Академический Документы
Профессиональный Документы
Культура Документы
MARCELO BRITTO PASSOS AMATO, M.D., CARMEN SILVIA VALENTE BARBAS, M.D., DENISE MACHADO MEDEIROS, M.D.,
RICARDO BORGES MAGALDI, M.D., GUILHERME DE PAULA PINTO SCHETTINO, M.D., GERALDO LORENZI-FILHO, M.D.,
RONALDO ADIB KAIRALLA, M.D., DANIEL DEHEINZELIN, M.D., CARLOS MUNOZ, M.D., ROSELAINE OLIVEIRA, M.D.,
TERESA YAE TAKAGAKI, M.D., AND CARLOS ROBERTO RIBEIRO CARVALHO, M.D.
M
ABSTRACT ECHANICAL ventilation can damage
Background In patients with the acute respiratory the lungs.1,2 Lesions at the alveolar–
distress syndrome, massive alveolar collapse and capillary interface,3 alterations in per-
cyclic lung reopening and overdistention during me- meability,4 and edema5-7 have repeat-
chanical ventilation may perpetuate alveolar injury. edly been shown to occur in animals subjected to
We determined whether a ventilatory strategy de- adverse patterns of mechanical ventilation.
signed to minimize such lung injuries could reduce In clinical practice, however, the “mechanical
not only pulmonary complications but also mortality stretch” caused by conventional ventilation has been
at 28 days in patients with the acute respiratory dis-
found to be detrimental in only a few uncontrolled
tress syndrome.
studies.8-11 Large variations in the susceptibility of
Methods We randomly assigned 53 patients with
early acute respiratory distress syndrome (includ- individual animal species12 and the apparent success
ing 28 described previously), all of whom were re- of mechanical ventilation based on a strategy of
ceiving identical hemodynamic and general support, using the lowest positive end-expiratory pressure
to conventional or protective mechanical ventilation. (PEEP) that results in acceptable oxygenation13,14
Conventional ventilation was based on the strategy suggest that the devastating effects observed in ani-
of maintaining the lowest positive end-expiratory mals cannot be easily extrapolated to humans.
pressure (PEEP) for acceptable oxygenation, with a We recently demonstrated that mechanical lung
tidal volume of 12 ml per kilogram of body weight protection can be provided in patients with the acute
and normal arterial carbon dioxide levels (35 to 38 respiratory distress syndrome, resulting in better pul-
mm Hg). Protective ventilation involved end-expira-
monary function and higher rates of weaning from
tory pressures above the lower inflection point on
the static pressure–volume curve, a tidal volume of the ventilator.15 Briefly, lung protection was based
less than 6 ml per kilogram, driving pressures of less on a strategy of maintaining low inspiratory driving
than 20 cm of water above the PEEP value, permis- pressures ( 20 cm of water above PEEP, with low
sive hypercapnia, and preferential use of pressure- tidal volumes and preferential use of limited airway
limited ventilatory modes. pressure over regulation of arterial carbon dioxide
Results After 28 days, 11 of 29 patients (38 per- levels), with the simultaneous circumvention of alve-
cent) in the protective-ventilation group had died, olar collapse through the use of high PEEP to keep
as compared with 17 of 24 (71 percent) in the con- end-expiratory pressures above the lower inflection
ventional-ventilation group (P0.001). The rates of point (PFLEX) on the static pressure–volume curve of
weaning from mechanical ventilation were 66 per- the respiratory system. The nearly maximal alveolar
cent in the protective-ventilation group and 29 per-
cent in the conventional-ventilation group (P 0.005);
recruitment and aeration accomplished with this
the rates of clinical barotrauma were 7 percent and 42 strategy were intended to minimize shear stresses in
percent, respectively (P 0.02), despite the use of the lung tissue during inspiration.15
higher PEEP and mean airway pressures in the pro- We have extended our earlier report15 and evalu-
tective-ventilation group. The difference in survival ated the effect of mechanical lung protection on sur-
to hospital discharge was not significant; 13 of 29 vival. We hypothesized that preventing the persistent
patients (45 percent) in the protective-ventilation collapse of recruitable units (alveolar units anatomi-
group died in the hospital, as compared with 17 of cally preserved but requiring high opening pressures
24 in the conventional-ventilation group (71 percent, for aeration) and reducing cyclic lung reopening and
P 0.37). stretch during mechanical breaths would result in
Conclusions As compared with conventional ven- lower rates of pulmonary complications and mortal-
tilation, the protective strategy was associated with
improved survival at 28 days, a higher rate of wean-
ing from mechanical ventilation, and a lower rate of
barotrauma in patients with the acute respiratory
distress syndrome. Protective ventilation was not as- From the Respiratory Intensive Care Unit, Pulmonary Division, Hospital
sociated with a higher rate of survival to hospital dis- das Clínicas, University of São Paulo (M.B.P.A., C.S.V.B., D.M.M., R.B.M.,
charge. (N Engl J Med 1998;338:347-54.) G.P.P.S., G.L.-F., R.A.K., D.D., T.Y.T., C.R.R.C.); and the General Inten-
sive Care Unit, Santa Casa de Misericórdia, Porto Alegre (C.M., R.O.) —
©1998, Massachusetts Medical Society.
both in Brazil. Address reprint requests to Dr. Amato at 135 Rua Dr. Joel
Lagos, CEP 05344-000 São Paulo, Brazil.
348 Febru ar y 5 , 1 9 9 8
PFLEX (corresponding to an upward shift in the slope of the curve equal PFLEX plus 2 cm of water. Finally, if a sharp PFLEX could not
and signaling an increment in lung compliance) could be deter- be determined on the pressure–volume curve, an empirical total-
mined for 49 patients, but the corresponding value was used to PEEP value of 16 cm of water was used.15 Recruiting maneuvers
adjust PEEP only in the group assigned to protective mechanical — aimed at reaerating alveolar units requiring very high opening
ventilation. Since this was the only curve calculated during the pressures — were frequently used, especially after inadvertent dis-
protocol, PEEP was then kept constant in this group until the in- connections from the ventilator. Continuous positive airway pres-
spiratory oxygen fraction was less than 0.4.15 After determining sures of 35 to 40 cm of water were applied for 40 seconds, followed
the pressure–volume curve, we randomly assigned the patients to by a careful return to previous PEEP levels. Finally, pressure-con-
one of the two groups. Randomization was performed with sealed trolled inverse-ratio ventilation was used whenever the inspiratory
envelopes and a 1:1 assignment scheme. oxygen fraction was higher than 0.5, in order to decrease minute-
volume requirements.24
General Ventilatory Support
General Support
Protective or conventional mechanical ventilation was rigorous-
ly maintained until the patient was extubated or died. Each pa- All patients were monitored with the Swan–Ganz catheter, and
tient was connected to a closed system for aspirating tracheal se- a stepwise algorithm for hemodynamic optimization15,16 was used.
cretions; the patient remained connected to the ventilator during Measurements of plasma lactate and mixed venous saturation were
aspiration, minimizing temporary drops in airway pressure. In both used to correct imbalances between oxygen transport and demand.
groups, the target partial pressure of arterial oxygen was 80 The pulmonary-artery wedge pressure never exceeded 15 mm Hg.
mm Hg, and the PEEP level was never set below 5 cm of water, Procedures for nutritional support, treatment of infections, and re-
even during weaning from the ventilator. The weaning proce- nal dialysis (when needed) were the same in both groups.15,16 Cor-
dure was the same in the two groups: a gradual decrease in the ticosteroids were given only to patients with Pneumocystis carinii
level of pressure support.15 Patients received ventilation exclusive- pneumonia. No patients received immunotherapy. The protocol
ly through endotracheal tubes. for sedation was the same for both groups, with only two sedatives
prescribed (fentanyl and diazepam) and only one neuromuscular
Conventional Approach paralyzing drug (pancuronium). Although larger doses (up to 9 mg
per day) were used in the protective-ventilation group, continuous
We sought to maintain an arterial carbon dioxide level of 35 to infusions of fentanyl were used in both groups to keep the patients
38 mm Hg, independent of airway pressures, and an inspiratory ox- comfortable. All patients received ranitidine (50 mg intravenously
ygen fraction of less than 0.6 with adequate systemic oxygen deliv- every eight hours) as prophylaxis against bleeding.
ery. To optimize this compromise, we used a stepwise algorithm for
PEEP increments.15,16 Other ventilatory settings were as follows: Statistical Analysis
tidal volume, 12 ml per kilogram (volume-cycled assisted or con-
trolled ventilation); square-wave inspiratory flow rate, 50 to 80 li- The primary end point was survival at 28 days. The effect of
ters per minute (adjusted to avoid auto-PEEP, or abnormal gas the protective approach was analyzed with a Cox proportional-
trapping leading to an elevated end-respiratory pressure); inspirato- hazards model, with the base-line adjusted APACHE II score (ad-
ry pause, 0.4 second; and backup respiratory rate, 10 to 24 cycles justed risk of death) included as a covariate.
per minute (depending on the value for arterial carbon dioxide). After the first block of 28 patients had been enrolled, a bene-
In addition to the administration of sedative drugs to keep the pa- ficial effect of the protective approach on pulmonary function be-
tients comfortable, additional doses of sedatives were given to pre- came evident,15 and we were concerned about the possibility of
subjecting the patients to an unnecessary continuation of the pro-
vent patient-triggered respiratory rates higher than 24 cycles per
tocol.25 Therefore, we performed an interim analysis after each
minute or arterial carbon dioxide values lower than 25 mm Hg.
new block of five patients. We estimated that a maximal sample
of 58 patients was required, assuming a type I error of 5 percent,
Protective Approach
a statistical power of 85 percent, and a survival rate in the protec-
The protective approach was intended to prevent alveolar col- tive-ventilation group that would be 2.4 times that in the con-
lapse and overdistention, regardless of arterial carbon dioxide lev- ventional-ventilation group, according to our initial results.15
els, and to maintain an “open lung” independently of hemo- To counterbalance the increased chance of prematurely stop-
dynamic conditions. The tidal volume was maintained at a level ping the study because of a type I error, we used the conservative
lower than 6 ml per kilogram, with a respiratory rate of less than correction for multiplicity proposed by Peto et al.26 and Geller
30 cycles per minute, even during pressure support. Permissive and Pocock,27 with a nominal significance level of 0.001 for an
hypercapnia and continuous infusions of fentanyl and diazepam interim analysis, if the study was stopped early, and a significance
were used to prevent discomfort and signs of increased respiratory level of 0.04 for the final analysis, if the study was completed.27
drive. Initial arterial carbon dioxide levels of up to 80 mm Hg The secondary end points were survival to hospital discharge,
were allowed, and slow intravenous sodium bicarbonate infusions occurrence of clinically detectable barotrauma, and weaning rate
(50 mmol per hour) were permitted if the arterial pH was less adjusted for APACHE II score (Cox model). Bonferroni’s adjust-
than 7.2. ment for multiple comparisons was performed for each secondary
Driving pressures (PPLATPEEP, with PPLAT defined as the pla- end point. All other statistical tests are described below. All P val-
teau pressure after the inspiratory pause) and peak airway pres- ues (two-tailed) were calculated with the BMDP software package
sures were kept below 20 and 40 cm of water, respectively. Only (BMDP Statistical Software, version 7.0, Los Angeles).
pressure-limited modes of ventilation (pressure-controlled in-
verse-ratio ventilation [ratio of inspiration to expiration, 1] RESULTS
and pressure-support ventilation, both generating constant air-
way pressure during inspiration) or combined modes (volume-
The study was stopped during the fifth interim
ensured pressure-support ventilation, in which a constant inspir- analysis, after 53 patients had been enrolled, because
atory pressure is targeted at the same time that a minimal tidal of a significant survival difference between the groups
volume is guaranteed23) were used, according to a stepwise algo- (Tables 2 and 3 and Fig. 1). After 28 days, 11 of
rithm.15 29 patients (38 percent) in the protective-ventila-
PEEP was preset at 2 cm of water above PFLEX. When auto-
PEEP (defined as the difference between alveolar pressures at end tion group had died, as compared with 17 of 24
expiration and airway pressures) was present, the total PEEP (ex- (71 percent) in the conventional-ventilation group
ternal PEEP plus auto-PEEP) was considered and adjusted to (P0.001). The results were similar when the groups
PROTECTIVE CONVENTIONAL
VENTILATION VENTILATION
OUTCOME (N29) (N24) P VALUE
COMPARISONS
CORRECTED
ISOLATED FOR MULTIPLE
COMPARISONS TESTING*
were stratified according to the initial severity of ill- (71 percent, P0.37 after adjustment for multiple
ness or the center where the patient was treated. comparisons).
The difference in weaning rates mirrored the Within the first 28 days, the most frequent causes
results for survival, with 19 of 29 patients (66 per- of death were refractory septic shock and progressive
cent) in the protective-ventilation group successful- respiratory failure (Table 2).15 Fourteen episodes of
ly weaned from the ventilator, as compared with 7 of accidental extubation (usually during repositioning
24 (29 percent) in the conventional-ventilation group of the patient) occurred in nine patients in the pro-
(P0.005 after adjustment for multiple compari- tective-ventilation group, as compared with 10 epi-
sons). The rate of clinical barotrauma was also sig- sodes in seven patients in the conventional-ventilation
nificantly lower in the protective-ventilation group group. In two of the patients in the protective-venti-
than in the conventional-ventilation group (7 per- lation group and one in the conventional-ventilation
cent vs. 42 percent, P0.02 after adjustment for group, irreversible cardiac events followed these epi-
multiple comparisons). The difference in survival to sodes. Although successfully extubated (at 48
hospital discharge was not significant; 13 of 29 pa- hours), four patients in the protective-ventilation
tients in the protective-ventilation group (45 per- group died before hospital discharge: one from mas-
cent) died in the hospital, as compared with 17 of sive hemothorax with arterial rupture during attempts
24 patients in the conventional-ventilation group at central venous cannulation (on day 7), one from
350 Febr u ar y 5 , 1 9 9 8
100
80
Protective
Survival (%)
60
P0.001
40
Conventional
20
0
0 10 20 30
Days after Randomization
NO. AT RISK
Protective 29 25 20 18
Conventional 24 11 9 7
Figure 1. Actuarial 28-Day Survival among 53 Patients with the Acute Respiratory Distress Syndrome
Assigned to Protective or Conventional Mechanical Ventilation.
The data are based on an intention-to-treat analysis. The P value indicates the effect of ventilatory
treatment as estimated by the Cox regression model, with the risk of death associated with the adjust-
ed base-line score on APACHE II included as a covariate.
the estimates of relative risk shown in Table 3 may be The strong protective effect associated with a high
imprecise. The corrections proposed for multiple se- PEEP value is consistent with recent experimental
quential analysis can properly control the overall type data,7,29-33 and the benefit seems to be more pro-
I error, but they cannot prevent associated distor- nounced than the deleterious effect of high distend-
tions of the magnitude of the treatment effect caused ing pressures.7,29,30 Had we not used high PEEP
by early termination or the small sample.28 levels (PFLEX), the results might have been very dif-
Since the effect of the protective-ventilation strat- ferent, with the isolated reduction in PPLAT potential-
egy on survival was observed in the context of many ly causing reabsorption atelectasis, loss of alveolar
concomitant maneuvers (permissive hypercapnia, low- surface, and hypoxemia in some patients.
er peak and driving pressures, higher PEEP, a tidal Recent evidence suggests that the minimization
volume of less than 6 ml per kilogram, and so forth), of ventilator-induced lung injury may have impor-
we performed a pooled “retrospective” analysis to de- tant systemic benefits, decreasing the release of pro-
termine the key combination of ventilatory variables inflammatory mediators,34-36 the dissemination of in-
responsible for the ventilatory treatment effect on fections,37-39 and possible complications related to air
mortality at 28 days (data not shown). When the embolism.40,41 In addition to preventing progres-
treatment assignment was removed from the Cox sive respiratory failure, the protective-ventilation ap-
mortality model, there were three significant prog- proach may be associated with these mechanisms.
nostic factors: the APACHE II score, the mean PEEP Despite the use of higher PEEP values (up to 24
used during the first 36 hours (with a protective ef- cm of water) and higher mean airway pressures, there
fect indicated by a coefficient of 0.15), and the was a lower incidence of barotrauma in the protec-
driving pressures (PPLATPEEP) during the first 36 tive-ventilation group. The protective-ventilation ap-
hours (with a deleterious effect of high driving pres- proach may thus not only improve pulmonary func-
sures indicated by a coefficient of 0.06). All other tion and oxygenation but also reduce clinically
respiratory variables were of secondary importance. apparent alveolar damage. Another study suggested a
Higher PEEP values (preferentially above the PFLEX protective effect of PEEP against clinical barotrau-
value) and lower driving pressures were independ- ma.42 The paucity of data in favor of this concept
ently associated with better survival. High initial may be explained by the correlation normally found
PEEP values appeared to be beneficial, even when between PEEP and peak pressures.43,44 In our study,
the PPLAT value increased, as long as the driving pres- however, the use of high PEEP levels did not neces-
sure did not change disproportionately. sarily result in high peak or plateau pressures.
352 Febr u ar y 5 , 1 9 9 8
mean SE
*The values are means of the average values for all measurements in each patient, with all 53 pa-
tients included (intention-to-treat analysis). At least three measurements of all respiratory variables,
along with blood-gas and hemodynamic variables, were performed each day. PEEP denotes positive
end-expiratory pressure, PaCO2 partial pressure of carbon dioxide, and PaO2:FiO2 the ratio of arterial
oxygen tension to the fraction of inspired oxygen. P values are for the comparison between the two
groups during the specified interval, with adjustment for differences in the incremental area under
the curve.15
†P0.001.
‡P0.01.
8. Hickling KG, Henderson SJ, Jackson R. Low mortality associated with clinical trials requiring prolonged observation of each patient. I. Introduc-
low volume pressure limited ventilation with permissive hypercapnia in se- tion and design. Br J Cancer 1976;34:585-612.
vere adult respiratory distress syndrome. Intensive Care Med 1990;16:372- 27. Geller NL, Pocock SJ. Interim analysis in randomized clinical trials:
7. ramifications and guidelines for practitioners. Biometrics 1987;43:213-23.
9. Hickling KG, Walsh J, Henderson S, Jackson R. Low mortality rate in 28. Pocock SJ, Hughes MD. Estimation issues in clinical trials and over-
adult respiratory distress syndrome using low-volume, pressure-limited ven- views. Stat Med 1990;9:657-71.
tilation with permissive hypercapnia: a prospective study. Crit Care Med 29. Muscedere JG, Mullen JBM, Gan K, Slutsky AS. Tidal ventilation at
1994;22:1568-78. low airway pressures can augment lung injury. Am J Respir Crit Care Med
10. Gattinoni L, Pesenti A, Mascheroni D, et al. Low-frequency positive- 1994;149:1327-34.
pressure ventilation with extracorporeal CO2 removal in severe acute respi- 30. Bshouty Z, Ali J, Younes M. Effect of tidal volume and PEEP on rate
ratory failure. JAMA 1986;256:881-6. of edema formation in in situ perfused canine lobes. J Appl Physiol 1988;
11. Lee PC, Helsmoortel CM, Cohn SM, Fink MP. Are low tidal volumes 64:1900-7.
safe? Chest 1990;97:430-4. 31. Sandhar BK, Niblett DJ, Argiras EP, Dunnill MS, Sykes MK. Effects
12. Mathieu-Costello O, Willford DC, Fu Z, Garden RM, West JB. Pul- of positive end-expiratory pressure on hyaline membrane formation in a
monary capillaries are more resistant to stress failure in dogs than in rabbits. rabbit model of the neonatal respiratory distress syndrome. Intensive Care
J Appl Physiol 1995;79:908-17. Med 1988;14:538-46.
13. Albert RK. Least PEEP: primum non nocere. Chest 1985;87:2-4. 32. Argiras EP, Blakeley CR, Dunnill MS, Otremski S, Sykes MK. High
14. Petty TL. The use, abuse, and mystique of positive end-expiratory PEEP decreases hyaline membrane formation in surfactant deficient lungs.
pressure. Am Rev Respir Dis 1988;138:475-8. Br J Anaesth 1987;59:1278-85.
15. Amato MBP, Barbas CSV, Medeiros DM, et al. Beneficial effects of the 33. Wyszogrodski I, Kyei-Aboagye K, Taeusch HW Jr, Avery ME. Surfac-
“open lung approach” with low distending pressures in acute respiratory tant inactivation by hyperventilation: conservation by end-expiratory pres-
distress syndrome: a prospective randomized study on mechanical ventila- sure. J Appl Physiol 1975;38:461-6.
tion. Am J Respir Crit Care Med 1995;152:1835-46. 34. Edmonds HL Jr, Spohr RW, Finnegan RF, et al. Indomethacin pre-
16. Carvalho CRR, Barbas CSV, Medeiros DM, et al. Temporal hemody- treatment in continuous positive-pressure ventilation. Crit Care Med 1981;
namic effects of permissive hypercapnia associated with ideal PEEP in 9:524-9.
ARDS. Am J Respir Crit Care Med 1997;156:1458-66. 35. Tremblay L, Valenza F, Ribeiro SP, Li J, Slutsky AS. Injurious ventila-
17. Murray JF, Matthay MA, Luce JM, Flick MR. An expanded definition tory strategies increase cytokines and c-fos m-RNA expression in an isolated
of the adult respiratory distress syndrome. Am Rev Respir Dis 1988;138: rat lung model. J Clin Invest 1997;99:944-52.
720-3. [Erratum, Am Rev Respir Dis 1989;139:1065.] 36. Sugiura M, McCulloch PR, Wren S, Dawson RH, Froese AB. Ventila-
18. Teboul JL, Besbes M, Andrivet P, et al. A bedside index assessing the tor pattern influences neutrophil influx and activation in atelectasis-prone
reliability of pulmonary artery occlusion pressure measurements during rabbit lung. J Appl Physiol 1994;77:1355-65.
mechanical ventilation with positive end-expiratory pressure. J Crit Care 37. Tilson MD, Bunke MC, Smith GJ, et al. Quantitative bacteriology
1992;7:22-9. and pathology of the lung in experimental Pseudomonas pneumonia
19. Yeung HC, Lu MW, Martinez EG, Puri VK. Critical Care Scoring Sys- treated with positive end-expiratory pressure (PEEP). Surgery 1977;82:
tem — new concept based on hemodynamic data. Crit Care Med 1990; 133-40.
18:1347-52. 38. Johanson WG Jr, Higuchi JH, Woods DE, Gomez P, Coalson JJ. Dis-
20. Smith PEM, Gordon IJ. An index to predict outcome in adult respi- semination of Pseudomonas aeruginosa during lung infection in hamsters:
ratory distress syndrome. Intensive Care Med 1986;12:86-9. role of oxygen-induced lung injury. Am Rev Respir Dis 1985;132:358-61.
21. Knaus WA, Draper EA, Wagner DP, Zimmerman JE. Apache II: a se- 39. Nahum A, Hoyt J, McKibben A, et al. Effect of mechanical ventilation
verity of disease classification system. Crit Care Med 1985;13:818-29. strategy on E. coli pneumonia in dogs. Am J Respir Crit Care Med 1996;
22. Amato MBP, Barbas CSV, Meyer EC, Grunauer MA, Magaldi RB, Car- 153:Suppl:A531. abstract.
valho CRR. Setting the “best PEEP” in ARDS: limitations of choosing the 40. Marini JJ, Culver BH. Systemic gas embolism complicating mechani-
PEEP according to the “best compliance.” Am J Respir Crit Care Med cal ventilation in the adult respiratory distress syndrome. Ann Intern Med
1995;151:Suppl:A550. abstract. 1989;110:699-703.
23. Amato MBP, Barbas CSV, Bonassa J, Saldiva PHN, Zin WA, Carvalho 41. Gregory GA, Tooley WH. Gas embolism in hyaline-membrane disease.
CRR. Volume-assured pressure support ventilation (VAPSV): a new ap- N Engl J Med 1970;282:1141-2.
proach for reducing muscle workload during acute respiratory failure. 42. Mathru M, Rao TLK, Venus B. Ventilator-induced barotrauma in con-
Chest 1992;102:1225-34. trolled mechanical ventilation versus intermittent mandatory ventilation.
24. Gattinoni L, Mascheroni D, Borelli M, Basilico E, Pesenti A. Ventila- Crit Care Med 1983;11:359-61.
tion in severe ARDS: inverted ratio ventilation and CO2 removal. In: Le- 43. Petersen GW, Baier H. Incidence of pulmonary barotrauma in a med-
maire F, ed. Mechanical ventilation. Berlin, Germany: Springer-Verlag, ical ICU. Crit Care Med 1983;11:67-9.
1991:129-45. 44. Zwillich CW, Pierson DJ, Creagh CE, Sutton FD, Schatz E, Petty TL.
25. Pocock SJ. When to stop a clinical trial. BMJ 1992;305:235-40. Complications of assisted ventilation: a prospective study of 354 consecu-
26. Peto R, Pike MC, Armitage P, et al. Design and analysis of randomized tive episodes. Am J Med 1974;57:161-70.
354 Febr u ar y 5 , 1 9 9 8