Improving the CHADS2 Score

Motivation: Recently, I met a man with CHADS2 score of one.  We shook hands, and he had atrial fibrillation.  Slowing his heart down was relatively easy, but what about anti-coagulation, aspirin or warfarin?  A CHADS2 score of one is a grey "intermediate risk" zone without clear indications for using warfarin or aspirin.  About two years ago, a group from England and Netherland proposed a further refinement to the CHADS2 score (called CHA2DS2-VASc) to help decide on anti-coagulation in these intermediate risk cases.

Paper: Lip, G.Y.H., Nieuwlaat, R., Pisters, R., et. al. "Refining Clinical Risk Stratification for Predicting Stroke and Thromboembolism in Atrial Fibrillation Using a Novel Risk Factor-Based Approach: The Euro Heart Survey on Atrial Fibrillation." Chest (2010); 137: 263-272.

Methods: The authors proposed a revised point scoring system with the following components:

Congestive Heart Failure or LV dysfunction: 1
Hypertension: 1
Age 75 or over: 2
Age 65 or over: 1
Diabetes: 1
Stroke, TIA, or other embolic disease: 2
Vascular disease (MI, PAD, or known aortic plaque): 1
Female gender: 1


To validate the predictive value of this scoring system, the authors analyzed data from 1,084 patients without mitral stenosis or heart valve surgery who did not use warfarin.  The authors tracked survival status and risk of thromboembolic disease.  The patients were derived from Euro Heart Survey cohort, which tracked patients among 182 hospitals in 35 countries.

Results:
Cohort Characteristics: The patients in the tracked cohort were on average 66 years of age and 40.8% were women.  The most common risk factor was hypertension (67.3%) followed by CAD (38.4%).  Overall, 34.9% of the cohort had CHADS2 score of one, 20.4% with score of zero, and rest with higher scores.  74% took anti-platelet agents.

Thromboembolic Risk: The annual risk of thromboembolic events by risk factor is shown below (only the low risk rates are shown).  Adjusted rate for aspirin use assumes a 22% risk reduction by aspirin use.

Score: 0 (103 patients, no risk factors) - Event rate: 0, annual rate adjusted for ASA use: 0
Score: 1 (162 patients) - Event rate: 0.6%, annual rate adjusted for ASA use: 0.7%
Score: 2 (184 patients) - Event rate: 1.6%, annual rate adjusted for ASA use: 1.9%
Score: 3 (203 patients) - Event rate: 3.9%, annual rate adjusted for ASA use: 4.7%

One of the novel risk factors that this scoring system adds is female gender.  In univariate analysis, female gender is associated with event rate odds ratio of 2.53 (1.08-5.92).

Discussion: This paper provides help in further distinguishing patients really at low risk not requiring further treatment.  I was pretty impressed with the zero annual event rate in patients with no risk factors.  On the other hand, for patients with score of 2, treatment with warfarin is likely indicated since the risk of symptomatic embolic complication  (1.9%) balances the risk of major bleeding (about the same risk).  Especially in an older population, aspirin carries a major bleeding risk (estimated to be about the same in elderly population) but provides substantially decreased protection against embolic disease.

The study, while helpful, has some limitations.  First, the analysis was done retrospectively in a cohort.  It is unclear why these patients were not anti-coagulated.  It is possible that factors which led to decision not to anti-coagulate also substantially modified the subsequent risk of embolic events (high risk behaviors such as IV drug use).  Nonetheless, I think that this study adds to the identification of the truly low risk group.    

Dizzy is dizzy

Motivation: During patient presentations, saying that a patient is "dizzy" without further qualifiers is sure to trigger further questions.  Did dizziness actually mean vertigo or fainting ("presyncope") or unsteadiness?  This approach to dissecting dizziness into further subdivisions stems from a 1972 paper which divided dizziness into vertigo, presyncope, disequilibrium, and "vague lightheadedness," with the implication that vertigo stems from vestibular causes, presyncope from cardiovascular causes, disequilibrium from neurological causes, and other dizziness from other causes.  This approach, while never rigorously validated, has permeated medicine and has almost become the standard framework for evaluating dizziness.  Recently, this approach has been questioned with growing evidence that the various subtypes of dizziness have many overlapping etiologies.

Results:  Unlike other posts, I will summarize three articles here that show that the approach of subdividing dizziness may not be valid because many dangerous causes of dizziness have variable presentations.

1. Culic, V., Miric, D. and Eterovic, D. "Correlation between symptomatology and site of acute myocardial infarction." Int. J. Cardiol. (2001) 77: 163-8:

In this paper, the authors attempted to correlate sites of myocardial infarction and symptoms of presentation among 1546 patients.  Among the many symptoms, the authors separated feeling of "vertigo" from "faintness."  Here are the findings:

SITE OF INFARCTION                   % WITH VERTIGO             % WITH FAINTING
      Anterior                                               11.1                                   6.2
      Inferior                                                  4.7                                    4.9
      Lateral                                                  8.3                                    4.2

 
2. Newman-Toker, D.E. and Camargo, C.A. "Cardiogenic Vertigo - true vertigo as the presenting manifestation of primary cardiac disease." Nature Clin. Prac. Neurol. (2006) 2: 167-172.

This is a case report of a 90 year old woman who presented to the ED after saying that "Everything's going around in a circle" followed by a brief period of decreased consciousness without prodromal feelings of palpitations or presyncope.  In the ED, presence of vertigo was used to rule out cardiac causes, and patient was admitted to Neurology.  Monitoring there revealed periods of transient asystole (14 second pauses) during which the woman had similar feelings of vertigo.  Placement of a pacemaker resolved these episodes.  This case-report also referred to another study summarized next.

3. Low, PA, Opfer-Gehrking, TL, McPhee, BR, et. al. "Prospective evaluation of clinical characteristics of orthostatic hypotension."  Mayo Clin. Proc. (1995) 70: 617-22.

This article examined 90 patients with documented orthostatic-hypotension undergoing tilt-table testing.  When patients were put in an upright position, 88% complained of lightheadedness while 37% complained of vertigo with some patients complaining of both symptoms.

Discussion: I thnk that these studies (along with many others) make the point that using the type of dizziness to exclude causes of dizziness is dangerous and can lead to mistakes.  Other more validated algorithms have been created (such as using "timing and trigger" by Dr. Newman-Toker), which use the onset of symptoms and factors provoking the symptoms to generate a differential.  Also, people experience dizziness in many ways, and the same cause could trigger many symptoms both in the same person and in different people.  As far I could tell, there have been no large scale studies which have particularly looked at errors generated by relying on dizziness symptoms.  At some level, dizzy is just, well, dizzy.

Spinning the Urine

Motivation: Tonight, at 8 pm, I was trudging up stairs with an empty stomach and one last task to do - looking at urine sediment under the microscope.  But, is urine microscopy a useful heritage of the past?  I was trying to use urine microscopy to distinguish pre-renal injury from acute tubular necrosis, which are often the most debated diagnoses in an inpatient setting.  How good is urine microscopy for such a purpose?  Turns out that these grumbling musings have been studied more conclusively in a prospective study.

Paper: Perazella, M.A., Cocoa, S.G., Kanbay, M. et. al. "Diagnostic Value of Urine Microscopy for Differential Diagnosis of Acute Kidney Injury in Hospitalized Patients." Clin J Am Soc Nephrol (2008) 3: 1615-1619.

Methods: At Yale-New Haven Hospital between April 2006 and May 2007, 267 inpatient nephrology consults were called for acute kidney injury.  Prior to looking at urine sediment, the nephrologists were asked to provide a "pre-test" clinical diagnosis consisting of: (1) ATN, (2) pre-renal azotemia, or (3) other.  Nephrologists could use all available information including trends in serum creatinine.  Subsequently, after patient discharge, renal biopsy, or death, nephrologists were asked to provide a "final" diagnosis.

Results:
ATN vs Pre-renal: Overall in "final" diagnosis, 123 patients were diagnosed with ATN and 108 patients with prerenal AKI. Using the "final" diagnosis after hospital course as gold standard, the performance characteristics of urine microscopy in distinguishing ATN vs pre-renal AKI are as follows:
- sensitivity: 0.76
- specificity: 0.86
- positive likelihood ratio: 5.75

On further subgroup analysis, the authors looked at specific findings in urine microscopy most helpful in distinguishing ATN from pre-renal AKI (expressed in terms of likelihood ratio for ATN):
Granular Casts per High Power Field (40x magnification) +LR for ATN vs Prerenal:
0 ------------ LR: 0.23
1 to 5 ------- LR: 2.97
6 to 10 ------ LR: 9.68
>10 --------- LR: Infinite (No instances in pre-renal cases)
Note that granular casts did not have to be muddy brown!

Renal Tubular Cells per High Power Field +LR for ATN vs Prerenal:
0-------------- LR: 0.72
1 to 5 --------- LR: 1.97
6 to 20 -------- LR: Infinite (No instances in pre-renal cases)
>20 ----------- LR: Infinite (No instances in pre-renal cases)

Concordance: Between "pre-test" and "final" diagnosis, the concordance for diagnosis of pre-renal AKI was 77% while the concordance for ATN was 86%.

Discussion: I found this paper useful in a couple of different ways.  First, I thought that the microscopic equivalent of ATN was muddy brown casts.  Seeing a lot of granular casts or many renal tubular cells predicts ATN just as strongly!  Secondly, urine microscopy remains useful.  Overall initial clinical assessment of pre-renal AKI is wrong about a quarter of the time - an error rate that is pretty high.  Observing many granular casts or renal tubular cells can certainly help reduce this error rate.

This paper, while helpful, has some curious methodological limitations.  The authors had the nephrologists give a "pre-test" diagnosis.  However, no "post-test" diagnosis was assessed after urine microscopy.  That way, the effect of urine microscopy on diagnosis could be better assessed.  Also, the "final" gold standard diagnosis used is an overall clinical assessment and not a renal biopsy, which could significantly change the diagnoses.