneumonia 14 days after stool transplant).  There was one treatment failure in the protocol who developed recurrent C. diff that was successfully treated after one additional course of oral vancomycin.

Discussion: I am amazed that the procedure works.  The cure rate was 83% in patients with refractory C. difficile infection.  I think that this therapy points out the important fact that C. diff flourishes because of the lack of normal bowel flora, and continuous rounds of metronidazole/vancomycin do little to fix that problem.  The authors chose close family members for donors because presumably their bowel flora is similar to the patient's bowel flora.  A cautionary note for the article are the two deaths.  While both patients were sick prior to the transplantation, it is possible that introducing stool to a very sick patient at risk for aspiration could be dangerous (one patient died of PNA after stool transplantation).  This procedure is probably best suited to the outpatient setting.  On the other hand, there have been reports of C. diff induced ICU level colitis being treated with stool transplantation introduced by colonoscopy, which bypasses the risk of aspiration and ileus in sick patients.  While this theory needs confirmation with a RCT, the stool transplant is cheap and probably effective!

PS: Sorry for the far between posts.  I will write more regularly.

PCP and Steroids

Motivation: For prevention of PCP in patients taking steroids, I have heard a variety of rumors on the needs of Pneumocystis pneumonia prophylaxis.  After what dose of steroid and what duration do patients need prophylaxis?  Some doctors prescribe Bactrim for more than 20 mg of prednisone use for more than a month to some GI doctors who do not prescribe prophylaxis even on higher doses of prednisone.  Recently, our medicine team wondered where the data came from.  One of the major studies addressing this question came from Mayo Clinic in 1996.

Paper: Yale, S.H. and Limper, A.H. "Pneumocystis carinii Pneumonia in Patients Without Acquired Immunodeficiency Syndrome: Associated Illnesses and Prior Corticosteroid Therapy." Mayo Clinical Proc.(1996) 71: 5-13.

Methods: Between 1985-1991, data on patients presenting to Mayo Clinic with Pneumocystis pneumonia but without HIV were retrospectively analyzed.  PCP was proven by bronchoalveolar lavage, lung biopsy, or autopsy.  Patients were excluded from analyses if clinical syndrome was suggestive of AIDS.

Results:
Cohort: Between 1985-1991, there were 116 patients with PCP pneumonia without AIDS.  These patients commonly had associated conditions of hematologic malignancy (30.2%), organ transplantation (25%), inflammatory diseases (22.4%), solid tumors (12.9%), and other diseases.

Steroid Use: Of the 116 patients, 105 (90.5%) had used steroid therapy within one month of diagnosis of PCP.  98 (84.5%) were using steroids at time of diagnosis.  Prednisone use dose and duration are depicted as below:

• Median dose: 30 mg of prednisone (25th percentile was 16 mg meaning 25% were using less than 16 mg)
• Median duration: 12 weeks (25th percentile was at 8 weeks)

Concurrent Infections: In 57.8% of patients, additional infectious agents were detected.  No significant differences in incidence of concurrent infections were noted among different underlying illnesses (such as malignancy vs inflammatory diseases).  The most common concurrent infection was CML (35.3%) followed by Candida infection (18.1%).

Outcome: Overall, in-hospital mortality was 34%.  Respiratory failure occurred in 43% of patients and was associated with 66% mortality.  Of note, PCP infection was associated with a 100% mortality in patients with solid malignancies.  Finally, survival to discharge from hospital was linked to lower dose of corticosteroid at the time of diagnosis.

Discussion: Overall, this paper provides a rough guide to the dose of steroid (median 30 mg) and duration (median 12 weeks) associated with PCP.  However, I think that this paper also shows that there is no absolute threshold effect beyond which the risk clearly skyrockets.  The 25th and 75th percentiles were very  wide.  Another interesting and cautionary outcome is that in the immunosuppressed, there is often co-infection of multiple pathogens, and the process of diagnosis should not stop after detecting PCP.  Also, 11 patients got PCP without using steroids.  PCP should remain on the differential for any immunosuppressed individual.

This paper also has some significant limitations.  First, patients who were selected in this analyses had received at least BAL or lung biopsy.  Currently, many patients are diagnosed by stains of sputum and may be less sick than those getting BAL.  Also, the paper does not address properly the more interesting question of what is a safe dose of prednisone to use without prophylaxis.  Perhaps if the authors picked a population with medium to low dose prednisone use, then a "safe" dose could be described.