View
215
Download
0
Category
Preview:
Citation preview
A Prospective Study of Cerebrovascular Disease inJapanese Rural Communities, Akabane and Asahi
Part 1: Evaluation of Risk Factors in the Occurrence of CerebralHemorrhage and Thrombosis
HIROSHI OKADA, M.D., M.P.H., HIROSHI HORIBE, M.D., YOSHIYUKI OHNO, M.D., M.P.H..
NORIHIKO HAYAKAWA, M.D., AND NOBUO AOKI, M.D.
599
SUMMARY An epidemiological study of cerebrovascular diseasein Akabane and Asahi, Japan, was made. (These cities are locatednear Nagoya, Japan.) The study population included 4,737 men andwomen aged 40 to 79 at the time of entry into the study. There were4,186 persons who were examined and, of these, 264 cases of cerebro-vascular attacks were observed between 1964 and 1970. The incidencerate of stroke in those persons not responding to the survey was 15.9times higher than in those persons examined according to person-yearobservation in Akabane. The risk factors for cerebral hemorrhageand thrombosis were evaluated by age-adjusted and sex-adjustedrelative risks. The predisposing factors to cerebral hemorrhageappeared to be high blood pressure, high left R wave, ST depression,
T abnormality, capillary fragility counts, previous medical history ofstroke, and albuminuria. For cerebral thrombosis, the predisposingfactors appeared to be high blood pressure, ST depression and fun-duscopic sclerotic findings, and those factors assumed to be signifi-cant were glycosuria and smoking habits. Ocular funduscopic abnor-mality was the most prominent risk factor for cerebral thrombosis,while high blood pressure and ECG abnormalities were highly relatedto cerebral hemorrhage. It was suggested that those subjects with arelatively higher blood pressure may have a higher relative risk ofcerebral hemorrhage than those with a lower (normal range) bloodpressure. A previous or family history of stroke also appearedsignificantly related to cerebral hemorrhage.
ESPECIALLY in Japan, cerebrovascular disease may havesome etiological features different from those of coronaryheart disease. Risk factors for coronary heart disease havebeen evaluated and those for cerebrovascular disease havebeen investigated only recently.1"8 The incidence ratio ofcerebral hemorrhage to thrombosis is still controversial inJapan,210 while it is consistently less than 1.0 in West Euro-pean countries and the United States.11
The incidence of cerebrovascular disease has been highestin the northern part of the Japanese mainland.12 Lowtemperatures associated with housing inadequacy and ex-cessive salt intake have been suspected as major factors inthe high morbidity and mortality rates.1316 Recently, it wasstated that unbalanced nutrition might be a risk factor forcerebral hemorrhage.3-' The incidence of cerebralhemorrhage has been decreasing steadily, possibly due to thecontinuous medical care (supported by a nationwide healthinsurance program) of hypertension in Japan.12
The study started in 1963 in order to investigate the in-cidence and prevalence of cerebrovascular diseases and toevaluate risk factors.8 An epidemiological follow-up studyhas been conducted in two rural Japanese communities,Akabane and Asahi (near Nagoya), located in the centralpart of the Japanese mainland.
This is a preliminary report describing the two com-munities, the survey method and the incidence of cerebro-vascular diseases in the two populations during a seven-yearperiod. The risk factors (medical history, physical signs andlaboratory values) for cerebral hemorrhage and thrombosisare assessed.
From the Department of Preventive Medicine, Nagoya University Schoolof Medicine, Tsurumai-cho 65, Showa-ku, Nagoya 466, Japan.
Supported in part by a grant from the Ministry of Education, JapaneseGovernment.
Reprint requests to Dr Hiroshi Horibe.
Methods
Akabane is located along the southern coast of the At-sumi peninsula south of Nagoya. The major industry isagriculture. Asahi is located in the low mountain area about70 km northeast of Nagoya. The principal industries areagriculture and forestry. The majority of the residents areonly part-time farmers, and they work at other occupations(fig. 1). Their economic level of life is steadily improving,but is still poorer than the urban area.
The total 1965 census of Akabane and Asahi was 6,696and 6,482 persons, respectively, gradually decreasing by1970 to 6,400 and 5,753, respectively. Akabane has threegeneral practitioners and a regional health center, AtsumiHospital. There are two general practitioners in Asahi andthe Asuke Hospital is the regional health center. Themedical personnel of both communities were verycooperative in this follow-up study by immediately reportingany new or suspected cases of cerebrovascular disease or,when periodically asked, by thoroughly listing their cases ofcerebrovascular disease. Death certificates also werestudied.
The general practitioners usually made differentialdiagnoses of cerebrovascular diseases based on specifieddiagnostic criteria, which were a modification of theclassification reported by Millikan et al ." When thediagnosis was uncertain, we tried to contact the generalpractitioner to discuss the case and to determine thediagnosis. Autopsied cases of cerebrovascular disease wererare, even in the regional key hospitals, due to traditional orreligious custom.
The age of the subjects was 40 or more at entry to thestudy in 1964. They were examined periodically by the studyphysicians along with paramedical personnel andnutritionists. Their symptoms and medical histories ofstroke, hypertension, cardiac and renal diseases, anddiabetes were checked and recorded by trained laymen
by guest on Novem
ber 11, 2017http://stroke.ahajournals.org/
Dow
nloaded from
600 STROKE VOL. 7, No. 6, NOVEMBER-DECEMBER 1976
PACIFIC OCEAN
FIGURE 1. Location of Akabane and Asahi, Japan.
against a self-administered questionnaire, which had beenmailed to the subjects prior to examination. The question-naire covered such topics as daily activities (casual orrestrained work) and smoking and drinking (occasional orregular) habits. "Casual work" meant sometimes and"restrained work" referred to the usual work-week withlimited working hours and/or physical activity. The oc-casional drinker did not have alcoholic beverages every daywhereas the regular drinker did.
The physical and laboratory examinations includedanthropometric measurements of height, weight and skin-folds on R-triceps and subscapular sites, neck cir-cumference, casual BP, resting and exercise (Master's two-step test) ECGs, ocular funduscopy, urinalysis by test tapeof protein and sugar, serum cholesterol determinations, andcapillary fragility test.
Funduscopic color pictures were taken with a Mamiyafundus camera. These were identified by simultaneous re-
cording of the date and name and identification number ofthe subject appearing side by side. The pupil of the right eyewas dilated by applying eye drops of tropicamide andphenylephrine hydrochloride. After extensive visual obser-vation, at least three different parts of the retina werephotographed. The funduscopic picture was enlarged anddiagnosed by the study physicians by using a viewer called"Medical Vision," which was based on a modification of theclassification cited by Aoki.8
All of the data were coded and punched on IBM cards andprocessed for analysis by an IBM sorting machine and aFACOM 230-60 electronic computer in the Nagoya Univer-sity Computation Center. The significance of each findingwas tested by X2 and t-test. Age-adjusted and sex-adjustedrelative risks for each category were calculated andstatistically tested against the overall incidence rate ofcerebral hemorrhage and thrombosis and then a ratio ofrelative risk for a specified category was calculated in orderto eliminate the error due to small numbers of observationsin a specific category.
Results
The cumulative response rate and the age and sex dis-tributions of the subjects are shown in table 1. The overallresponse rate was 90.3% in Akabane and 86.7% in Asahi.Generally, the response rates in younger age groups werelower than in older age groups, and those in men were con-sistently lower than in women.
As shown in table 2, cerebral hemorrhage occurred in 143of 264 patients diagnosed as having cerebrovascular diseasein Akabane and Asahi between 1964 and 1970, whilecerebral thrombosis occurred in 109 subjects during thesame time period. The incidence ratio of cerebralhemorrhage to thrombosis was 1.31. Subarachnoidhemorrhage (eight), cerebral embolism (one) and un-classified event (three) also were diagnosed in these two pop-ulations.
After the first examination, cerebral hemorrhage andthrombosis occurred in 71 and 58 patients (age: 40 to 79years), respectively, and these patients had further analysesto evaluate the significance of risk factors (table 3). In thefirst survey of persons 40 years of age or more with cerebro-vascular events, the incidence rate in the non-respondentswas much higher than in those who were examined; the ratioof incidence rates was 15.9 (table 4). Thus, the non-respondents were assumed to be the high-risk group forcerebrovascular disease.
TABLE 1 Cumulalive Response Rales by Age and Sex in Population Survey of Akabane and Asahi (1964 to 1968)
Men (age)40-49 50-59 60-69 70-79 Total
Women (age)40-49 50-59 60-69 70-79 Total
AkabaneRegisteredExaminedResponse rate (%)
AsahiRegisteredExaminedResponse rate (%)
33229388.3
35726072.8
31227186.9
32927683.9
24022995.4
30528593.4
12211594.3
17015490.6
1,035 368 399 370 160 1,197908 338 371 268 144 1,10787.7 91.9 93.0 99.3 90.0 92.5
1,161 457 379 326 182 1,344975 396 351 317 132 1,19684.0 86.7 92.6 97.2 72.5 89.0
Overall response rate: Akabane 90.3% and Asahi 86.7%.
by guest on Novem
ber 11, 2017http://stroke.ahajournals.org/
Dow
nloaded from
602 STROKE VOL. 7, No. 6, NOVEMBER-DECEMBER 1976
TABLE 6 Age-Adjusted and Sex-Adjusted Relative Risk of ECG and Funduscopic Findings
ECGNormalQ-QSHigh left R-waveST depressionT abnormality
FunduscopyHypertension
Grade: 0123-4
Not examinedSclerosis
Grade: 0123-4
Not examined
No. pts.
2,19451
869412229
1,7001,275
618115405
1,3241,973
367136405
Cerebral hemorrhageNo. cases RRa (ratio)
133
322113
162116
212
113275
12
0.42 (1.0)2.54 (3.3x)1.90 (3.2x)*2.56 (4.5x)*2.83 (5.1x)*
0.77 (1.0)0.97 (1.3x)1.26 (1.6x)0.87 (l.lx)1.37 (1.8x)
0.65 (1.0)1.04 (1.6x)0.98 (1.5x)1.84 (2.8x)1.37 (2.1x)
CerebralNo. cases
193
1712
4
1113174
13
32515
213
thrombosisRRa (ratio)
0.78 (1.0)2.75 (3.2x)1.14 (1.4x)1.90 (2.5x)f1.14 (1.4x)
0.69 (1.0)0.74 (l.lx)1.60 (2.3x)2.00 (2.9x)1.74 (2.5x)
0.23 (1.0)t0.99 (4.2x)2.53 (10.8x)*0.91 (3.9x)1.74 (7.5x)
•p <0.001, tp <0.05,
BP and ST-T abnormality had a relative risk for cerebralhemorrhage of 3.7, and a low risk (1.9) of cerebral throm-bosis. In those subjects with funduscopic abnormality andECG ST-T abnormality, the relative risk was 3.0 forcerebral thrombosis. The relative risk of cerebralhemorrhage in those subjects with high BP, abnormal ST-Tand funduscopic findings was not higher (2.7) than in thosewith only high BP and ST-T abnormality (risk: 3.7); whilefor cerebral thrombosis the risk was not significantly higher(2.7) in those who had high BP, ST-T abnormality and ab-normal funduscopic findings than in those having only ab-normal ST-T and funduscopic findings (3.0).
Anthropometric Measurements
Relative weight of those examined from Akabane andAsahi was less than the average figure in Japan as seen fromthe distribution cited in table 8. The first category of relativeweight had the lowest risk of cerebral hemorrhage and thesecond category of relative weight had the lowest risk ofcerebral thrombosis, but the difference from the othercategories was not statistically significant.
We found that the lowest incidence rate of cerebralhemorrhage and thrombosis was not necessarily in the firstcategory of skinfolds. None of the differences in relativerisks reached the statistically significant level because theskinfold measurement was limited to subjects from Asahi.
The third category of triceps skinfold (3.0 to 4.9 mm) hadthe lowest relative risks for both cerebral hemorrhage andthrombosis.
The second category (34 to 35 cm) of neck circumferencehad the lowest relative risk for cerebral hemorrhage, whilethe first category (< 34 cm) of neck circumference had thelowest relative risk for cerebral thrombosis. The larger theneck circumference, the higher were the relative risks ofcerebral hemorrhage and thrombosis, although therelationship did not reach a statistically significant level.
Urinalysis
Albuminuria was detected in 169 of 4,231 subjects, andthe positive cases showed a relative risk of 2.7 for cerebralhemorrhage with statistical significance (table 9). In thosepatients with a trace level of albuminuria the relative risksfor cerebral hemorrhage and thrombosis were 1.4 and 1.9,respectively.
Those persons with glycosuria had a 1.7 relative risk ofboth cerebral hemorrhage and thrombosis.
Serum Cholesterol Level and Capillary Fragility Test
Age-adjusted and sex-adjusted relative risk of cerebralhemorrhage by serum cholesterol levels was highest (1.8) inthe first category (< 130 mg/dl) with statistical significancein the second category (130 to 159 mg/dl) which had the
TABLE 7 Age-Adjusled and Sex-Adjusted Relative Risk According to BP and ECG and Ftmduscopic Findings
Factors
NoneHigh BP (A)ST-T abnormality (B)Abnormal fundus (C)A + BA+ CB + CA + B + C
No. pts.
2,3051,137
4831,032
239369155100
Cerebral hemorrhageNo. cases RRa (ratio)
105824212220
77
0.33 (1.0)*2.16 (6.5x)*2.47 (7.4x)*1.03 (3.1x)3.68 (ll.Ox)*2.25 (6.7x)*1.87 (5.6x)2.73 (8.2x)J
CerebralNo. cases
10371226
915
96
thrombosisRRa (ratio)
0.42 (1.0)t1.59 (3.8x)f1.64 (3.9x)1.53 (3.6)t1.85 (4.4x)1.97 (4.7x)J2.95 (7.0x)f2.73 (6.5x)t
High BP: 160/95 mm Hg or more; ST-T abnormality: Minnesota code IV 1-3 or V 1-3; abnormal fundus: H2-4 or S2-4 by Scheie'sclassification.
*P <0.001, tp <0.05.
by guest on Novem
ber 11, 2017http://stroke.ahajournals.org/
Dow
nloaded from
PROSPECTIVE STUDY OF CVD IN JAPAN. PART l/Okada el al. 603
TABLE 8 Age-Adjusted and Sex-Adjusted Relative Risk by Anthropometric Measurements
Relative weight (%)- 1 5
-14 ~ -5-4 H4+ 5 ~
Skinfold (mm)Triceps
-1.92.0-2.93.0-4.95.0
Subscapular- 7.9
8.0-11.912.0-
Umbilical-4.9
5.0-9.910.0-
Neck circumference (cm)-33
34-3536-3738-
No. pts.
1,4491,398
756569
316642471349
566681549
354725697
904475262156
Cerebral hemorrhageNo. cases RRa (ratio)
23251310
4845
886
6106
9564
0.89 (1.0)1.11 (1.3x)1.12 (1.3x)1.20 (1.4x)
0.91 (1.2x)1.03 (1.4x)0.73 (1.0)1.50 (2.1x)
0.98 (1.0)1.03 (l.lx)1.10 (l.lx)
1.18 (1.3x)1.07 (1.2x)0.91 (1.0)
0.84 (l.lx)0.75 (1.0)1.71 (2.3x)2.05 (2.7x)
CerebralNo. cases
2018127
6732
983
497
6744
thrombosisRRa (ratio)
1.02 (l.lx)0.92 (1.0)1.17 (1.3x)1.06 (1.2x)
1.34 (1.7x)0.88 (l.lx)0.79 (1.0)1.25 (1.6x)
1.23 (1.6x)1.14 (1.5x)0.77 (1.0)
0.78 (1.0)0.96 (1.2x)1.28 (1.7x)
0.80 (1.0)1.01 (1.3x)0.97 (1.2x)1.80 (2.3x)
lowest (0.6) relative risk. Relative risks of cerebral hemor-rhage showed a slight increase from the second (risk = 0.56)to fifth (risk = 1.08) category, but the relative risks of cere-bral thrombosis showed a slight increase and then decreasefrom the second (risk = 1.0 to 1.2) to the fifth (risk = 0.6)category (table 10).
Relative risks of cerebral hemorrhage and cerebralthrombosis by capillary fragility counts tended to increasewith the advancing categories but never reached statisticalsignificance.
Daily Activity and Smoking and Drinking Habits
Subjects with no or casual daily activity had a relative riskof 2.5 for cerebral hemorrhage with statistical significance(table 11). It was quite impressive that the relative risks ofcerebral hemorrhage and thrombosis were higher in subjectsworking casually than in those not working.
Relative risks of cerebral hemorrhage and thrombosis innon-smokers was lower than in smokers or ex-smokers, butthe difference was not statistically significant. Habitualdrinkers had the highest relative risk for cerebral throm-bosis, and occasional drinkers the lowest incidence rate ofboth cerebral hemorrhage and thrombosis.
A reverse relationship was found between the relative
risks of cerebral hemorrhage and thrombosis and food in-take, i.e., those persons ingesting 20 bowls of rice or moreper day had the lowest relative risk of cerebral hemorrhagebut the highest relative risk of cerebral thrombosis (table 11).
It was understandable that patients who previouslyreceived or who regularly used antihypertensive agentsshowed a higher relative risk (2.3) of cerebral hemorrhage(statistically significant) than thrombosis; however, therewas no difference between the relative risks of cerebralthrombosis in patients who either previously received orregularly used antihypertensive agents.
Medical History
Subjects previously having a stroke had a very highrelative risk of cerebral hemorrhage (5.4) and thrombosis(2.7) (table 12). Those persons with a previous history ofhypertension had relative risks for cerebral hemorrhage at2.2 and for cerebral thrombosis at 1.6, results comparablewith those of patients taking antihypertensive agents. On theother hand, patients with a family history of stroke or hyper-tension showed a significantly high relative risk of cerebralhemorrhage rather than thrombosis. The incidence rate inpatients with previous or family history of heart disease wasless but not significantly different.
TABLE 9 Age-Adjusted and Sex-Adjusted Relative Risk of Albuminuria and Glycosuria
Albuminuria
±
+ ~ + + +Glycosuria
+— = negative, ±*p <0.01.
= trace, + r
No. pts.
3,743320169
4,097133
—1- + + = positive.
Cerebral hemorrhageNo.
529
10
665
cases RRa (ratio)
0.88 (1.0)1.38 (1.6x)2.70 (3.1x)*
1.00 (1.0)1.67 (1.7x)
CerebralNo. cases
42104
505
thrombosisRRa (ratio)
0.88 (1.0)1.91 (2.2x)1.14 (1.3x)
0.93 (1.0)1.66 (1.8x)
by guest on Novem
ber 11, 2017http://stroke.ahajournals.org/
Dow
nloaded from
604 STROKE VOL. 7, No. 6, NOVEMBER-DECEMBER 1976
TABLE 10 Age-Adjusted and Sex-Adjusted Relative Risk bySerum Cholesterol Levels and Capillary Fragility Counts
Cerebral hemorrhage Cerebral thrombosisNo.pts.
No.cases
Serum cholesterol (mg/dl)-129
130-159160-189190-219220-
370878
1,159936622
108
141511
Capillary fragility counts- 3
4-1112-1920-
661410273566
464
12
RRa (ratio)
1.75 (3.1x)0.56 (1.0)0.76 (1.4x)0.97 (1.7x)1.08 (1.9x)
0.47 (1.0)*1.18 (2.5x)1.11 (2.4x)1.63 (3.5x)
No.cases
3131714
4
6548
RRa (ratio)
0.58 (l.lx)1.03 (1.8x)1.09 (2.0x)1.22 (2.2x)0.56 (1.0)
0.74 (1.0)1.16 (1.6x)1.25 (1.7x)1.42 (1.9x)
*p <0.05.
Subjective Symptoms
The subjects with any one of 26 subjective symptoms(table 13) had nonsignificant age-adjusted and sex-adjustedrelative risks for cerebral hemorrhage and thrombosis.More than four cases of cerebral hemorrhage with subjectivesymptoms (speech disturbance, fainting, insomnia, stiff gait)had a relative risk more than 2.0, while none of these oc-curred in cerebral thrombosis.
Discussion
Some investigators in the United States and Europe doubtthe diagnostic accuracy of cerebrovascular diseases inJapan. The Classification of Cerebrovascular Diseases16 hasbeen familiar to Japanese general practitioners as well as in-vestigators of cerebrovascular diseases, but this has notaltered the general practitioner's habits of differentialdiagnosis. Hirota and Lau" reported that clinical diagnosisof cerebral hemorrhage was correct in 70% of the autopsied
cases and that the accuracy rate was unchanged during thelast ten years.18 They pointed out that in Japan the accuracyrate for diagnosing cerebral infarction and subarachnoidhemorrhage was less than that of cerebral hemorrhage butdid improve during the last ten years.
The incidence ratio of cerebral hemorrhage to thrombosisin these communities (Akabane and Asahi) was slightlyhigher than that reported in other epidemiological studies ofJapanese populations,2"7 but was lower than that in nationaldeath statistics.12 Some investigators reported that cerebralhemorrhage was overdiagnosed as compared to throm-bosis,17- 19 while others3 considered the high ratiocharacteristic of strokes in Japan.
High blood pressure is the most predominant risk factorof stroke, especially hemorrhage,2' '•13> 2° but there havebeen many various opinions regarding borderline hyper-tension.21 Relative risk of borderline systolic hypertension(i.e., 140 through 159 mm Hg) was three times as high as thelowest figure for cerebral hemorrhage and double the figurefor thrombosis. Although the difference in incidence rateswas statistically tested to the overall rate (because a fewcases were found in the lowest category of systolic BP), itwas concluded that borderline systolic hypertension couldgive an additional risk of cerebral hemorrhage or throm-bosis. This fact does not directly justify the lifelong phar-macological therapy for borderline hypertensives, but atleast suggests that their BP should be periodically checkedand that they should control every aspect of their daily life.
Furthermore, the subjects with systolic pressures of 120through 139 mm Hg showed a higher relative risk ofcerebral thrombosis than those with a BP of 119 mm Hg orless; the risks are similar for diastolic pressures and cerebralhemorrhage (even when pressures are within normal ranges).
It was noteworthy that in subjects with systolic pressuresof 160 through 179 mm Hg the relative risk of cerebralhemorrhage was much higher than for thrombosis. Forpressures of 180 mm Hg or more the relative risk for
TABLE 11 Age-Adjusted and Sex-Adjusted Relative Risk by Daily Activity and Habits and Medications
Work activityNoneCasualRestrainedRegular
Smoking (cigarettes/day)None
1- 910-192 0 -
DrinkingNoneOccasionalRegular
Rice intake (bowls/day)- 5
6-1112-192 0 -
Antihypertensive agentsNoneRegularPrevious
No. pts.
203168441
2,837
2,420323621414
2,4271,070
295
7401,916
869206
3,104351376
Cerebral hemorrhageNo. cai
9109
31
331316
7
43198
1336112
231819
ses RRa (ratio)
1.74 (2.3x)2.48 (3.2x)*0.98 (1.3x)0.77 (1.0)
0.90 (1.0)1.75 (1.9x)1.32 (1.5x)1.07 (1.2x)
1.03 (l.lx)0.93 (1.0)1.28 (1.4x)
1.00 (1.4x)1.11 (1.6x)0.91 (1.3x)0.70 (1.0)
0.51 (1.0)2.21 (4.4x)J2.29 (4.5x)J
CerebralNo. cases
46
1224
15121510
25179
1025
45
291011
thrombosisRRa (ratio)
0.98 (1.2x)1.69 (2.3x)1.68 (2.3x)0.74 (1.0)
0.61 (1.0)1.67 (2.7x)1.18 (l-9x)1.47 (2.4x)
0.95 (l.Ox)0.93 (1.0)1.49 (1.6x)
1.04 (2.9x)0.98 (2.7x)0.36 (1.0)t1.75 (4.8x)
0.78 (1.0)1.45 (1.9x)1.63 (2.1x)
.01. tp <0.05, Jp <0.001.
by guest on Novem
ber 11, 2017http://stroke.ahajournals.org/
Dow
nloaded from
PROSPECTIVE STUDY OF CVD IN JAPAN. PART \/Okada et al. 605
TABLE 12 Age-Adjusted and Sex-Adjusted Relative Risk by Medical History
Previous historyHealthyStrokeHypertensionHeart disease
Family historyStrokeHypertensionHeart disease
No. pts.
2,12045
635477
1,034224360
Cerebral hemorrhageNo. cases RRa (ratio)
176
296
2555
0.61 (1.0)5.35 (8.8x)*2.24 (3.7x)*0.73 (l.2x)
1.44f1.770.93
CerebralNo. cases
193
176
1046
thrombosisRRa (ratio)
0.85 (1.0)2.71 (3.2x)1.60 (1.9x)0.97 (1.2x)
0.711.801.41
*p <0.001, tp <0.05.
cerebral hemorrhage rose remarkably (from 9.8 to 20.9)while decreasing for cerebral thrombosis (from 7.3 to 5.4).Some epidemiological studies of cerebrovascular and car-diovascular diseases suggest that the relative risk of cerebralhemorrhage and thrombosis would be high in subjects withthe highest diastolic BP. This controversial statement, in ouropinion, needs further investigation for clarification.
The incidence rates of cerebral hemorrhage and throm-bosis were lowest in those with diastolic pressures of 69 mmHg or less. The relative risks increased sharply with the in-crease in diastolic pressure. It is noteworthy that diastolicblood pressures of 70 through 89 mm Hg gave additionalrisk of stroke compared with much lower diastolic pressures,although both categories have been assumed to be in the nor-mal range.
Most stroke epidemiological studies indicate that ECGabnormalities were very important risk factors not only forischemic heart disease but also for stroke.3"7 While therelative risk of cerebral hemorrhage was higher thancerebral thrombosis in subjects with ECG abnormalities, the
TABLE 13 Age-Adjusted and Sex-Adjusted Relative Risk ofSubjects With Subjective Symptoms
Symptoms
HeadacheDull headacheVertigoTinnitusWeak memorySpeech disturbanceFaintingFastidiousMore sentimentalInsomniaNumbnessStiff shoulderTremorSluggishFlushFeverishStiff gaitPalpitationChest oppressionShortness of breathExertional dyspneaEdemaNosebleedJoint painJoint swellingConstipation
No. pte.
761818573629
1,258113137343502252530
1,469184568225
92235845372555495549
35332
6013
CerebralhemorrhageNo.cases
1217111229
446989
273
1332
11181013
511
2423
RRa
1.011.251.151.071.282.022.141.121.022.071.131.130.921.150.861.622.071.321.651.280.661.162.370.842.008.00
CerebralthrombosisNo.cases
121149
23315656
1349124
1078671321
RRa
1.341.050.560.991.251.580.711.190.971.730.990.701.460.940.412.410.850.991.511.021.020.951.310.833.112.74
risk of cerebral thrombosis was remarkably higher than forcerebral hemorrhage in patients with abnormal funduscopicfindings.
The low incidence rate of cerebral thrombosis in thosesubjects without sclerotic findings (by Scheie's classification)was quite impressive, and this accounted for the higher riskratio of the other categories, including the risk ratio to thecontraindicated group for funduscopic examination.
Aoki reported the significance of further precise fun-duscopic findings in cerebral hemorrhage and thrombosis,based on similar material. Although several Japanese in-vestigators have stressed the significance of funduscopicfindings in predicting the occurrence of cerebrovascular dis-eases,5- 22- 23 there still is some hesitation in applying theocular funduscopic examination for screening purposes,even in Japan. Glaucoma patients (high-risk group) shouldbe excluded from the ocular funduscopic examination. Thefundus camera is still expensive and inconvenient and re-quires a specialist, thus limiting the practical use of thisequipment.
It was quite impressive that most of our subjects had alower weight than the average value (obtained by theNational Nutrition Survey conducted in Japan) in personsof the same age, sex and height.
Several reports have suggested that excess body weightalone was not related to the severity of cerebral athero-sclerosis,3' *• "•2S and we confirmed that in this study relativeweight did not show any significant relationship with therelative risk of either cerebral hemorrhage or thrombosis.Although the leanest people had the lowest relative risk, thelowest relative risk according to triceps skinfold was foundin the third category (3.0 to 4.9 mm). In other words, thosewith thin triceps skinfold showed higher relative risk ofcerebral hemorrhage and thrombosis, and this was true ofsubscapular skinfold and cerebral thrombosis. Thin skin-fold could be considered a sign of poor unbalanced nutrition,which is an important contributing factor to the occurrenceof stroke, especially hemorrhage.3-'
Neck circumference has been considered a representativemeasure of constitutional predisposition to stroke (a popularidea in Japan), and it was suggested in this study that thelarger the neck circumference, the more frequently cerebralhemorrhage and thrombosis occurred.
Albuminuria was found in 169 cases during the first ex-amination, and showed a significantly higher relative risk ofcerebral hemorrhage. Several investigators have reportedthat albuminuria showed a modestly high risk of stroke.'- '•26
Albuminuria (common in these communities) might be dueto hypertension or arteriosclerotic renal dysfunction rather
by guest on Novem
ber 11, 2017http://stroke.ahajournals.org/
Dow
nloaded from
PROSPECTIVE STUDY OF CVD IN JAPAN. PART X/Okada el al. 601
TABLE 2 Cerebrovascular Attacks Observed in Akabane andAsahi (1964 to 1970)
TABLE 4 Incidence Rates of Cerebrovascular Diseases in theExamined Persons and in Non-Respondents (Akabane, 1964
Cerebral hemorrhageCerebral thrombosisSAHCerebral embolismUnclassified
Total
Akabane
6834203
107
Asahi
7575610
157
Total
143109
813
264
to 1970)
Observed person-yearNumber of strokesIncidence rate/100,000/yearRelative rate
Examined
36,6966625.7
1
Non-respondents
1,88854
408.615.9
SAH = subarachnoid hemorrhage.
Blood Pressure
Relative risk in those with systolic pressures of 180 mmHg or higher was 3.34 for cerebral hemorrhage comparedwith the overall incidence rate in the same age ranges, andthe ratio was 20.9 compared with those having systolic pres-sures between 120 and 139 mm Hg (table 5). The latterfigure was significant. It should be noted that the lowest in-cidence rate of cerebral hemorrhage was not necessarilyobserved in the lowest systolic pressure level, although thedifference from the second lowest was not statisticallysignificant.
The highest relative risk for cerebral thrombosis wasobserved in those persons with the second highest systolicblood pressure (160 to 179 mm Hg), and the lowest risk wasin the lowest systolic BP (< 120 mm Hg).
The highest relative risk for cerebral hemorrhage wasobserved in those persons with the highest diastolic BP(>110 mm Hg). The lowest relative risk for both cerebralhemorrhage and thrombosis occurred at the lowest diastolicBP (<70 mm Hg).
ECC Findings
The Q-QS pattern on ECG at entry had the highestrelative risk for three patients with cerebral hemorrhage(2.5) and for three patients with cerebral thrombosis (2.7)(table 6). These figures are impressive since most of thepatients had no myocardial infarction. For subjects withhigh left R waves the relative risk was greater for cerebralhemorrhage than for cerebral thrombosis. In those patientswith ST depression and T abnormality, the high relativerisks might be related to hypertensive heart disease. The riskof cerebral hemorrhage was greater in patients with T ab-
TABLE 3 Index Cases of Cerebral Hemorrhage and Throm-bosis for the Evaluation of Risk Factors (Akabane and Asahi,1964 to 1970)
Cerebral hemorrhageMenWomen
Subtotal
Cerebral thrombosisMenWomen
SubtotalTotal
40-49
12
O O
C
O
03
Age50-59
83
11
91
1021
60-69
1916
35
206
2661
70-79
1210
22
148
2244
Total
4031
71
4315
58129
normality than in those with ST depression; however, thereverse was true for thrombosis, i.e., the risk of cerebralthrombosis was greater in patients with ST depression thanin those with T abnormality.
Ocular Funduscopic Findings
The relationship of funduscopic findings to the occurrenceof stroke was more remarkable for cerebral thrombosis thancerebral hemorrhage (table 6). Those patients whose fun-duscopic examination could not be performed due to variouseye diseases showed a fairly high relative risk for cerebralhemorrhage, especially thrombosis.
Grade 1 hypertensive retinal changes in patients withcerebral hemorrhage and thrombosis and Grade 1 scleroticfindings in patients with cerebral hemorrhage, according toScheie's classification (cited by Aoki8), were usuallyassumed to be within normal range. In those patients havingcerebral thrombosis with Grade 1 sclerotic findings, therelative risk was 4.2. Although, as usual, the grade classifica-tion was sometimes inconsistent (even by the samephysician), it would be advisable to discriminate the subjectswith Grade 1 sclerotic findings from the others as a high-riskgroup for cerebral thrombosis.
Combination of BP, ECG and Funduscopic Findings
Those subjects with BP of 160/95 mm Hg or higher andan ECG ST-T abnormality had greater age-adjusted andsex-adjusted relative risk of cerebral hemorrhage thanthrombosis, respectively (table 7). Abnormal funduscopicfindings were significant risk factors for cerebral thrombosisand hemorrhage. When combined, those persons with a high
TABLE 5BP Level
Age-Adjusted and Sex-Adjusted Relative Risk by
No.pts.
Cerebral hemorrhageNo.cases RRa (ratio)
Cerebral thrombosisNo.cases RRa (ratio)
Systolic BP (mm Hg)-119
120-139140-159160-179180-
8931,3891,015
545432
23
101938
Diastolic BP (mm Hg)- 69
70- 8990- 99
100-109110-
7842,070
748266192
220131522
0.22 (1.4x)*0.16 (l.O)f0.52 (3.3x)1.56 (9.8x)3.34 (20.9x)f
0.21 (1.0)*0.62 (3.0x)»0.86 (4.1x)2.46 (11.7x)t4.83 (23.0x)f
210102115
32017144
0.28 (1.0)0.71 (2.5x)0.62 (2.2x)2.05 (7.3x)f1.51 (5.4x)
0.41 (1.0)0.75 (1.8x)1.31 (3.2x)2.61 (6.4x)f1.03 (2.5x)
*p <0.05. tp <0.001. tp <0.01RRa: adjusted relative risk (ratio): relative risk ratio to the lowest.
by guest on Novem
ber 11, 2017http://stroke.ahajournals.org/
Dow
nloaded from
606 STROKE VOL. 7, No. 6, NOVEMBER-DECEMBER 1976
than to chronic nephritis or nephrosis (which are uncommonin these communities).
While elevated blood sugar as the primary determinant ofstroke risk is mentioned,26-27 our subjects with glycosuriashowed only a slightly higher relative risk of cerebralhemorrhage and thrombosis. Although the diabetics did notalways have glycosuria, those with consistent fastingglycosuria frequently had diabetes by further laboratorytests. It is well known that diabetes frequently accompaniesarteriosclerosis rather than hypertension, and this fact couldaccount for the high relative risk of stroke in patients withglycosuria. Some Japanese investigators, however, reportedthat glycosuria or diabetes mellitus was not significantlyrelated to stroke.4 In our study, albuminuria and glycosuriawere correlated with the occurrence of cerebral hemorrhageand thrombosis. To investigate the relationship of renal dis-eases or diabetes to cerebral hemorrhage and thrombosis,further investigation should be carried out. Low serumcholesterol levels could be a risk factor for cerebralhemorrhage;3-4 in this study, the relative risk ratio ofcerebral hemorrhage in those with serum cholesterol of 129mg/dl or less also was significantly higher (3.1) than that inthrombosis. The lowest incidence rate of cerebralhemorrhage was found in the group with cholesterol levels of130 to 159 mg/dl.
In the Framingham study28 the risk of atherothromboticbrain infarction was proportional to the serum cholesterolvalues in persons under age 50; however, this observation didnot apply to intracerebral hemorrhage. Serum cholesterollevels are much higher in persons in the United States thanin Japan, and this fact suggests that hypercholesterolemiawas not a risk factor for cerebral hemorrhage or thrombosisin this study.
Serum triglyceride levels might be a more significantpredisposing factor for stroke than the serum cholesterollevels,28 but we were unable to present follow-up data oftriglyceride levels at this time.
Although fragile capillaries would not directly causebleeding in the brain, capillary fragility counts might berelated to the incidence rate of stroke. High capillaryfragility counts proved to be associated with high relativerisk of cerebral hemorrhage. This may be related to the factthat cerebral hemorrhage was more prevalent than throm-bosis in women (table 3).
In daily life, work activity could'be an important index ofthe physical condition of our subjects, reflecting old age,retirement from work, or illness. The reason for a higherrelative risk of cerebral hemorrhage or thrombosis in thecasual work activity group was not very clear, but this groupmight be forced to work daily with some health handicapwhile other groups may not. The casual work activity groupconsisted of those who positively answered the question:"Are you working sometimes?" Details of the occupationwere unclear, and further analysis of the group is necessary.
It was unexpectedly remarkable that smoking habits wererelated to the occurrence of stroke, especially thrombosis,but the dose-response relationship was not demonstrated,and would be modified by other related factors such asmoderate smoking by sick persons. As it is widely acceptedthat smoking is harmful to persons with cardiovascular dis-eases, one reason for the high relative risk in mild smokersmight be their hesitation to admit excessive smoking. The
precise pathophysiological mechanism of smoking in rela-tion to the high incidence of stroke is not known, but onecould be that smoking affects peripheral vascular constric-tion leading to higher systemic BP,2' and also raises thecoagulability of blood contributing to the occurrence ofcerebral thrombosis. We obtained no history of smoking foranalysis and therefore are unable to report on the chroniceffects of continuous smoking. Although it was suggestedthat drinking habits did not show any significant relation-ship with the incidence rate of stroke, daily drinking ofalcoholic beverages accompanied the tendency toward ahigher rate of stroke and especially thrombosis. The effect ofdrinking would be the direct effect of alcohol in the cir-culatory system and the indirect effect of high-calorie foods,usually salty in Japan, taken with an alcoholic beverage.There is some doubt about the chronic effect of alcohol onthe occurrence of stroke.
No statistically significant relationship between rice in-take by bowls per day and cerebral hemorrhage was found,although relative risk was lower in our subjects ingesting 12to 19 bowls of rice per day. Excessive rice intake, usually ac-companied by large amounts of salt, and a monotonous dietcould reflect a healthy condition in our subjects.
Oral administration of antihypertensive drugs isassociated with a high relative risk of stroke, especiallyhemorrhage; this simply is another facet of the strongrelationship between hypertension and stroke. It might besuspected that hypertensive patients without therapy wouldshow a much higher relative risk of stroke than these figuresindicate.
Subjects with a previous history of stroke showed a veryhigh relative risk of having another attack, especiallycerebral hemorrhage. Relative risk in our subjects with aprevious history of hypertension was well matched to that ofpersons with previous oral administration of anti-hypertensive drugs.
Alter and Kluznik30 reported that in the absence of certainpredisposing illnesses, close relatives of CVA patientsappeared to have no greater risk of CVA than geneticallyunrelated individuals. However, in our subjects with a familyhistory of stroke there was a statistically significant high riskof cerebral hemorrhage. The possibility of genetic or socialinheritance of stroke exists. A study of twins or further in-vestigation of hereditary aggregation may disclose some in-heritance.26
It is not simple to evaluate subjective symptoms for theoccurrence of stroke because of unreliable reporting by oursubjects. Although some investigators reported that the in-cidence of stroke was high in subjects complaining of chestpain, nocturia, vertigo, dyspnea, and numbness, and soforth,5 all 26 subjective symptoms in our study showed nosignificantly higher relative risks of cerebral hemorrhageand thrombosis. Age-adjusted and sex-adjusted relative riskof cerebral hemorrhage (2.0) was observed in more than fourpatients with the following subjective symptoms: speech dis-turbance, fainting, insomnia, and stiff gait. These symptomsmay be related to cerebrovascular dysfunction or peripheralcirculatory disturbance.31
Conclusions
In Akabane and Asahi (Japan), 4,186 of 4,737 registeredresidents (40 to 79 years of age) were examined, and 264
by guest on Novem
ber 11, 2017http://stroke.ahajournals.org/
Dow
nloaded from
PROSPECTIVE STUDY OF CVD IN JAPAN. PART \/Okada et al. 607
cases of cerebrovascular attacks were observed between1964 and 1970. Age-adjusted and sex-adjusted relative risksof cerebral hemorrhage and thrombosis were evaluated onthe following bases: physical signs and laboratory valuessuch as anthropometric measurements, BP levels, ECG find-ings, ocular funduscopy findings, serum cholesterol levels,and medical history. It was stressed that elevated BP (evenwhen within normal limits) could be a recognizable risk fac-tor to the occurrence of stroke and especially thrombosis.Relative risk of cerebral hemorrhage in our subjects with thehighest category of BP was formidable, while risk ofcerebral thrombosis was less than in those with the secondhighest category of BP. Abnormal ECG findings were morepredictive of cerebral hemorrhage than thrombosis, and theocular funduscopic findings showed a much higherrelationship to cerebral thrombosis than hemorrhage.Among the concomitant presence of high BP, ST-T abnor-mality, and/or abnormal fundus, BP and ECG were mostpredictive of cerebral hemorrhage and ECG and fun-duscopic findings were predictive of cerebral thrombosis.
Among the less significant anthropometric measurements,neck circumference was most impressively related to the oc-currence of cerebral hemorrhage and thrombosis, and thosewith the least skinfold did not necessarily show the lowestrelative risk of stroke. While relative risk of cerebralhemorrhage in subjects with albuminuria was significantlyhigh, risk of cerebral thrombosis was highest in those withglycosuria. The patients in the lowest category of cholesterollevel (< 130 mg/dl) had the highest relative risk of cerebralhemorrhage. Capillary fragility had a high relative risk ofstroke, especially hemorrhage.
No or restricted daily work activity suggested high risk ofcerebral hemorrhage and thrombosis. Smoking habit wassignificantly related to cerebral thrombosis rather than tocerebral hemorrhage, although this report lacks the dose-response relationship between the number of cigarettessmoked and the relative risks. Previous history of stroke orhypertension was reasonably related to the occurrence ofstroke, especially cerebral hemorrhage; family history ofstroke did show a significantly higher relative risk. Among26 subjective symptoms, those of speech disturbance, faint-ing, insomnia, and stiff gait showed a slightly higher relativerisk of cerebral hemorrhage only.
These risk factors were related together and not inde-pendently compared to the occurrence of stroke. Multi-variate analysis (to be reported in another paper) mightgive additional information about the relationship betweenthese risk factors and cerebrovascular disease.
Acknowledgment
The authors would like to express their appreciation to Drs. YasushiMizuno, Haruo Takahashi, Tohru Iwatsuka, Masao Ito, and NaohikoOzawa for their cooperation and for the assistance of their colleagues in theFirst Department of Internal Medicine and in the Department of Health,Aichi Prefecture.
References
1. McDowell F: Report of the New Orleans (1970) meeting concerning theepidemiology of cerebrovascular disease. Stroke 2: 101-104, 1971
2. Katsuki S, Hirota Y: Recent trends in incidence of cerebral hemorrhageand infarction in Japan. A report based on death rate, autopsy case andprospective study on cerebrovascular disease. Jap Heart J 7: 26-34, 1966
3. Komachi Y: Hypertension and cerebral apoplexy. Mod Med 26:1897-1906, 1971 (in Japanese)
4. Owada K, Tanaka H, Ueda U, et al: An epidemiological study on riskfactors for cerebral stroke and ischemic heart disease of Japanese. Jap JPublic Health 21: 425-433, 1974 (in Japanese)
5. Hashimoto T, Kuboki S, Kawano S, et al: An epidemiological study onrisk factors for developing cerebrovascular diseases. Jap J Public Health17: 1131-1140, 1970 (in Japanese)
6. Konishi M: Epidemiological studies on cerebrovascular diseases with thespecial reference to the pathological findings on stroke in Akita district.Jap J Public Health 21: 485-500, 1974 (in Japanese)
7. Nishida T: An epidemiologica! study on cerebrovascular disease — 10years' follow-up data in Hakushu-cho, Yamanashi, Japan. Jap J Geriatr11: 382-392, 1974 (in Japanese)
8. Aoki N: Epidemiological evaluation of funduscopic findings in cerebro-vascular diseases. 1. Funduscopic findings as risk factors for cerebro-vascular diseases. Jap Circ J 39: 257-269, 1975
9. Horibe H: Cerebro- and cardio-vascular diseases in a Japanese ruralcommunity, Akabane. Jap Circ J 33: 1477-1481, 1969
10. Okada H, Ohno Y, Horibe H: A study of cerebrovascular accidents inNagoya. Jap Med J (No. 2401): 30-34, 1970 (in Japanese)
11. WHO: Cerebrovascular diseases: Prevention, treatment, and rehabilita-tion. WHO Tech Rep Ser #469, Geneva, 1971
12. Ministry of Health and Welfare: Vital statistics 1972, Japan. Jap AssocHealth Stat, Tokyo, 1974
13. Kojima S: Distinctive features of cerebral apoplexy seen around Akitadistrict. Jap J Public Health 13: 907-924, 1966 (in Japanese)
14. Ohno Y: Biometeorologic studies on cerebrovascular diseases. 1. Effectsof meteorological factors on the death from cerebrovascular accident.Jap Circ J 33: 1285-1298, 1969
15. Ohno Y, Aoki N, Horibe H, et al: Biometeorologic studies on cerebro-vascular diseases. V. A multivariate analysis of meteorologic effects oncerebrovascular accident. Jap Circ J 38: 195-208, 1974
16. Millikan CH, Adams RD, Fang HCH, et al: A classification and outlineof cerebrovascular diseases. Neurology 8: 395-434, 1958
17. Hirota T, Lau A: A study of the validity of differential diagnosis and 10years' trend of autopsies in Japan. Jap J Geriatr 12: 7-11, 1975 (inJapanese)
18. Japanese Pathological Society (eds): Annal of Pathological AutopsyCases in Japan
19. Florey CV, Senter MG, Acheson RM: A study of the validity of thediagnosis of stroke in mortality data. II. Comparison by computer ofautopsy and clinical records with death certificates. Am J Epidemiol 98:15-24, 1969
20. Kannel WB, Blaisdell FW, Gifford R, et al: Risk factors in stroke due tocerebral infarction — a statement for physicians. Stroke 2:423-428, 1971
21. Veterans Administration Cooperative study group on antihypertensiveagents: Effects of treatment on morbidity in hypertension. II. Results inpatients with diastolic blood pressure averaging 90 through 114 mm Hg.JAMA 213: 1143-1152, 1970
22. Arai H, Akiyama M, Hata T, et al: Evaluation of antihypertensionprogram by ocular funduscopic findings. Especially on the prospect ofcerebral apoplexy. Public Health 34: 55-63, 1970 (in Japanese)
23. Shibata S: Reinvestigation of eye fundus findings in cerebrovascular sur-vey. Jap J Public Health 20: 449-457, 1973
24. Klassen AC, Loewenson RB, Resch JA: Body weight, cerebral athero-sclerosis and cerebral vascular disease: An autopsy study. Stroke 5:312-317, 1974
25. Olivares L, Castaneda E, Grife A, et al: Risk factors in stroke: A clinicalstudy in Mexican patients. Stroke 4: 773-781, 1973
26. Kuller K, Tonascia S: A follow-up study of the commission on chronic ill-ness morbidity survey in Baltimore. IV. Factors influencing mortalityfrom stroke and arteriosclerotic heart disease (1954-1967). J Chron Dis24: 111-124, 1971
27. Lavy S, Melamed E, Cahane E, et al: Hypertension and diabetes as riskfactors in stroke patients. Stroke 4: 751-759, 1973
28. Kannel WB, Gordon T, Dawber TR: Role of lipids in the development ofbrain infarction: The Framingham study. Stroke 5: 679-685, 1974
29. Okada H, Horibe H, Ohno Y, et al: Essential hypertension and smokinghabits in a Japanese rural community. Jap J Public Health 15: 901-914,1968 (in Japanese)
30. Alter M, Kluznik J: Genetics of cerebrovascular accidents. Stroke 3 :41-48, 1972
31. Okada H, Aoki K, Horibe H, et al: A study of subjective symptoms ofarteriosclerosis in a population. Jap J Geriatr 4: 246-256, 1967 (inJapanese)
by guest on Novem
ber 11, 2017http://stroke.ahajournals.org/
Dow
nloaded from
H Okada, H Horibe, O Yoshiyuki, N Hayakawa and N Aokithrombosis.
Asahi. Part 1: evaluation of risk factors in the occurrence of cerebral hemorrhage and A prospective study of cerebrovascular disease in Japanese rural communities, Akabane and
Print ISSN: 0039-2499. Online ISSN: 1524-4628 Copyright © 1976 American Heart Association, Inc. All rights reserved.
is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231Stroke doi: 10.1161/01.STR.7.6.599
1976;7:599-607Stroke.
http://stroke.ahajournals.org/content/7/6/599World Wide Web at:
The online version of this article, along with updated information and services, is located on the
http://stroke.ahajournals.org//subscriptions/
is online at: Stroke Information about subscribing to Subscriptions:
http://www.lww.com/reprints Information about reprints can be found online at: Reprints:
document. Permissions and Rights Question and Answer available in the
Permissions in the middle column of the Web page under Services. Further information about this process isOnce the online version of the published article for which permission is being requested is located, click Request
can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial Office.Stroke Requests for permissions to reproduce figures, tables, or portions of articles originally published inPermissions:
by guest on Novem
ber 11, 2017http://stroke.ahajournals.org/
Dow
nloaded from
Recommended