Exercises for Chapter 7: Answers

Evidence-based Diagnosis:

A Workshop on Evaluating and Using Medical Tests

Small Group 3: Friday, June 10, 11:00 – 12:30

Problems with Answers

1. Oseltamivir for Post Exposure Prophylaxis

Case: The patient is a 6-year-old girl with fever, myalgias, cough and sore throat for 1 day. A rapid antigen test for influenza is positive. (See SG-1, Problem 1.) Her mother is 12 weeks pregnant and does not have any flu symptoms at this time.

Setting aside the question of whether to treat the patient herself with oseltamivir, we will focus on the question of whether to treat her mother with prophylactic oseltamivir.

Welliver et al(1) did a randomized trial of prophylactic oseltamivir to prevent influenza in household contacts of patients with clinical influenza. The table below summarized their results in the subgroup where the index case was (like the 6-year-old in this question) positive for influenza.

a) Calculate the NNT explain how to interpret it.

Absolute Risk Reduction = 12.6% - 1.4% = 11.2%

NNT = 1/11.2% = 9

You need to treat 9 household contacts of patients who are test-positive for influenza to prevent one from contracting influenza.

Prophylactic oseltamivir dosing is 75 mg daily for 5 days. This drug cost is about $50.

b) In household contacts such as the mother in this problem, how much to you have to spend in oseltamivir to prevent one case of the flu.

NNT × Treatment Cost = 9 × $50 = $450

Now consider the question of whether it was worth doing a rapid antigen test on the 6-year-old girl for flu purely to guide prophylaxis in the mother. Things get complicated here, because Welliver et al used the PCR (gold standard) to determine whether the index case had the flu. The rapid flu test can be assumed to be 100% specific relative to the PCR but only 50% sensitive.

Assumptions:

• Prophylactic oseltamivir is of no benefit if the index case is gold-standard (PCR) negative for the flu.
• It is worth $250 in oseltamivir costs to reduce the mother’s risk of flu by the amount shown in the table above.
• Since the actual treatment cost is only $50, net gain of prophylactic treatment is $250 - $50 = $200.
• You could also look at this as the cost B of failing to treat the mother if the index patient is gold-standard (PCR) positive.
• The rapid influenza test costs $10.
• Ignore the side effects of oseltamivir (which are minimal) and the issue of increasing oseltamivir resistance.

c) At what range of flu probabilities does it make sense to test the index case in order to guide prophylactic treatment of the household contact? (You may use this link.)

C = $50

B = $200

T = $10

Sensitivity = 50%

Specificity = 100%

No Treat-Test Threshold = 0.10

Test-Treat Threshold = 0.267

If the probability of influenza in the index case is less than 0.10, you shouldn’t test the index case or treat the mother prophylactically. (Remember, the issue here is purely whether to treat the mother prophylactically.)

If the probability of influenza is the index case is greater than 0.267, you should just treat the mother prophylactically without testing. The reason for this low threshold for treating empirically is the poor sensitivity of the rapid test. You can only rely on a negative result if the patient has a low pre-test probability of disease.

d) This is a good opportunity to play with the spreadsheet. What happens if the rapid test is free?

Since the test is perfectly specific (no false positives), the only reason not to do it on a low probability index case is the cost T. Setting T= 0, makes the no treat-test threshold 0.0. The test-treat threshold increases a little from 0.267 to 0.333 because what really drives that is the problem of false negatives.

2. Marcolide Antibiotics for Pertussis

3. Warfarin for VTE

4. Early Goal-Directed Therapy for Sepsis

Rivers et al (2) performed a noteworthy and somewhat controversial clinical trial of Early Goal Directed Therapy (EGDT) to reduce 28-day mortality in patients with severe sepsis and septic shock. The main results are shown in the following table.

a) How many patients with severe sepsis do you have to treat with EGDT instead of standard care to save one from dying within 28 days?

ARR = 0.46 – 0.31 = 0.15 NNT = 1/0.15 = 6.7 We think it is okay to report a factional NNT, but if you want to round up to7 , that’s okay too.

“Early Goal Directed Therapy” was a bundle of interventions performed according to a protocol in a special ED unit by a trained team consisting of an attending emergency physician, 2 emergency medicine residents, and 3 nurses.

Here is the paper’s description of the protocol:

The protocol was as follows. A 500-ml bolus of crystalloid was given every 30 minutes to achieve a central venous pressure of 8 to 12 mm Hg. If the mean arterial pressure was less than 65 mm Hg, vasopressors were given to maintain a mean arterial pressure of at least 65 mm Hg. If the mean arterial pressure was greater than 90 mm Hg, vasodilators were given until it was 90 mm Hg or below. If the central venous oxygen saturation was less than 70 percent, red cells were transfused to achieve a hematocrit of at least 30 percent. After the central venous pressure, mean arterial pressure, and hematocrit were thus optimized, if the central venous oxygen saturation was less than 70 percent, dobutamine administration was started at a dose of 2.5 μg per kilogram of body weight per minute, a dose that was increased by 2.5 μg per kilogram per minute every 30 minutes until the central venous oxygen saturation was 70 percent or higher or until a maximal dose of 20 μg per kilogram per minute was given. Dobutamine was decreased in dose or discontinued if the mean arterial pressure was less than 65 mm Hg or if the heart rate was above 120 beats per minute. To decrease oxygen consumption, patients in whom hemodynamic optimization could not be achieved received mechanical ventilation and sedatives.

“Standard Therapy” was also performed in the same unit by the same team. Both EGDT and standard therapy involved placing a central venous line, but once the line was placed the standard therapy patients were admitted to the ICU, while the EGDT patients were kept in the ED unit for 6+ hours. The physicians who ultimately assumed care in the ICU were blinded to initial treatment group.

One of the main enrollment criteria for this study was a serum lactate >= 4.0 mmol/L.

At many community hospitals, including Mills-Peninsula Hospital in Burlingame, this study has resulted in an imperative to place a central line in all patients with a significant infection and a lactate >= 4.0.

b) Is this imperative to place a central line in all patients with severe sepsis consistent with the results to this study? Explain your answer.

While I do think the general directive may be a good idea, it cannot be a specific response to this study, because both the EGDT group and the “Standard Care” group in this study received central lines. Clearly, it was not the simple placement of a central line that conferred the mortality advantage. (Clinical note: The EGDT group were in the ED 6+ hours being monitored by ED nurses and physicians, with transduction of CVP and monitoring of SvO2. Although we place the lines, we cannot monitor CVP, much less SvO2, and we certainly can’t keep these very sick patients in the ED for 6 hours. The ProCESS Study is a 5-year multicenter trial attempting to replicate these results with 3 arms: EGDT, EGDT – CVP line, standard therapy: )

5. Aspirin and t-PA for Acute MI (Same as EBD Problem 9.5)

In the chapter, we learned from the ISIS-2 Study (3) that aspirin therapy (one month of 160 mg/day) in patients with acute myocardial infarction (AMI) reduced 30-day cardiovascular mortality from 11.8% in the placebo group to 9.4% in the aspirin group (P < 0.00001).

a) What is the absolute risk reduction (ARR)?

11.8% - 9.4% = 2.4%

b) How many AMI patients need to be treated with aspirin for 30 days to prevent one death?

NNT = 1/ARR = 1/2.4% ≈ 42

c) A 120-pill bottle of 81 mg aspirin tablets costs about $5.00. Considering only the cost of the treatment drug, use this to calculate the cost effectiveness ratio ($/Cardiovascular death prevented) for aspirin therapy of AMI. (Be careful, you only need 60 aspirin pills to treat one patient for 30 days.)

2 pills/person treated per day × $5/120 pills ×42 people need to treat × 30 days/cardiovascular death prevented = $104 / cardiovascular death prevented

ISIS-2 and other studies also showed that, when added to aspirin, thrombolysis with streptokinase reduced mortality even further. The GUSTO Study (4) compared thrombolysis using tissue plasminogen activator (tPA) with thrombolysis using streptokinase (SK) in reducing the 30-day mortality of patients with acute myocardial infarction (AMI) receiving aspirin. This study showed a 14% relative risk reduction in the tPA-treated patients. Controversy exists because other large studies found no difference in 30-day mortality. Also, only the subgroup of US (not European) patients showed a benefit, and it was not a blinded study. However, for this problem assume that the 14% relative risk reduction is real. The 30-day mortality risk in the SK group was 7.3%.

d) What was the 30-day mortality risk in the tPA group?

The RRR was 14%, so the RR would be 1-14% = 0.86.

The risk in the tPA group would be the risk in the SK group × the RR: 0.86×7.3% = 6.3%

e) What was the absolute risk reduction associated with tPA (vs. SK)?

The ARR = 7.3% - 6.3% = 1.0%

(Alternatively, could also do 14%×7.3% = 1.0%.)

f) How many patients need to be treated with tPA instead of SK to prevent one death?

The NNT = 1/ARR = 1/1% = 100

g) The cost of tPA is about $3400 per course of treatment, and the cost of SK is about $560[1]. (5) What is the approximate treatment drug cost per death prevented by using tPA instead of SK?

Additional cost of tPA = $3400 - $560 = $2840

NNT = 100

NNT×Cost = 100×$2840 = $284,000 per death prevented.

6. Clopidogrel (Plavix) for TIA

The Fast Assessment of Stroke and TIA to prevent Early Recurrence (FASTER) Trial was a randomized controlled trial of clopidogrel (Plavix®) vs. placebo to prevent stroke at 90 days in patients within 24 hours of TIA.(6) All patients in the trial received aspirin. Here are the results:

A protocol referring to this study stated

“… clopidogrel combined with aspirin reduced the 90-day risk of stroke by 36% compared to aspirin alone in a pilot trial of 392 patients treated acutely after minor stroke or TIA, and it was well tolerated.”

a) To what measure of treatment effect (e.g. RR, OR, RRR, ARR, etc.) does the 36% refer?

RRR

b) Based on this study result, how many TIA patients do we have to treat with colpidogrel plus aspirin instead of aspirin alone to prevent 1 from having a stroke in the subsequent 90 days?

R(Placebo) = 21/194 = 10.8%

R(Clopidogrel) = 14/198 = 7.1%

ARR = 3.8% (not 3.7% which represents a rounding error)

1/ARR = NNT = 27

Currently clopidogrel is available only as Plavix® manufactured by Bristol-Myers Squibb in conjunction with Sanofi-Aventis Pharmaceuticals. (It is expected to go off patent in 2011.) A quick look at drugstore.com (11/7/2010) shows that 1 pill costs about $5.

A 90 day course of clopidogrel requires 97 pills (each 75 mg). This is because the first dose is 8 pills, then 1 pill per day afterwards. At the price of $5 per pill, how much does it cost in Plavix alone to prevent 1 stroke at 90 days in a TIA patient (whom you can treat within 24 hours of symptoms onset).

Rough numbers are fine. ~ 100 pills × $5 per pill = $500 per treatment × 27 = $13,500.

Exact (and spuriously precise) answer is $12,919.43

7. t-PA for Acute Ischemic Stroke

A New York Times article (5/28/2007) summarized the results of the NINDS randomized blinded trial of t-PA (tissue plasminogen activator) within 3 hours of symptom onset in acute ischemic stroke.(7)

In 1995, after 40 years of trying to find something to break up blood clots in the brain, the cause of most strokes, researchers announced that tPA worked. A large federal study showed that, without it, about one patient in five escaped serious injury. With it, one in three escaped. The drug had a serious side effect — it could cause potentially life-threatening bleeding in the brain in about 6 percent of patients.

This is an accurate summary of the study’s results, although we emphasize that the study was limited to patients within 3 hours of symptom onset.

a) What was the risk of serious injury in the placebo group?

1 in 5 escaped serious injury, so 4 in 5 suffered serious injury.

4/5 = 80%

b)What was the risk of serious injury in the t-PA group?

1 in 3 escaped, so 2 in 3 suffered serious injury

2/3 = 67%

c)What is the absolute risk reduction (ARR) and number needed to treat (NNT) with tPA to prevent one ischemic stroke patient from suffering serious injury?

ARR = 80% - 67% = 13%

NNT = 1/13% ≈ 8

d)What is tPA’s absolute risk increase (ARI) for serious intracranial bleed (“bleeding in the brain”)? What is the number needed to harm?

From the abstract of the NINDS paper: “Symptomatic intracerebral hemorrhage within 36 hours after the onset of stroke occurred in 6.4 percent of patients given t-PA but only 0.6 percent of patients given placebo (P < 0.001).” So the NY Times article is correct in saying that tPA caused bleeds in about 6% of patients.

ARI = 6%, NNH = 1/6% ≈17

e)How many serious injuries are prevented for each serious intracranial bleed caused? (Warning: the serious intracranial bleeds were also counted as serious injuries – the primary study outcome.)

This is a trick question. The initial inclination is to calculate the ratio as follows:

ARR/ARI = 13%/6% = 2.2

However, this only applies when the negative effect of the treatment is not included among the primary bad outcome. In the chapter, this was true of nausea associated with oseltamivir prophylaxis (which was obviously not counted as a case of the flu). However, in the NINDS Study, the serious intracranial bleeds were also counted as serious injuries included in the 67% risk of the TPA group. The 6% serious injuries due to intracranial bleeds should be subtracted from 67% injuries in the tPA group to give 61% serious injuries due to ischemia – not bleed. The ARR due to tPA is then 80% - 61% = 19% and the correct ratio would be 19%/6% = 3.2 severe ischemic injuries prevented per intracranial bleed caused.

1. Coiling versus Clipping for Bleeding Cerebral Aneurysms

Case: A previously health 62 year old man presents to the ED with 24-hours of headache. (See SG-1, Problem 8.) A CT of the brain shows subarachnoid hemorrhage and magnetic resonance angiography shows a leaking cerebral artery aneurysm. (For those who want more detail, it is an “8 x 7 x 7 mm leaking anterior communicating artery aneurysm with localized adjacent subarachnoid hemorrhage in the interhemispheric fissure and medial margin of the right sylvian fissure.”)

Clinical Information: The question arises as to whether to treat this problem with “clipping” or “coiling.” “Clipping” is an operating room procedure that involves a craniotomy (opening the skull) and placing a metal clip at the base of the aneurysm. “Coiling” is a transvascular procedure done in the interventional radiology suite and involves passing an arterial catheter through the neck and into the brain, then occluding the aneurysm by feeding a thrombogenic wire into it. Clipping has higher operative risk but a lower chance of re-bleeding.

The International Subarachnoid Aneurysm Treatment (ISAT) Trial(8) was a randomized trial of the effect of coiling vs. clipping on the outcome of death or dependency in patients with bleeding cerebral aneurysms who could be treated with either modality. Here are the results.

Dead or Dependent at 12 months:

Coil: 250/1063 = 24% (10 lost to f/u)

Clip: 326/1055 = 31% (15 lost to f/u)

a) What is the NNT? Explain it in a sentence that might be comprehensible to someone unfamiliar with NNT.

ARR = 31% - 24% = 7%; NNT = 1/ARR = 1/7% = 14.

“You will prevent one death or serious disability for every 14 people you treat with coiling vs. clipping (when you have the choice).”

b) It was impossible to blind the patients or their treating physicians to treatment group (coiling vs. clipping). The assessment of outcome (death or disability) was done by examiners/interviewers blinded to treatment group. How would these facts affect the trial results?

Probably very little. You blind patients and treating physicians to avoid differential co-interventions, but the opportunities for differential co-interventions that would have a significant effect on the outcome (death or disability) are minimal. The outcome of death is obviously objective. Patients are also involved in assessment of disability as an outcome, since they report their level of disability to the (blinded) examiners, so there could be some bias here, but it’s not clear whether this would favor coiling or clipping, and the effect would probably be very small.

REFERENCES

1.Welliver R, Monto AS, Carewicz O, Schatteman E, Hassman M, Hedrick J, et al. Effectiveness of oseltamivir in preventing influenza in household contacts: a randomized controlled trial. Jama. 2001 Feb 14;285(6):748-54.

2.Rivers E, Nguyen B, Havstad S, Ressler J, Muzzin A, Knoblich B, et al. Early goal-directed therapy in the treatment of severe sepsis and septic shock. N Engl J Med. 2001 Nov 8;345(19):1368-77.

3.ISIS-2. Randomised trial of intravenous streptokinase, oral aspirin, both, or neither among 17,187 cases of suspected acute myocardial infarction: ISIS-2. ISIS-2 (Second International Study of Infarct Survival) Collaborative Group. Lancet. 1988 Aug 13;2(8607):349-60.

4.GUSTO. An international randomized trial comparing four thrombolytic strategies for acute myocardial infarction. The GUSTO investigators. N Engl J Med. 1993 Sep 2;329(10):673-82.

5.Peacock WF, Hollander JE, Smalling RW, Bresler MJ. Reperfusion strategies in the emergency treatment of ST-segment elevation myocardial infarction. Am J Emerg Med. 2007 Mar;25(3):353-66.

6.Kennedy J, Hill MD, Ryckborst KJ, Eliasziw M, Demchuk AM, Buchan AM. Fast assessment of stroke and transient ischaemic attack to prevent early recurrence (FASTER): a randomised controlled pilot trial. Lancet Neurol. 2007 Nov;6(11):961-9.