- Email: [email protected]
Body Mass Index and Stroke: Overweight and Obesity Less Often Associated with Stroke Recurrence
Body Mass Index and Stroke: Overweight and Obesity Less Often Associated with Stroke Recurrence Klaus Kaae Andersen, MS, PhD,* and Tom Skyhøj Olsen, MD, PhD†
Background: Although obesity is associated with excess mortality and morbidity, mortality is lower in obese than in normal weight stroke patients (the obesity paradox). Studies now indicate that obesity is not associated with increased risk of recurrent stroke in the years after first stroke. We studied the association between body mass index (BMI) and stroke patient’s risk of having a history of previous stroke (recurrent stroke). Methods: A registry designed to collect data on all hospitalized stroke patients in Denmark 2000-2010 includes 61,872 acute stroke patients with information on BMI in 38,506. Data include age, sex, civil status, stroke severity (Scandinavian Stroke Scale), computed tomography, and cardiovascular risk factors. There were 28,382 patients with complete covariate information. We used multiple logistic regression models on this data set to compare the risk of stroke being recurrent in the 4 BMI groups: underweight (BMI , 18.5), normal weight (BMI 18.5-24.9), overweight (BMI 25.0-29.9), and obese (BMI $ 30.0). Results: Of the patients with complete covariate information, 22,811 (80.1%) had first-ever stroke; in 5571 patients (19.6%), stroke was recurrent. Multiple logistic regression analysis adjusting for age, stroke severity, sex, BMI, civil status, and cardiovascular risk factors showed that being obese and overweight in comparison with normal weight was associated with a significantly lower risk of stroke being recurrent (obese: odds ratio [OR] 5 .90, confidence interval [CI] .82-.98; overweight: OR 5 .89, CI .83-.96). Being underweight was associated with a significantly higher risk of stroke being recurrent (OR 5 1.23; CI 1.06-1.43). Conclusions: The obesity paradox in stroke can be extended to include also stroke recurrence. Obese and overweight stroke patients had experienced less previous strokes than normal weight stroke patients. Key Words: Obesity—stroke—mortality—recurrent stroke—body mass index. Ó 2013 by National Stroke Association
Introduction ‘‘The stroke obesity paradox’’ originally referred to the paradox that obesity although being associated with higher mortality and morbidity (including stroke) in the
From the *Statistical department, Danish Cancer Society Research Center, Copenhagen, Denmark; and †The Stroke Unit, Frederiksberg University Hospital, Frederiksberg, Denmark. Received April 5, 2013; revision received June 4, 2013; accepted June 19, 2013. Grant support: None. Address correspondence to Tom Skyhøj Olsen, MD, PhD, The Stroke Unit, Frederiksberg University Hospital, DK-2000 Frederiksberg, Denmark. E-mail: [email protected]. 1052-3057/$ - see front matter Ó 2013 by National Stroke Association http://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2013.06.031
e576
general population is associated with lower death rates after stroke when compared with normal weight stroke patients.1 Recent studies now report that the stroke obesity paradox may even be extended to include risk of stroke recurrence. Three studies reported that obesity was not associated with increased risk of stroke recurrence,2-4 and 1 study reported that readmission rates for recurrent stroke was even lower in obese than in normal weight stroke patients.5 The underlying cause of the stroke obesity paradox remains still unknown, and although reported from rather large study populations,2-6 the phenomenon has been questioned as being an artifact.7,8 If obesity is not associated with higher risk of recurrent stroke as observed in prospective studies,2-5 then one would not expect a higher incidence of patients with a history of stroke in obese stroke patients. We studied the relation between body mass index (BMI)
Journal of Stroke and Cerebrovascular Diseases, Vol. 22, No. 8 (November), 2013: pp e576-e581
OVERWEIGHT, OBESITY, AND STROKE
and stroke patient’s risk of having a history of previous stroke based on a Danish cohort of 38,506 stroke patients with information on BMI.
Materials and Methods The study is based on data from the Danish Stroke Registry (previously referred to as the Danish National Indicator Project), described in detail elsewhere.9,10 All Danish hospitals are committed to reporting a predefined set of data into the Danish Stroke Registry on all patients admitted to the hospital with acute stroke. Data include age, gender, civil status, stroke severity measured by the Scandinavian Stroke Scale (SSS),11 stroke subtype, and a predefined cardiovascular profile including BMI. The SSS is a validated neurologic stroke scale evaluating stroke severity on a score from 0 (worst) to 58 (best).9 The stroke subtype (hemorrhage/infarction) is determined after computed tomography/magnetic resonance scan. Measures of BMI on admission (weight [kg]/height [m2] measured by the nursing staff on admission) were available in 38,506 patients. Patients were assigned to 1 of the 4 BMI categories: underweight (BMI , 18.5), normal weight (BMI 18.5-24.9), overweight (BMI 25.029.9), and obese (BMI $ 30.0). The cardiovascular profile also included information on alcohol consumption (#14/21/.14/21 drinks per week for women and men, respectively, representing under/over the limit set by the Danish National Board of Health), current daily smoking, diabetes mellitus (DM), atrial fibrillation (AF; chronic or paroxysmal), arterial hypertension, previous myocardial infarction (MI), and previous stroke. Diagnosis of DM, AF, arterial hypertension, previous MI, and previous stroke is made following the current Danish standards10 and is either known before the onset of stroke or diagnosed during hospitalization. Stroke is defined according to the World Health Organization criteria.12 The inclusion started on May 8, 2000, and the end of follow-up was on June 20, 2010. For patients with multiple records (stroke events within the inclusion period), only the first event was included in the analysis. Patients with transient ischemic attacks and those younger than 18 years were excluded from the study as well as patients in whom computed tomography/magnetic resonance scan was not performed (.4%) or unavailable (.7%). Time origin for the analysis was the date of hospital admission. The Danish Stroke Registry coverage was by professional consensus estimated to be about 80% of all stroke admissions in Denmark.13 The very high proportion of stroke patients were admitted to hospitals (90%) in Denmark.14 The study was approved by the board of the Danish Stroke Registry and the Danish Data Protection Agency.
e577
Statistical Analysis The association between BMI and a history of previous stroke defined as a binary outcome (previous stroke yes/ no) was studied by applying generalized additive logistic regression models. In the analysis, BMI was categorized into 4 groups: underweight (BMI , 18.5), normal weight (BMI 18.5-24.9), overweight (BMI 25.0-29.9), and obese (BMI $ 30.0). Both unadjusted and adjusted estimates were obtained for the categorized BMI variable. In the adjusted analysis, the categorized BMI variable was included together with the following cardiovascular risk factors: diabetes (yes/no), previous MI (yes/no), AF (yes/no), hypertension (yes/no), previous stroke (yes/no), smoking (current smoker/ex-smoker/neversmoker), alcohol consumption (over limit: .14/21 drinks per week for women and men, respectively, versus under limit: #14/21 drinks per week for women and men, respectively), and stroke type (ischemic/hemorrhage). The analysis was further adjusted for age and the SSS score by applying smooth splines (penalized cubic regression spline with 4 degrees of freedom) to study the nonlinear effects of the 2. The estimated association and deviations from linearity between the age and SSS score with BMI were presented graphically with 95% confidence intervals. To describe the missingness mechanism in the data, we estimated and compared the risk factor profile in those with BMI measured and those that did not have BMI measurements available. All analyses were performed using the R statistical software (R Foundation for Statistical Computing, Vienna, Austria),15 and effect estimates are given as odds ratios (ORs) with corresponding 95% confidence intervals (CIs). The difference between 2 means was tested by student t test and between 2 proportions by chi-square test. Test for significance in the adjusted model was done by means of likelihood ratio tests. Effects that met the 5% significance level were considered statistically significant, and all tests were 2 sided.
Results In total, 61,872 patients were admitted with acute stroke and registered in the Danish Stroke Registry on June 20, 2010. Mean age was 71.5 years (SD 13.2), 47.8% were women, and mean SSS score was 41.8 (SD 16.5). BMI was measured in 38,506 (62%) patients. Of these, mean age was 72.3 years, 47% were women, mean SSS score was 43.1, and mean BMI was 23.0 (5.0% were underweight, 44.5% normal weight, 34.0% overweight, and 16.5% obese). Clinical characteristics of patients with BMI recorded versus those not recorded are given in Table 1. A multiple logistic regression analysis adjusting for age, sex, stroke severity, civil status, and cardiovascular risk factors showed that patients in whom BMI was not available had more severe stroke (OR 5 1.09; the SSS score per 10-point decrease; P , .001) and were more likely to have had hemorrhagic stroke (OR 5 1.56;
K.K. ANDERSEN AND T.S. OLSEN
e578
Table 1. Clinical characteristics of patients with BMI recorded versus those not recorded BMI not recorded (%)
BMI recorded (%)
P value
12,008 (51.4) 2907 (12.5) 2107 (9.0) 4994 (21.4) 3754 (16.1) 10,043 (43.1) 2802 (12.0) 1528 (6.6) 6841 (29.4) 11,613 (49.8) 10,264 (57.0) 10,047 (43.0)
20,309 (52.7) 5205 (13.5) 3488 (9.1) 7728 (20.1) 6103 (15.9) 18,933 (49.3) 2419 (6.3) 2832 (7.4) 12,906 (33.6) 20,283 (52.7) 19,184 (56.1) 17,725 (46.0)
.001 ,.001 ,.001 ,.001 ,.001 ,.001 ,.001 ,.001 ,.001 ,.001 .056 ,.001
Gender (male) Diabetes (yes) Previous MI (yes) Previous stroke (yes) AF (yes) Hypertension (yes) Stroke type (hemorrhage) Alcohol consumption (over limit) Smoking (current) Civil status (living with someone) SSS , 50 Age , 71 y
Abbreviations: AF, atrial fibrillation; BMI, body mass index; MI, myocardial infarction; SSS, Scandinavian Stroke Scale.
P , .001). Patients with previous strokes were more likely to have missing information on BMI (OR 5 1.13; P ,.001). Also smoking status (OR 5 1.10; P , .001), hypertension (OR 5 1.13; P , .001), and civil status (OR 5 1.24; P , .001) were associated with missing BMI. Gender, age, and other variables included in our analysis did not differ significantly between patients in whom BMI was available and unavailable. There were 28,382 patients with complete covariate information. Risk factor distribution, age, sex, stroke severity, and civil status according to the BMI category in these patients are shown in Table 2. Prevalence of diabetes, previous MI, and hypertension increased with increasing BMI, whereas prevalence of smoking, AF, previous stroke, and hemorrhagic stroke increased with decreasing BMI. More severe stroke and higher age were associated with lower BMI. Of the 28,382 patients, stroke was first-ever in 22,811 (80.4%; mean age, 69.4 years [SD 13.3 years]; mean SSS
score, 45.6 [SD 13.6]) and in 5571 patients (19.6%) stroke was recurrent (mean age, 72.4 years [SD 11.4 years]; mean SSS score, 43.6 [SD 13.4]). P value is less than .001 for age and the SSS score. Table 3 shows sex, BMI, other cardiovascular risk factors, and civil status in patients with first-ever stroke and those in whom stroke was recurrent. Patients with recurrent stroke were significantly more often men, and they had significantly more often diabetes, previous MI, AF, and hypertension. They were significantly more often living alone, and they were significantly more often underweight and less often overweight, whereas prevalence of obesity did not differ significantly between patients with recurrent and firstever stroke. Figure 1 shows the univariate association between BMI and risk of stroke being recurrent; the risk decreases as BMI increases. The independent association between BMI and risk of stroke being recurrent in a multiple logistic regression analysis adjusting for age, stroke severity
Table 2. Risk factor distribution according to BMI category
Gender (male) Diabetes (yes) Previous MI (yes) Previous stroke (yes) AF (yes) Hypertension (yes) Stroke type (hemorrhage) Alcohol consumption (over limit) Smoking (current) Civil status (living with someone) SSS , 50 Age , 71 y
Underweight (%)
Normal weight (%)
Overweight (%)
Obese (%)
P value
334 (25.7) 66 (5.1) 96 (7.4) 293 (22.5) 221 (17.0) 534 (41.0) 117 (9.0) 113 (8.7) 718 (55.1) 479 (36.8) 785 (68.9) 424 (32.6)
6144 (49.6) 1155 (9.3) 1019 (8.2) 2480 (20.0) 1857 (15.0) 5758 (46.4) 719 (5.8) 1057 (8.5) 5494 (44.3) 6532 (52.7) 6084 (54.6) 5462 (44.1)
6320 (64) 1401 (14.2) 933 (9.5) 1870 (18.9) 1345 (13.6) 5284 (53.5) 487 (4.9) 846 (8.6) 3661 (37.1) 6138 (62.2) 4405 (48.7) 5217 (52.9)
2708 (56.3) 1108 (23.0) 479 (10.0) 928 (19.3) 591 (12.3) 2946 (61.2) 228 (4.7) 390 (8.1) 1686 (35.0) 3024 (62.8) 2132 (48.4) 3096 (64.3)
,.001 ,.001 ,.001 .011 ,.001 ,.001 ,.001 .782 ,.001 ,.001 ,.001 ,.001
Abbreviations: AF, atrial fibrillation; BMI, body mass index; MI, myocardial infarction; SSS, Scandinavian Stroke Scale. The table is based on completed cases.
OVERWEIGHT, OBESITY, AND STROKE
e579
Table 3. Gender, BMI, cardiovascular risk factors, and civil status in patients with first-ever stroke and patients in whom stroke was recurrent
Gender (male) Diabetes (yes) Previous MI (yes) AF (yes) Hypertension Stroke type (infarction) Alcohol consumption (over limit) Smoking (current) Civil status (living with someone) Underweight Normal weight Overweight Obese
All, n (%)
First stroke, n (%)
Recurrent stroke, n (%)
P
15,506 (54.6) 3730 (13.1) 2527 (8.9) 4014 (14.1) 14,522 (51.0) 26,831 (94.5) 2406 (8.5) 11,559 (40.1) 16,173 (57.0) 1302 (4.6) 12,399 (43.7) 9869 (34.8) 4812 (17.0)
12,307 (53.9) 2764 (12.1) 1810 (7.9) 3022 (13.2) 11,044 (48.4) 21,549 (94.4) 1968 (8.6) 9415 (33.2) 13,167 (57.7) 1009 (4.4) 9919 (43.5) 7999 (35.1) 3884 (17.0)
3199 (57.4) 966 (17.3) 717 (12.9) 992 (17.8) 3478 (62.4) 5282 (94.8) 438 (7.9) 2144 (38.5) 3006 (54.0) 293 (5.2) 2480 (44.5) 1870 (33.6) 928 (16.7)
,.001 ,.001 ,.001 ,.001 ,.001 .33 .07 ,.001 ,.001 .03 Reference .05 .29
Abbreviations: AF, atrial fibrillation; BMI, body mass index; MI, myocardial infarction.
(SSS), sex, BMI (4 groups), civil status, and cardiovascular risk factors is shown in Table 4. It appears that being obese and overweight (using normal weight as reference) is associated with a significantly lower risk of stroke being recurrent (obese: OR 5 .90; CI .82-.98; overweight: OR 5 .89; CI .83-.96). Being underweight is associated with a significantly higher risk of stroke being recurrent (OR 5 1.23; CI 1.06-1.43). Diabetes (OR 5 1.31; CI 1.20-1.43), hypertension (OR 5 1.68; CI 1.57-1.79), previous MI (OR 5 1.35; CI 1.23-1.49), AF (OR 5 1.19; CI 1.09-1.30), and being men (OR 5 1.29; CI 1.20-1.38) were associated with a significantly higher risk of stroke being recurrent, whereas living with someone was associated with a significantly
lower risk of stroke being recurrent (OR 5 .93; CI .87-1.00). The partial effect of age and stroke severity (SSS) on BMI is shown in Figure 2. From the age of 60 years and older, BMI was decreasing steeply with age, about 1 point BMI per 10 years increase. Milder strokes tended to have higher BMI.
Discussion Among patients admitted with stroke, the risk that stroke was recurrent was inversely related to BMI: previous stroke was less often reported in overweight and obese stroke patients than in normal weight patients, no Table 4. Multivariate estimates on the risk of stroke being recurrent
Gender (male) Diabetes (yes) Previous MI (yes) AF (yes) Hypertension Stroke type (infarction) Alcohol consumption (over limit) Smoking (current) Civil status (living with someone) Underweight Normal weight Overweight Obese
Figure 1. Univariate association between BMI and the risk of stroke being recurrent. The confidence intervals are 95% limits of the fit. Abbreviations: BMI, body mass index; OR, odds ratio.
OR
95% CI
P
1.29 1.31 1.35 1.19 1.68 1.10 .98
1.20-1.38 1.20-1.43 1.23-1.49 1.09-1.30 1.57-1.79 .96-1.27 .88-1.12
0 0 0 0 0 .16 .86
1.07 .93
.98-1.15 .87-1.00
.12 .04
1.23 Reference .89 .90
1.06-1.42
.01
.83-.96 .82-.98
0 .02
Abbreviations: AF, atrial fibrillation; CI, confidence interval; MI, myocardial infarction; OR, odds ratio. Multiple logistic regression analysis adjusting for age, sex, stroke severity, stroke type, civil status, and cardiovascular risk factors.
K.K. ANDERSEN AND T.S. OLSEN
27
27
26
26 BMI
BMI
e580
25 24
24
23
23
20
40
60 Age
80
100
Figure 2. Association between age and BMI (left) and the SSS score and BMI (right). The confidence intervals are 95% limits of the fit. Abbreviations: BMI, body mass index; SSS, Scandinavian Stroke Scale.
25
0
10
matter age, sex, stroke severity, and cardiovascular risk factor profile. Conversely, underweight stroke patients had more often a history of a previous stroke. There is accumulating evidence that overweight and obesity are not associated with increased risk of stroke recurrence.2-4 Recently, data from the Danish Stroke Registry indicated that the risk of readmission for recurrent stroke was even lower in obese than in normal weight stroke patients.5 These findings are further supported by the findings of the present study indicating that obesity is not only associated with decreased risk of recurrent stroke in the years after stroke. In patients admitted to the hospital with acute stroke, a history of a previous stroke is also less frequent in obese and overweight patients than in normal weight and underweight stroke patients. Our study now adds to the evidence of another dimension of the obesity paradox1 that obesity although associated with increased risk of stroke in primary prevention studies16,17 is not associated with higher risk of stroke recurrence in secondary prevention studies. There is still no explanation of this new aspect of the obesity paradox in stroke, also observed in prospective studies.2-5 Higher rate of antihypertensive use in obese patients has been suggested as a potential explanation,3 but in our study, the lower risk of stroke being recurrent was independent of hypertension and other cardiovascular risk factors. Another suggestion was metabolic consequences of skeletal muscle tissue wasting because of the incident stroke,4 but this does not explain lower prevalence of previous stroke seen in this study. Surprisingly, positive health indicators such as nonsmoking and nondrinking are more common among obese and overweight stroke patients.5 Also, cohabitation is more common among obese and overweight stroke patients.5 Hence, lifestyle or social factors hitherto not addressed when evaluating health effects of increased BMI may counteract adverse effects of DM and hypertension occurring more often in obese and overweight stroke patients. However, the findings in our study were independent of hypertension, diabetes, and civil status.
20
30 40 SSS score
50
60
Prevalence of traditional cardiovascular risk factors for stroke was significantly associated with BMI. Hence, prevalence of diabetes, hypertension, and previous MI increased as BMI increased, whereas prevalence of AF and smoking decreased as BMI increased. The remarkably higher prevalence of current smoking in addition to higher age and higher prevalence of AF might explain higher prevalence of recurrent stroke among underweight stroke patients. However, the higher prevalence remained even after adjustment for these variables. In accordance with already established knowledge on risk factors for stroke,15 we found that diabetes, previous MI, AF, and hypertension were associated with increased frequency of previous stroke in our multivariate analysis also when adjusting for BMI. Our study has weaknesses and strengths. Our study’s strength is first of all in its large sample size allowing for sufficient statistical power. Second, we included patients without limitations on age (.18 years), gender, or stroke severity. Third, in our patients, stroke severity was measured on admission to the hospital using a wellvalidated neurologic scale. Although the Danish Stroke Registry is designed as a nationwide registration of all patients admitted with acute stroke, coverage is not yet complete (presently about 80%).11 It is a weakness of concern that of the 61,872 patients registered in the Danish Stroke Registry, only 62% had information on BMI; however, for all other variables, data completeness is high exceeding 85% for all individual variables. The relatively high rate of missing data on BMI when compared with other variables in stroke databases is a well-known problem,4 reflecting that the clinical usefulness of this information still has not achieved general acceptance. However, we performed an analysis of the missing observations and did not find any evidence of informative missingness and, thus, biased estimates. Except for higher prevalence of hemorrhagic stroke and severe strokes, differences between patients with or without measurement of BMI in regard to age, sex, and risk factors were of no clinical importance. Thus, for the purpose of establishing a model of high
OVERWEIGHT, OBESITY, AND STROKE
statistical validity, we believe that our data set is appropriate as the large sample size allows us to adjust reliably for differences in age, sex, civil status, stroke severity, and risk factor profile. We cannot exclude, however, the possibility of bias because of variables not recorded in the Danish Stroke Registry. We have no information as to treatments or interventions that might have influenced stroke recurrence, and we have no information on possible differences between BMI groups in regard to preventive treatment and compliance to treatment. In conclusion, it is generally accepted to recommend patients with stroke to aim at normal weight, that is, BMI between 18.5 and 25.16 There is no doubt that overweight and obesity are associated with increased risk of a number of diseases, among these also stroke.17,18 However, our study supports findings from previous research that obesity is not associated with increased risk of recurrent stroke in stroke patients.2-5 These findings are still unexplained calling for further studies as more than half of stroke patients are overweight or obese, and the effect of weight reduction on prognosis in stroke is still unknown.
References 1. McAuley PA, Blair SN. Obesity paradoxes. J Sports Sci 2011;29:773-782. 2. Vemmos K, Ntaios G, Spengos K, et al. Association between obesity and mortality after acute first-ever stroke: the obesity–stroke paradox. Stroke 2011;42:30-36. 3. Ovbiagele B, Bath PM, Cotton D, et al. Obesity and recurrent vascular event after recent ischemic stroke. Stroke 2011;42:3397-3402. 4. Doehner W, Schenkel J, Anker SD, et al. Overweight and obesity are associated with improved survival, functional outcome, and stroke recurrence after acute stroke or transient ischaemic attack: observations from the TEMPiS trial. Eur Heart J 2013;34:268-277. 5. Andersen KK, Olsen TS. The obesity paradox in stroke: lower mortality and lower risk of readmission for recurrent stroke in obese stroke patients. Int J Stroke
e581
6.
7. 8. 9.
10.
11.
12.
13.
14.
15. 16.
17.
18.
2013; March 12. http://dx.doi.org/10.1111/ijs.12016. [Epub ahead of print]. Olsen TS, Dehlendorff C, Petersen HG, et al. Body mass index and poststroke mortality. Neuroepidemiology 2008;30:93-100. Bray AB. The obesity paradox—an artifact of small sample size? Nat Rev Cardiol 2009;6:561-562. Gielen S, Sandri M. The obesity paradox—a scientific artifact? Int J Cardiol 2013;162:140-142. Olsen TS, Dehlendorff C, Andersen KK. Sex-related timedependent variations in poststroke survival—evidence of a female stroke survival advantage. Neuroepidemiology 2007;29:218-225. Mainz J, Krog BR, Bjønshave B, Bartels P. Nationwide continuous quality improvement using clinical indicators: the Danish National Indicator Project. Int J Qual Health Care 2004;16(Suppl 1):145-150. Lindenstrøm E, Boysen G, Christiansen LW, et al. Reliability of Scandinavian Stroke Scale. Cerebrovasc Dis 1991;1:103-107. Report of the WHO Task Force on stroke and other cerebrovascular disorders: stroke—1989: recommendations on stroke prevention, diagnosis, and therapy. Stroke 1989;20:1407-1431. NIP-apopleksi. National auditrapport 2010. Version 3; Aarhus, Denmark, 2011. Available at: https://www.sundhed. dk/content/cms/69/4669_2010_apopleksi_nat_sfa.pdf. Accessed March 20, 2013. Jørgensen HS, Plesner AM, H€ ubbe P, et al. Marked increase of stroke incidence in men between 1972 and 1990 in Frederiksberg, Denmark. Stroke 1992;23: 1701-1704. Ihaka I, Gentlemann RR. A language for data analysis and graphics. J Comput Graph Stat 1995;5:299-314. Sacco RL, Adams R, Albers G, et al. Guidelines for prevention of stroke in patients with ischemic stroke or transient ischemic attack: a statement for healthcare professionals from the American Heart Association/American Stroke Association Council on Stroke. Stroke 2006; 37:577-617. Kurth T, Gaziano JM, Berger K, et al. Prospective study of body mass index and risk of stroke in apparently healthy women. Circulation 2005;111:1992-1998. Mann GV. The influence of obesity on health (second of two parts). N Engl J Med 1974;291:226-232.
Background: Although obesity is associated with excess mortality and morbidity, mortality is lower in obese than in normal weight stroke patients (the obesity paradox). Studies now indicate that obesity is not associated with increased risk of recurrent stroke in the years after first stroke. We studied the association between body mass index (BMI) and stroke patient’s risk of having a history of previous stroke (recurrent stroke). Methods: A registry designed to collect data on all hospitalized stroke patients in Denmark 2000-2010 includes 61,872 acute stroke patients with information on BMI in 38,506. Data include age, sex, civil status, stroke severity (Scandinavian Stroke Scale), computed tomography, and cardiovascular risk factors. There were 28,382 patients with complete covariate information. We used multiple logistic regression models on this data set to compare the risk of stroke being recurrent in the 4 BMI groups: underweight (BMI , 18.5), normal weight (BMI 18.5-24.9), overweight (BMI 25.0-29.9), and obese (BMI $ 30.0). Results: Of the patients with complete covariate information, 22,811 (80.1%) had first-ever stroke; in 5571 patients (19.6%), stroke was recurrent. Multiple logistic regression analysis adjusting for age, stroke severity, sex, BMI, civil status, and cardiovascular risk factors showed that being obese and overweight in comparison with normal weight was associated with a significantly lower risk of stroke being recurrent (obese: odds ratio [OR] 5 .90, confidence interval [CI] .82-.98; overweight: OR 5 .89, CI .83-.96). Being underweight was associated with a significantly higher risk of stroke being recurrent (OR 5 1.23; CI 1.06-1.43). Conclusions: The obesity paradox in stroke can be extended to include also stroke recurrence. Obese and overweight stroke patients had experienced less previous strokes than normal weight stroke patients. Key Words: Obesity—stroke—mortality—recurrent stroke—body mass index. Ó 2013 by National Stroke Association
Introduction ‘‘The stroke obesity paradox’’ originally referred to the paradox that obesity although being associated with higher mortality and morbidity (including stroke) in the
From the *Statistical department, Danish Cancer Society Research Center, Copenhagen, Denmark; and †The Stroke Unit, Frederiksberg University Hospital, Frederiksberg, Denmark. Received April 5, 2013; revision received June 4, 2013; accepted June 19, 2013. Grant support: None. Address correspondence to Tom Skyhøj Olsen, MD, PhD, The Stroke Unit, Frederiksberg University Hospital, DK-2000 Frederiksberg, Denmark. E-mail: [email protected]. 1052-3057/$ - see front matter Ó 2013 by National Stroke Association http://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2013.06.031
e576
general population is associated with lower death rates after stroke when compared with normal weight stroke patients.1 Recent studies now report that the stroke obesity paradox may even be extended to include risk of stroke recurrence. Three studies reported that obesity was not associated with increased risk of stroke recurrence,2-4 and 1 study reported that readmission rates for recurrent stroke was even lower in obese than in normal weight stroke patients.5 The underlying cause of the stroke obesity paradox remains still unknown, and although reported from rather large study populations,2-6 the phenomenon has been questioned as being an artifact.7,8 If obesity is not associated with higher risk of recurrent stroke as observed in prospective studies,2-5 then one would not expect a higher incidence of patients with a history of stroke in obese stroke patients. We studied the relation between body mass index (BMI)
Journal of Stroke and Cerebrovascular Diseases, Vol. 22, No. 8 (November), 2013: pp e576-e581
OVERWEIGHT, OBESITY, AND STROKE
and stroke patient’s risk of having a history of previous stroke based on a Danish cohort of 38,506 stroke patients with information on BMI.
Materials and Methods The study is based on data from the Danish Stroke Registry (previously referred to as the Danish National Indicator Project), described in detail elsewhere.9,10 All Danish hospitals are committed to reporting a predefined set of data into the Danish Stroke Registry on all patients admitted to the hospital with acute stroke. Data include age, gender, civil status, stroke severity measured by the Scandinavian Stroke Scale (SSS),11 stroke subtype, and a predefined cardiovascular profile including BMI. The SSS is a validated neurologic stroke scale evaluating stroke severity on a score from 0 (worst) to 58 (best).9 The stroke subtype (hemorrhage/infarction) is determined after computed tomography/magnetic resonance scan. Measures of BMI on admission (weight [kg]/height [m2] measured by the nursing staff on admission) were available in 38,506 patients. Patients were assigned to 1 of the 4 BMI categories: underweight (BMI , 18.5), normal weight (BMI 18.5-24.9), overweight (BMI 25.029.9), and obese (BMI $ 30.0). The cardiovascular profile also included information on alcohol consumption (#14/21/.14/21 drinks per week for women and men, respectively, representing under/over the limit set by the Danish National Board of Health), current daily smoking, diabetes mellitus (DM), atrial fibrillation (AF; chronic or paroxysmal), arterial hypertension, previous myocardial infarction (MI), and previous stroke. Diagnosis of DM, AF, arterial hypertension, previous MI, and previous stroke is made following the current Danish standards10 and is either known before the onset of stroke or diagnosed during hospitalization. Stroke is defined according to the World Health Organization criteria.12 The inclusion started on May 8, 2000, and the end of follow-up was on June 20, 2010. For patients with multiple records (stroke events within the inclusion period), only the first event was included in the analysis. Patients with transient ischemic attacks and those younger than 18 years were excluded from the study as well as patients in whom computed tomography/magnetic resonance scan was not performed (.4%) or unavailable (.7%). Time origin for the analysis was the date of hospital admission. The Danish Stroke Registry coverage was by professional consensus estimated to be about 80% of all stroke admissions in Denmark.13 The very high proportion of stroke patients were admitted to hospitals (90%) in Denmark.14 The study was approved by the board of the Danish Stroke Registry and the Danish Data Protection Agency.
e577
Statistical Analysis The association between BMI and a history of previous stroke defined as a binary outcome (previous stroke yes/ no) was studied by applying generalized additive logistic regression models. In the analysis, BMI was categorized into 4 groups: underweight (BMI , 18.5), normal weight (BMI 18.5-24.9), overweight (BMI 25.0-29.9), and obese (BMI $ 30.0). Both unadjusted and adjusted estimates were obtained for the categorized BMI variable. In the adjusted analysis, the categorized BMI variable was included together with the following cardiovascular risk factors: diabetes (yes/no), previous MI (yes/no), AF (yes/no), hypertension (yes/no), previous stroke (yes/no), smoking (current smoker/ex-smoker/neversmoker), alcohol consumption (over limit: .14/21 drinks per week for women and men, respectively, versus under limit: #14/21 drinks per week for women and men, respectively), and stroke type (ischemic/hemorrhage). The analysis was further adjusted for age and the SSS score by applying smooth splines (penalized cubic regression spline with 4 degrees of freedom) to study the nonlinear effects of the 2. The estimated association and deviations from linearity between the age and SSS score with BMI were presented graphically with 95% confidence intervals. To describe the missingness mechanism in the data, we estimated and compared the risk factor profile in those with BMI measured and those that did not have BMI measurements available. All analyses were performed using the R statistical software (R Foundation for Statistical Computing, Vienna, Austria),15 and effect estimates are given as odds ratios (ORs) with corresponding 95% confidence intervals (CIs). The difference between 2 means was tested by student t test and between 2 proportions by chi-square test. Test for significance in the adjusted model was done by means of likelihood ratio tests. Effects that met the 5% significance level were considered statistically significant, and all tests were 2 sided.
Results In total, 61,872 patients were admitted with acute stroke and registered in the Danish Stroke Registry on June 20, 2010. Mean age was 71.5 years (SD 13.2), 47.8% were women, and mean SSS score was 41.8 (SD 16.5). BMI was measured in 38,506 (62%) patients. Of these, mean age was 72.3 years, 47% were women, mean SSS score was 43.1, and mean BMI was 23.0 (5.0% were underweight, 44.5% normal weight, 34.0% overweight, and 16.5% obese). Clinical characteristics of patients with BMI recorded versus those not recorded are given in Table 1. A multiple logistic regression analysis adjusting for age, sex, stroke severity, civil status, and cardiovascular risk factors showed that patients in whom BMI was not available had more severe stroke (OR 5 1.09; the SSS score per 10-point decrease; P , .001) and were more likely to have had hemorrhagic stroke (OR 5 1.56;
K.K. ANDERSEN AND T.S. OLSEN
e578
Table 1. Clinical characteristics of patients with BMI recorded versus those not recorded BMI not recorded (%)
BMI recorded (%)
P value
12,008 (51.4) 2907 (12.5) 2107 (9.0) 4994 (21.4) 3754 (16.1) 10,043 (43.1) 2802 (12.0) 1528 (6.6) 6841 (29.4) 11,613 (49.8) 10,264 (57.0) 10,047 (43.0)
20,309 (52.7) 5205 (13.5) 3488 (9.1) 7728 (20.1) 6103 (15.9) 18,933 (49.3) 2419 (6.3) 2832 (7.4) 12,906 (33.6) 20,283 (52.7) 19,184 (56.1) 17,725 (46.0)
.001 ,.001 ,.001 ,.001 ,.001 ,.001 ,.001 ,.001 ,.001 ,.001 .056 ,.001
Gender (male) Diabetes (yes) Previous MI (yes) Previous stroke (yes) AF (yes) Hypertension (yes) Stroke type (hemorrhage) Alcohol consumption (over limit) Smoking (current) Civil status (living with someone) SSS , 50 Age , 71 y
Abbreviations: AF, atrial fibrillation; BMI, body mass index; MI, myocardial infarction; SSS, Scandinavian Stroke Scale.
P , .001). Patients with previous strokes were more likely to have missing information on BMI (OR 5 1.13; P ,.001). Also smoking status (OR 5 1.10; P , .001), hypertension (OR 5 1.13; P , .001), and civil status (OR 5 1.24; P , .001) were associated with missing BMI. Gender, age, and other variables included in our analysis did not differ significantly between patients in whom BMI was available and unavailable. There were 28,382 patients with complete covariate information. Risk factor distribution, age, sex, stroke severity, and civil status according to the BMI category in these patients are shown in Table 2. Prevalence of diabetes, previous MI, and hypertension increased with increasing BMI, whereas prevalence of smoking, AF, previous stroke, and hemorrhagic stroke increased with decreasing BMI. More severe stroke and higher age were associated with lower BMI. Of the 28,382 patients, stroke was first-ever in 22,811 (80.4%; mean age, 69.4 years [SD 13.3 years]; mean SSS
score, 45.6 [SD 13.6]) and in 5571 patients (19.6%) stroke was recurrent (mean age, 72.4 years [SD 11.4 years]; mean SSS score, 43.6 [SD 13.4]). P value is less than .001 for age and the SSS score. Table 3 shows sex, BMI, other cardiovascular risk factors, and civil status in patients with first-ever stroke and those in whom stroke was recurrent. Patients with recurrent stroke were significantly more often men, and they had significantly more often diabetes, previous MI, AF, and hypertension. They were significantly more often living alone, and they were significantly more often underweight and less often overweight, whereas prevalence of obesity did not differ significantly between patients with recurrent and firstever stroke. Figure 1 shows the univariate association between BMI and risk of stroke being recurrent; the risk decreases as BMI increases. The independent association between BMI and risk of stroke being recurrent in a multiple logistic regression analysis adjusting for age, stroke severity
Table 2. Risk factor distribution according to BMI category
Gender (male) Diabetes (yes) Previous MI (yes) Previous stroke (yes) AF (yes) Hypertension (yes) Stroke type (hemorrhage) Alcohol consumption (over limit) Smoking (current) Civil status (living with someone) SSS , 50 Age , 71 y
Underweight (%)
Normal weight (%)
Overweight (%)
Obese (%)
P value
334 (25.7) 66 (5.1) 96 (7.4) 293 (22.5) 221 (17.0) 534 (41.0) 117 (9.0) 113 (8.7) 718 (55.1) 479 (36.8) 785 (68.9) 424 (32.6)
6144 (49.6) 1155 (9.3) 1019 (8.2) 2480 (20.0) 1857 (15.0) 5758 (46.4) 719 (5.8) 1057 (8.5) 5494 (44.3) 6532 (52.7) 6084 (54.6) 5462 (44.1)
6320 (64) 1401 (14.2) 933 (9.5) 1870 (18.9) 1345 (13.6) 5284 (53.5) 487 (4.9) 846 (8.6) 3661 (37.1) 6138 (62.2) 4405 (48.7) 5217 (52.9)
2708 (56.3) 1108 (23.0) 479 (10.0) 928 (19.3) 591 (12.3) 2946 (61.2) 228 (4.7) 390 (8.1) 1686 (35.0) 3024 (62.8) 2132 (48.4) 3096 (64.3)
,.001 ,.001 ,.001 .011 ,.001 ,.001 ,.001 .782 ,.001 ,.001 ,.001 ,.001
Abbreviations: AF, atrial fibrillation; BMI, body mass index; MI, myocardial infarction; SSS, Scandinavian Stroke Scale. The table is based on completed cases.
OVERWEIGHT, OBESITY, AND STROKE
e579
Table 3. Gender, BMI, cardiovascular risk factors, and civil status in patients with first-ever stroke and patients in whom stroke was recurrent
Gender (male) Diabetes (yes) Previous MI (yes) AF (yes) Hypertension Stroke type (infarction) Alcohol consumption (over limit) Smoking (current) Civil status (living with someone) Underweight Normal weight Overweight Obese
All, n (%)
First stroke, n (%)
Recurrent stroke, n (%)
P
15,506 (54.6) 3730 (13.1) 2527 (8.9) 4014 (14.1) 14,522 (51.0) 26,831 (94.5) 2406 (8.5) 11,559 (40.1) 16,173 (57.0) 1302 (4.6) 12,399 (43.7) 9869 (34.8) 4812 (17.0)
12,307 (53.9) 2764 (12.1) 1810 (7.9) 3022 (13.2) 11,044 (48.4) 21,549 (94.4) 1968 (8.6) 9415 (33.2) 13,167 (57.7) 1009 (4.4) 9919 (43.5) 7999 (35.1) 3884 (17.0)
3199 (57.4) 966 (17.3) 717 (12.9) 992 (17.8) 3478 (62.4) 5282 (94.8) 438 (7.9) 2144 (38.5) 3006 (54.0) 293 (5.2) 2480 (44.5) 1870 (33.6) 928 (16.7)
,.001 ,.001 ,.001 ,.001 ,.001 .33 .07 ,.001 ,.001 .03 Reference .05 .29
Abbreviations: AF, atrial fibrillation; BMI, body mass index; MI, myocardial infarction.
(SSS), sex, BMI (4 groups), civil status, and cardiovascular risk factors is shown in Table 4. It appears that being obese and overweight (using normal weight as reference) is associated with a significantly lower risk of stroke being recurrent (obese: OR 5 .90; CI .82-.98; overweight: OR 5 .89; CI .83-.96). Being underweight is associated with a significantly higher risk of stroke being recurrent (OR 5 1.23; CI 1.06-1.43). Diabetes (OR 5 1.31; CI 1.20-1.43), hypertension (OR 5 1.68; CI 1.57-1.79), previous MI (OR 5 1.35; CI 1.23-1.49), AF (OR 5 1.19; CI 1.09-1.30), and being men (OR 5 1.29; CI 1.20-1.38) were associated with a significantly higher risk of stroke being recurrent, whereas living with someone was associated with a significantly
lower risk of stroke being recurrent (OR 5 .93; CI .87-1.00). The partial effect of age and stroke severity (SSS) on BMI is shown in Figure 2. From the age of 60 years and older, BMI was decreasing steeply with age, about 1 point BMI per 10 years increase. Milder strokes tended to have higher BMI.
Discussion Among patients admitted with stroke, the risk that stroke was recurrent was inversely related to BMI: previous stroke was less often reported in overweight and obese stroke patients than in normal weight patients, no Table 4. Multivariate estimates on the risk of stroke being recurrent
Gender (male) Diabetes (yes) Previous MI (yes) AF (yes) Hypertension Stroke type (infarction) Alcohol consumption (over limit) Smoking (current) Civil status (living with someone) Underweight Normal weight Overweight Obese
Figure 1. Univariate association between BMI and the risk of stroke being recurrent. The confidence intervals are 95% limits of the fit. Abbreviations: BMI, body mass index; OR, odds ratio.
OR
95% CI
P
1.29 1.31 1.35 1.19 1.68 1.10 .98
1.20-1.38 1.20-1.43 1.23-1.49 1.09-1.30 1.57-1.79 .96-1.27 .88-1.12
0 0 0 0 0 .16 .86
1.07 .93
.98-1.15 .87-1.00
.12 .04
1.23 Reference .89 .90
1.06-1.42
.01
.83-.96 .82-.98
0 .02
Abbreviations: AF, atrial fibrillation; CI, confidence interval; MI, myocardial infarction; OR, odds ratio. Multiple logistic regression analysis adjusting for age, sex, stroke severity, stroke type, civil status, and cardiovascular risk factors.
K.K. ANDERSEN AND T.S. OLSEN
27
27
26
26 BMI
BMI
e580
25 24
24
23
23
20
40
60 Age
80
100
Figure 2. Association between age and BMI (left) and the SSS score and BMI (right). The confidence intervals are 95% limits of the fit. Abbreviations: BMI, body mass index; SSS, Scandinavian Stroke Scale.
25
0
10
matter age, sex, stroke severity, and cardiovascular risk factor profile. Conversely, underweight stroke patients had more often a history of a previous stroke. There is accumulating evidence that overweight and obesity are not associated with increased risk of stroke recurrence.2-4 Recently, data from the Danish Stroke Registry indicated that the risk of readmission for recurrent stroke was even lower in obese than in normal weight stroke patients.5 These findings are further supported by the findings of the present study indicating that obesity is not only associated with decreased risk of recurrent stroke in the years after stroke. In patients admitted to the hospital with acute stroke, a history of a previous stroke is also less frequent in obese and overweight patients than in normal weight and underweight stroke patients. Our study now adds to the evidence of another dimension of the obesity paradox1 that obesity although associated with increased risk of stroke in primary prevention studies16,17 is not associated with higher risk of stroke recurrence in secondary prevention studies. There is still no explanation of this new aspect of the obesity paradox in stroke, also observed in prospective studies.2-5 Higher rate of antihypertensive use in obese patients has been suggested as a potential explanation,3 but in our study, the lower risk of stroke being recurrent was independent of hypertension and other cardiovascular risk factors. Another suggestion was metabolic consequences of skeletal muscle tissue wasting because of the incident stroke,4 but this does not explain lower prevalence of previous stroke seen in this study. Surprisingly, positive health indicators such as nonsmoking and nondrinking are more common among obese and overweight stroke patients.5 Also, cohabitation is more common among obese and overweight stroke patients.5 Hence, lifestyle or social factors hitherto not addressed when evaluating health effects of increased BMI may counteract adverse effects of DM and hypertension occurring more often in obese and overweight stroke patients. However, the findings in our study were independent of hypertension, diabetes, and civil status.
20
30 40 SSS score
50
60
Prevalence of traditional cardiovascular risk factors for stroke was significantly associated with BMI. Hence, prevalence of diabetes, hypertension, and previous MI increased as BMI increased, whereas prevalence of AF and smoking decreased as BMI increased. The remarkably higher prevalence of current smoking in addition to higher age and higher prevalence of AF might explain higher prevalence of recurrent stroke among underweight stroke patients. However, the higher prevalence remained even after adjustment for these variables. In accordance with already established knowledge on risk factors for stroke,15 we found that diabetes, previous MI, AF, and hypertension were associated with increased frequency of previous stroke in our multivariate analysis also when adjusting for BMI. Our study has weaknesses and strengths. Our study’s strength is first of all in its large sample size allowing for sufficient statistical power. Second, we included patients without limitations on age (.18 years), gender, or stroke severity. Third, in our patients, stroke severity was measured on admission to the hospital using a wellvalidated neurologic scale. Although the Danish Stroke Registry is designed as a nationwide registration of all patients admitted with acute stroke, coverage is not yet complete (presently about 80%).11 It is a weakness of concern that of the 61,872 patients registered in the Danish Stroke Registry, only 62% had information on BMI; however, for all other variables, data completeness is high exceeding 85% for all individual variables. The relatively high rate of missing data on BMI when compared with other variables in stroke databases is a well-known problem,4 reflecting that the clinical usefulness of this information still has not achieved general acceptance. However, we performed an analysis of the missing observations and did not find any evidence of informative missingness and, thus, biased estimates. Except for higher prevalence of hemorrhagic stroke and severe strokes, differences between patients with or without measurement of BMI in regard to age, sex, and risk factors were of no clinical importance. Thus, for the purpose of establishing a model of high
OVERWEIGHT, OBESITY, AND STROKE
statistical validity, we believe that our data set is appropriate as the large sample size allows us to adjust reliably for differences in age, sex, civil status, stroke severity, and risk factor profile. We cannot exclude, however, the possibility of bias because of variables not recorded in the Danish Stroke Registry. We have no information as to treatments or interventions that might have influenced stroke recurrence, and we have no information on possible differences between BMI groups in regard to preventive treatment and compliance to treatment. In conclusion, it is generally accepted to recommend patients with stroke to aim at normal weight, that is, BMI between 18.5 and 25.16 There is no doubt that overweight and obesity are associated with increased risk of a number of diseases, among these also stroke.17,18 However, our study supports findings from previous research that obesity is not associated with increased risk of recurrent stroke in stroke patients.2-5 These findings are still unexplained calling for further studies as more than half of stroke patients are overweight or obese, and the effect of weight reduction on prognosis in stroke is still unknown.
References 1. McAuley PA, Blair SN. Obesity paradoxes. J Sports Sci 2011;29:773-782. 2. Vemmos K, Ntaios G, Spengos K, et al. Association between obesity and mortality after acute first-ever stroke: the obesity–stroke paradox. Stroke 2011;42:30-36. 3. Ovbiagele B, Bath PM, Cotton D, et al. Obesity and recurrent vascular event after recent ischemic stroke. Stroke 2011;42:3397-3402. 4. Doehner W, Schenkel J, Anker SD, et al. Overweight and obesity are associated with improved survival, functional outcome, and stroke recurrence after acute stroke or transient ischaemic attack: observations from the TEMPiS trial. Eur Heart J 2013;34:268-277. 5. Andersen KK, Olsen TS. The obesity paradox in stroke: lower mortality and lower risk of readmission for recurrent stroke in obese stroke patients. Int J Stroke
e581
6.
7. 8. 9.
10.
11.
12.
13.
14.
15. 16.
17.
18.
2013; March 12. http://dx.doi.org/10.1111/ijs.12016. [Epub ahead of print]. Olsen TS, Dehlendorff C, Petersen HG, et al. Body mass index and poststroke mortality. Neuroepidemiology 2008;30:93-100. Bray AB. The obesity paradox—an artifact of small sample size? Nat Rev Cardiol 2009;6:561-562. Gielen S, Sandri M. The obesity paradox—a scientific artifact? Int J Cardiol 2013;162:140-142. Olsen TS, Dehlendorff C, Andersen KK. Sex-related timedependent variations in poststroke survival—evidence of a female stroke survival advantage. Neuroepidemiology 2007;29:218-225. Mainz J, Krog BR, Bjønshave B, Bartels P. Nationwide continuous quality improvement using clinical indicators: the Danish National Indicator Project. Int J Qual Health Care 2004;16(Suppl 1):145-150. Lindenstrøm E, Boysen G, Christiansen LW, et al. Reliability of Scandinavian Stroke Scale. Cerebrovasc Dis 1991;1:103-107. Report of the WHO Task Force on stroke and other cerebrovascular disorders: stroke—1989: recommendations on stroke prevention, diagnosis, and therapy. Stroke 1989;20:1407-1431. NIP-apopleksi. National auditrapport 2010. Version 3; Aarhus, Denmark, 2011. Available at: https://www.sundhed. dk/content/cms/69/4669_2010_apopleksi_nat_sfa.pdf. Accessed March 20, 2013. Jørgensen HS, Plesner AM, H€ ubbe P, et al. Marked increase of stroke incidence in men between 1972 and 1990 in Frederiksberg, Denmark. Stroke 1992;23: 1701-1704. Ihaka I, Gentlemann RR. A language for data analysis and graphics. J Comput Graph Stat 1995;5:299-314. Sacco RL, Adams R, Albers G, et al. Guidelines for prevention of stroke in patients with ischemic stroke or transient ischemic attack: a statement for healthcare professionals from the American Heart Association/American Stroke Association Council on Stroke. Stroke 2006; 37:577-617. Kurth T, Gaziano JM, Berger K, et al. Prospective study of body mass index and risk of stroke in apparently healthy women. Circulation 2005;111:1992-1998. Mann GV. The influence of obesity on health (second of two parts). N Engl J Med 1974;291:226-232.