Determinants of lipid and lipoprotein level in elderly men

Atherosckrosis. 60 (1986) 221-229 Elsevier Scientific Publishers Ireland.

221 Ltd

ATH 03777

Determinants Sigismond

of Lipid and Lipoprotein

Deutscher,

Level in Elderly Men

Margaret W. Bates, Myrven J. Caines, Ronald Anthony Puntereri and Floyd H. Taylor

E. LaPorte,

Veterans Admrnrstration Medical Center and Departments of Medrcrne. Community Medicine and Pathologv. Unwersity of Pittsburgh School of Medicine, and Department of Epldemlolog: Graduate School of Pubhc Health, Pittsburgh. PA (U.S.A.) (Received 25 July. 1985) (Revised. received 18 December. 1985) (Accepted 20 December, 1985)

Summary In an effort to better understand the relationship existing between lipoprotein pattern and longevity, we studied the lipid and lipoprotein distribution of 94 men over age 80 who lived in a nursing home. and assessed the role of selective mortality, body mass and sex hormone secretion in determining these distributions. High density lipoprotein subfraction and serum testosterone measurements were obtained on subsamples. The main findings were: (a) Presence of a lipoprotein pattern characterized by low LDL (total serum cholesterol: 179.6 t- 36.0 mg/dl; LDL cholesterol: 106.3 f 31.2 mg/dl) and high HDLz cholesterol (18.5 f 10.2 mg/dl) levels. (b) Occurrence of a positive association between LDL and HDL, (r = 0.51, P < O.Ol), resulting in an overall high HDL,/HDL, ratio. Mortality over a 6-month period was directly related to LDL level and possibly inversely related to HDL, level, suggesting that selective mortality played a major role in determining the pattern observed. Body mass and serum testosterone concentration. which tended to be low, were independently correlated with lipoprotein level; a particularly strong (r = 0.68, P < 0.01). The latter correlation (positive) existed between free testosterone and triglyceride results suggest that changes related to senescence also influenced lipid and lipoprotein levels.

Key words:

Atherosclerotic disease - Lipoprotein mortalitv - Serum testosterone

inter-relationships

Introduction Low density lipoprotein an accelerating effect upon

(LDL) elevation has atherogenesis whereas

Reprint requests: Service. VA Medical PA 15240, U.S.A.

Sigismond Deutscher. M.D., Medical Center. University Drive C, Pittsburgh.

0021.9150/86/.$03.50

@)1986 Elsevier Scientific

Publishers

Ireland,

- Longeuity

- Risk factors ~ Selectrue

high density lipoprotein (HDL) seems to protect against it. It is tempting therefore to hypothesize that low LDL and/or high HDL levels play a crucial role in ensuring survival to an advanced age. Studies focusing on kindreds (1) provide some supportive evidence for such a hypothesis. Findings from studies dealing with the general population appear to be contradictory in this respect, Ltd

222

however. On the one hand, cross-sectional investigations [2-41 indicate that total serum cholesterol and LDL levels gradually decline in old age after peaking in middle-age; this is consistent with the notion that older individuals with higher cholesterol values are being selectively removed from the population through death. On the other hand, total cholesterol and LDL are known to constitute weak predictors of coronary artery disease (CAD) after middle-age [5], which in turn seems to suggest that factors other than selective mortality account for the low levels found among the elderly. Similar ambiguity exists with respect to HDL: low levels increase CAD risk at middleage and beyond, but mean cross-sectional measurements seem to remain essentially stable throughout the late adult period [2-41, contrary to what might be expected if selective mortality were operative. In an effort to help clarify the above points, we examined the lipoprotein-cholesterol distribution, including cholesterol distribution between HDL occurring among a group of ocsubclasses, tagenarians and nonagenarians living in a nursing home, and assessed the role of selective mortality as well as other potential factors such as nutritional and endocrine status, in determining lipid and lipoprotein level. Methods Study population All male patients above the age of 80 who figured on the active roster of an extended care facility serving veterans, at a given point in time (first week of May 1983), were automatically included in the study. The population thus selected, comprising 94 subjects, was 85% white and 15% black. Two main categories of patients were distinguishable among the participants according to whether their general health was compromised or not: (a) A large group of patients constituting the majority of the study population, who suffered from senile dementia or similar conditions but were otherwise in relatively good health, and who tended therefore to survive for prolonged periods of time. (b) A smaller group (roughly one fourth of the study population) consisting of patients with a much shorter survival expectancy who in

most instances were in the late stages of ultimately fatal chronic illnesses; the latter sub-group contained a high proportion of individuals afflicted with advanced vascular disease. Variables studied At entry into the study, blood was drawn after an overnight fast, and total cholesterol (TC), triglyceride (TG) and HDL cholesterol (HDL-C) determinations performed on the serum. Automated enzymatic procedures were used to quantify TC [6] and TG [7]; these procedures were carried out at VA Medical Center Pittsburgh under the direction of one of us (M.C.). HDL-C quantification, in turn, was carried out by a commercial laboratory (Smith-Kline Clinical Laboratory Inc., Ring of Prussia, PA): The approach followed consisted in precipitating apo-B-containing lipoproteins by adding heparin manganese chloride [S] and subsequently measuring the cholesterol remaining in the supernatant by an enzymatic method [9]. LDL cholesterol (LDL-C) and body mass were estimated by indirect methods: the former by using the Friedenwald formula [lo] and the latter by calculating an index based on available weight and height measurements (weight in kg divided by the square of height in m*). Two-thirds of the participants were still alive at the end of one year. Repeat blood samples were obtained on 37 of these survivors for follow-up and other purposes. TC, TG and HDL-C were re-analysed in the same manner as at entry into the study. Additional serum aliquots were frozen at - 70°C for future HDL, and HDL, cholesterol (HDL,-C and HDL,-C) quantification; the latter analyses were performed approximately 3 weeks later at the Heinz Nutrition Laboratory (a University of Pittsburgh research facility). It has been shown that HDL,-C and HDL,-C concentrations remain stable for several months when plasma samples are frozen at this temperature (Bates, M.W. et al: unpublished data). HDL fractionation was carried out according to the method recommended by Gidez et al. [ll], which includes the following steps: (a) precipitating HDL, by addition of dextran sulfate after prior precipitation of apo-B-containing lipoproteins by means of heparin and manganese, (b) determining total

223

HDL-C and HDL,-C on the heparin-manganese and dextran sulfate supernatants, respectively, and then (c) calculating HDL,-C by difference. Serum aliquots were also used to assess free and total testosterone levels by equilibrium dialysis and radioimmunoassay methods among a sub-sample encompassing 17 of the 37 survivors mentioned above. Stutistical ma&is In order to test for a possible relationship of high TC and/or low HDL-C with increased mortality, participants were cross-classified by variable level at entry and survival status at the end of specified periods of time. TC values exceeding 220 mg/dl were considered ‘high’ for the purpose of the present study; HDL-C values below 45 mg/dl as ‘low’. Adjustment for potential confounders such as age was done by stratification. Age-adjusted risk ratios for ‘high’ TC and ‘low’ HDL-C (expressed as odds ratios) were calculated by the Mantel-Haenszel procedure [12] and Mantel extension test [13], as indicated. The factors under study were also treated as continuous variables whenever appropriate: Summary statistics were computed for pertinent sub-groups and differences between them assessed by means of parametric tests (two-sample t-test and one-way analysis of variance). Relationships between selected variables were assessed by computing simple and partial correlation coefficients.

TABLE

Results Age-related differences in cross-sectional dutu Mean TC, LDL-C, TG, HDL-C and body mass index at entry into the study are shown in Table 1; members of the study population have been further divided into 3 sub-groups according to age reached. It can be seen that the trend towards a decline in TC, LDL-C and TG level with advancing age previously reported in relation to somewhat younger populational groups [2-41, seems to persist among the very old. The age-related differences detected in our data did not reach statistical significance when tested by analysis of variance, however (0.10 < P < 0.25, in the case of TC). In contrast with the pattern observed for TC. LDL-C and TG, HDL-C did not appear to vary appreciably with age. Effect of selective mortalit] The effect of ‘high’ TC and ‘low’ HDL-C upon risk of dying from any cause within 6- and 12month periods is described in Table 2: three distinct age groups are again represented: ‘high’ TC was clearly related to increased mortality, but no such relationship was apparent for ‘low’ HDL-C. None of the 7 ‘high’ TC patients who died within 6 months had secondary hyperlipidemia attributable to renal disease, myxedema or uncontrolled diabetes. Death was ascribed by the attending physician to myocardial infarction in 3 instances

1

SERUM

LIPIDS

AND

LIPOPROTEINS,

AND BODY MASS INDEX

Figures shown are mean & SD. TC. LDL-C. HDL-C and TG values are in mg/dl.

AT ENTRY

INTO STUDY,

BY AGE

body mass index in kg/m2.

Age group (yr)

Total cholesterol

LDL cholesterol

HDL cholesterol

Triglyceride

Body mass index ’

80-84 (n = 20) 85-89 (n = 52) 90+ (n=22) All (n = 94)

190.0+ 179.0* 169.8 f 179.6 +

114.0 + 33.8 ’ 107.3 * 30.1 c 91.6 k 29.9 106.3rt31.2d

51.1* 49.5 * 49.4+ 49.8 f

115.6k60.6 110.5 + 65.6 107.0 + 54.9 110.8 + 62.2

21.9k3.3 22.2 k 3.8 20.5 + 2.7 21.8 f 3.5

a ’ ’ ’

Based Based Based Based

on on on on

82 19 51 92

observations. observations. observations. observations.

35.3 33.9 38.8 36.0

14.7 12.0 13.8 13.1

224 TABLE

2

EFFECT

OF LIPID

Age Survivor group status

AND

LIPOPROTEIN

LEVEL

UPON

RISK OF DYING

Risk of dying within 6 months

n

WITHIN

SPECIFIED

PERIODS

OF TIME

Risk of dying within 12 months

Total cholesterol

HDL cholesterol

No. with level 2 220

No. with level < 45

n

Total cholesterol

HDL cholesterol

No. with level 2 220

No. with level < 45

(yr) Odds ratio

Odds ratio

Odds ratio

Odds ratio

80-84

Deceased Alive

146

2(14%) 2 (33%)

3.00

25 (36%) (33%)

0.90

137

22(15%) (29%)

2.20

25 (38%) (29%)

o.64

85-89

Alive Deceased

41 11

3 (27%) (7%)

4.75

193 (27%) (46%)

o,43

18 34

3 (17%) (9%)

2.07

157 (39%) (44%)

0.81

90+

Deceased Alive

5 17

2 (40%) 0 (0%)

+ o.

2 (40%) 5 (29%)

1.60

6 16

2(33%) 0 (0%)

+cc

2 (33%) 5 (31%)

1.10

All

Deceased

22

7 (32%)

7 (32%)

o,71 a

31

7(23%)

Alive

72

5 (7%)

a b ’ d

5.73 a.d (1.44-26.74)

b 29 (40%)

(0.22-2.16)

Mantel-Haenzel age-adjusted summary odds ratio. 95% Confidence interval calculated by Cornfeld method. Overall Mantel-Haenszel &i-square: 3.74, P = 0.05. Overall Mantel-Haenszel chi-square: 8.57, P < 0.01. TC and HDL-C

(infarct occurrence being fully documented in only one instance, however); a fourth patient died of a cerebra-vascular accident and a fifth of what appears to have been sudden cardiovascular death without demonstrable MI. As might be expected on account of the short survival span encountered among persons with advanced vascular disease, risk ratio estimates pertaining to ‘high’ TC were greater for 6-months than 12-months end points. Mean TC concentration for individuals alive at the end of 6 months was 177.1 f 29.1 mg/dl as compared to 188.0 * 50.5 mg/dl for those deceased within that period (0.10

TABLE

b 63

5 (8%)

3.18 a,C (0.83-14.03)

11 (35%) b 25 (40%)

0.82 a (0.30-2.10)

b

values are in mg/dl.

continuous over the range of the distribution, the patients were also grouped into ordered categories corresponding to concentration quintiles, and age-adjusted risk ratios calculated in relation to the lowest quintile: The risk ratios obtained for the second, third, fourth and fifth quintiles were 1.4, 1.5,0.7 and 4.4 respectively (the data referring to 6- months mortality), indicating that the association detected is restricted to the upper end of the distribution. Comparison between baseline and follow-up measurements Mean TC and HDL-C values observed at the end of one year did not differ appreciably from

3

COMPARISON BETWEEN BASELINE AND FOLLOW-UP (1 YEAR) TOTAL BODY MASS INDEX, IN SUB-SAMPLE CONSISTING OF 37 PATIENTS TC and HDL-C

values are in mg/dl,

CHOLESTEROL,

HDL CHOLESTEROL

body mass index in kg/m2.

Variable

Mean value at entry

Mean value at end of one year

Difference

Correlation between readings

Total cholesterol HDL cholesterol Body mass index

175.4 48.4 22.2

180.5 44.2 22.0

+ 5.1 - 4.2 - 0.2

0.79 0.83 0.85

AND

225

TABLE

4

DISTRIBUTION SUB-CLASSES Figures

OF CHOLESTEROL (n = 37)

BETWEEN

HDL

shown are mean * SD. Values are in mg/dl. 44.3*11.5 18.5 f 10.2 25.8f 5.3

Total HDL cholesterol HDL, cholesterol HDL, cholesterol Percent of total HDL cholesterol carried by HDL, sub-fraction

41.8%

those found at entry; furthermore, the two sets of measurements were highly correlated (Table 3). These results indicate that no substantive longitudinal changes within individuals took place over the period studied. Distribution of cholesterol between HDL sub-classes Recent evidence suggests that the protective effect shown by HDL in relation to CAD is primarily associated with the HDL, sub-fraction [14,15]. HDL,-C levels tend to be lower among men than women, mean values being 15 mg/dl or less for the former and about 20 mg/dl for the latter; HDL,-C levels, on the other hand, tend to be similar in the two sexes, averaging approximately 30 mg/dl [ll]. (Also, LaPorte R. et al., unpublished data referring to values obtained at the Heinz Nutrition Laboratory under the same conditions as in the present study.) These observations, it should be noted, refer to persons under age 65; to the best of our knowledge, no previous data specifying cholesterol distribution between

TABLE

Inter-relationships among lipoprotein classes Correlations between pertinent variables are given in Table 5. As anticipated, TC and LDL-C were highly correlated, and an inverse relationship was detectable between TG and HDL,-C (but not with HDL,-C). The most interesting aspect of these data, however, lies in the fact that: (a) a relatively strong positive correlation seems to exist between TC (and consequently also LDL-C) and HDL,-C, (b) not only TG but also TC and LDL-C appear to be inversely correlated with HDL,-C. The latter inverse correlations were not dependent upon body mass (partial correlation between LDL-C and HDL,-C, holding body mass index

5

INTER-RELATIONSHIPS Figures

HDL sub-classes are available for persons above that age, using the Gidez procedure. It can be seen from Table 4 that HDL,-C concentrations among the elderly men included in the present study were substantially higher and HDL,-C concentrations lower than the values mentioned for younger men. Since, however, these age-related cross-sectional differences tend to counter-balance one another, total HDL-C does not vary with advancing age as cholesterol distribution between sub-classes seems to. It is interesting to note that the HDL,-C concentrations seen in our subjects are comparable to the values previously mentioned in relation to younger women. The pattern observed further resembles the one existing for women in that, HDL,-C being relatively low among our patients, a high proportion of total HDL-C (almost half as compared to less than one-third in the case of younger men) is carried by the HDL, fraction.

AMONG

shown are linear correlation

LIPID

AND

coefficients

LIPOPROTEIN

CLASSES

(n = 37)

(r).

Variable

LDL cholesterol

Triglyceride

Total HDL cholesterol

HDL, cholesterol

HDL, cholesterol

Total cholesterol LDL cholesterol Triglyceride Total HDL cholesterol HDL, cholesterol

0.92 ****

0.41 *** 0.25

0.08 - 0.06 -0.43 ***

- 0.20 - 0.33 ** -0.44 *** 0.88 ****

0.54 **** 0.51 *** - 0.07 0.43 *** - 0.05

* P -c 0.10. ** P c 0.05. *** P < 0.01. **** P < 0.001, Correlations involving LDL cholesterol based on 35 observations. Correlations involving triglyceride based on 36 observations.

226

TABLE

6

RELATION

WITH

BODY

MASS AND

Figures shown are linear correlation index constant. Variable

Total cholesterol LDL cholesterol Triglyceride Total HDL cholesterol HDL, cholesterol HDL, cholesterol

SERUM

coefficients

Body mass index (n = 35) 0.34 ** 0.29 * 0.42 *** - 0.23 - 0.27 0.01

TESTOSTERONE

(r). Parenthetical

entries represent

Serum testosterone

partial

correlation

coefficients

holding

body mass

(n = 17)

Total

Free

Percent

0.47 * (0.32) 0.37 = 0.20 0.33 0.09 0.48 ** (0.37)

0.56 ** (0.37) 0.48 a.* (0.24) 0.68 *** (0.68) *** - 0.09 - 0.24 0.30

0.49 0.44 0.84 - 0.45 - 0.48 0.05

free ** (0.31) a.* (0.24) **** (0.85) **** * ( - 0.43) * ** ( - 0.37)

a Based on 16 observations only. * P < 0.10. ** P-c 0.05. *** P < 0.01. **** P -c 0.001.

constant, was -0.30, P < 0.10). On the other hand, HDL, and HDL, were not significantly correlated in these data. Relationship with body mass and serum testosterone level Correlations between TC, LDL-C, TG, and HDL-C and its component fractions on the one hand, and body mass index and serum testosterone on the other, are shown in Table 6. Predictably, there was a positive correlation (albeit a modest one) between the first three variables mentioned and body mass, and a negative correlation of similar magnitude between the latter and HDL,-C; HDL,-C on the other hand, was unrelated to body mass. Stronger correlations were observed with free testosterone (positive correlations for TC, LDL-C and TG; negative correlation for HDL,); a particularly strong correlation existed between free testosterone and TG. Both testosterone and lipoprotein levels were correlated with body mass in these data; this could have given rise to confounding. Controlling for body mass reduced somewhat the correlations existing between testosterone, TC and LDL, but did not alter appreciably those with TG and HDL (see partial correlation coefficients shown in Table). Discussion Two main findings emerge from the present study which focused upon elderly males living in a

nursing home: (a) The existence of a lipoprotein pattern characterized by low LDL and high HDL, levels, (b) the occurrence of a positive association between LDL and HDL, resulting in relatively high HDL,/HDL, ratios: the latter being of approximately the same order of magnitude as noted among younger women who tend to be immune to the manifestations of atherosclerosis. Our results, taken at their face value, seem to indicate that not only selective mortality but also body mass and level of circulating free testosterone played a role in producing the seemingly protective pattern observed in the data. The following basic questions arise in connection with these results: (a) Do elderly men who live freely in the community (as opposed to institutionalized patients) display the same pattern? (b) If so, is that pattern due mainly to changes related to senescence or to the cumulative effect of selective mortality? A number of our subjects were afflicted with debilitating conditions (see description of study population under Methods) that resulted in malnutrition and might thus have affected overall lipid and lipoprotein levels. If this were indeed the case, a similar pattern should be observable among younger subjects with compromised nutrition. In order to test this possibility, we measured TC, HDL, and HDL, levels among a group of 43 men residing in the same nursing home, who were under the age of 65. This group consisted mainly of severely impaired individuals with dementia

227

due to various causes, cord injury, multiple sclerosis or schizophrenia, whose nutritional status was by and large comparable to that of the older subjects we studied; some of these patients were admitted to the facility because they had had multiple strokes or suffered brain damage as a result of prolonged cardiac arrest. The lipoprotein pattern displayed by the younger subjects differed substantially from that previously described for the older ones: mean TC, HDL,-C, HDL,-C and percentage of cholesterol carried by the HDL, sub-fraction were 198.2 f 55.7 mg/dl, 11.6 + 8.3 mg/dl, 29.9 + 5.4 mg/dl and 26% (as compared to 179.6, 18.5, 25.8 and 41.8%. respectively); there was no significant correlation between TC and HDL,. (Excluding the patients with past strokes or known CAD did not alter appreciably the mean values observed for this younger group.) We also compared the members of our study population (comparison restricted to the 82 elderly members for whom body mass index had been computed) to a group of 25 men of similar age who lived freely in the community and received routine geriatric care at the clinic of one of us (SD.). Although these men were heavier than the nursing home patients studied (mean body index being 24.9 k 4.1 as compared to 21.8 + 3.5, P < O.Ol), they displayed total serum cholesterol concentrations that tended to be only marginally higher (184.7 + 40.1 mg/dl compared to 181.2 + 40.8 mg/dl. P > 0.5). Thus, body mass (used here as an index of malnutrition), seems to have played little or no role in producing the overall low level of total cholesterol observed in our data. Similarly, there were no appreciable differences in mean HDL cholesterol values between the comparison group and the nursing home subjects (51.5 & 18.4 mg/dl compared to 49.8 + 13.1 mg/dl). We did not measure HDL sub-fractions in the former group and do not known therefore whether these free-living elderly tend to have the same high HDL: and low HDL, concentrations as the institutionalized elderly seemed to. It is worth mentioning in this connection that most of the 37 individuals on whom HDL sub-fractionation was carried out at the end of one year enjoyed relatively good general health, as evidenced by the fact that 27 of these one-year survivors were still alive at the end of 2 years.

Moreover, many of the institutionalized elderly included in the study were taking various medications that might also have affected lipoprotein concentrations. We therefore searched carefully for a possible relationship between lipoprotein pattern and drug usage; none was found. We show (see Table 2) as others have done before [16] that mortality (presumably from vascular disease) among the very old may be related to level of serum cholesterol (LDL), in selected populations. (The fact that we dealt with a nursing home population may have led to an overestimation of the risk associated with cholesterol elevation. It is unlikely therefore that hypercholesterolemia constitutes as important a determinant of mortality in the elderly male population at large as it appears on the basis of this study.) The above finding appears somewhat surprising in view of the fact that serum cholesterol elevation does not constitute a significant risk factor by itself in relation to CAD, after middle-age [5]. It should be kept in mind, however, that: (a) death at an advanced age may represent the belated consequence of a disease process initiated many decades earlier. (b) Since serum cholesterol was negatively correlated with HDL, in our data (see Table 5) it seems reasonable to assume that individuals with increased LDL who died of vascular disease also tended to have decreased levels of HDL,. Actual HDL, measurements were not available for the deceased group, but the unusually high concentrations found among the survivors are consistent with the notion that HDL, level plays a continuing role in influencing mortality. The individuals who died of vascular disease may thus have displayed a lipoprotein constellation (unfavourable LDL/HDL, ratios) conducive to more rapid atherosclerosis progression. We are referring hereto a possible effect of hypercholesterolemia occurring in conjunction with a low HDL, level as opposed to isolated total cholesterol elevation. Failure to detect an inverse relationship between HDL (total HDL cholesterol) and mortality in our data (see Table 2) does not invalidate the foregoing argument since such failure could have been due to negative confounding on the part of total serum cholesterol (LDL) which was positively associated with both HDLl and mortality.

228

Even though the approach followed was not suitable for adequate risk estimation in the general population, as indicated above, it nevertheless enabled us to demonstrate rather unequivocally that selective mortality continues to be operative in old age. Moreover, the fact that HDL, and HDL, age-related variations tended to cancel one another (increased levels of the former accompanied by decreased levels of the latter) seems to provide an explanation for the apparent paradox previously referred to (see Introduction), namely that low levels of total HDL cholesterol are known to increase CAD risk at middle-age and beyond but mean cross-sectional measurements appear to remain stable. The study also seems to suggest that changes related to senescence are capable of influencing lipoprotein levels. Sex hormone concentration in males declines after middle-age; there is, however, marked inter-individual variation with regard to the rate at which such decline occurs [17,18]. Levels of free testosterone in the population we studied tended to be very low (59.6 + 39.3 pg/ml; the normal range observed with the methods of assay used being loo-330 pg/ml for adult males covering a broad age range). The correlations detected between free testosterone and lipoprotein concentrations (see Table 6) point to the existence of a link between the generally low levels of that sex hormone observed in the study population and the distinctive lipoprotein pattern it exhibits. Since controlling for differences in body mass did not reduce substantially the strength of the aforementioned correlations (see Table 6) it seems unlikely that the latter arose as a result of a common association with nutritional status. One cannot. however, rule out the possibility that other, unspecified, confounding variables account for the correlations described. Whether these relationships between free testosterone and lipoprotein concentrations (assuming that they are of cause and effect) also hold for the somewhat higher testosterone levels one might find among elderly men who live freely in the community and for the much higher levels existing among younger men, cannot be determined on the basis of the information available to us and remains therefore open to question. In fact, studies dealing with middle-aged men [19,20] seem to indicate that the afore-men-

tioned relationships do not hold in the case of younger men. Nor does it necessarily follow from the above findings that any such hormone-related changes in lipoprotein pattern should affect CAD risk. Our results do not preclude, however, such a possibility, since there might have been a connection between the hormone-lipoprotein (Table 6) and lipoprotein-mortality (Table 2) relationships detected in this study. Viewed in such a context, our observation showing that endogenous testosterone secretion and triglyceride level were strongly correlated (positive correlation) appears to be interesting on account of the close metabolic relationship (inverse relationship) existing between VLDL (triglyceride) and HDL, a key factor correlating with protection against atherosclerosis. The fact that total cholesterol (LDL) and HDL, were positively correlated in our data seems to represent an intriguing and potentially important finding which calls for further clarification. It is interesting to note in this respect that HDL, and HDL, were both correlated with LDL, but that the correlations had opposite signs, lending thus added credence to the notion that the two sub-fractions fulfill different (but possibly interrelated) metabolic functions. It should also be mentioned that the positive correlation between LDL and HDL, here reported does not constitute an isolated occurrence restricted to the institutionalized elderly: A similar correlation has been reported for the Modesto, California population consisting of younger men and women [21]. Furthermore, preliminary results from a study dealing with college students (LaPorte, R. et al: unpublished data) suggest that total cholesterol and HDL, are positively correlated among females (n = 19, r = 0.42, P < 0.1). but not among males (n = 26, r = -0.03). It would seem thus that the TC-HDL, (LDL-HDL,) relationship uncovered is hormone-dependent too. References Glueck, C.J., Gartside, P.S., Steiner, P.M. et al.. Hyperalpha and hypobeta lipoproteinemia in octagenarian kindreds. Atherosclerosis. 27 (1977) 387. Rifkind, B.M., Tamir, I.. Heiss, G., Wallace, R.B. and Tyroler, H.A., Distribution of high density and other lipoproteins in selected LRC prevalance study populations - A brief survey, Lipids, 14 (1979) 105.

229

3 Nicholson, J., Gartside. P.S.. Siegel, M.. Spencer, W., Steiner. P.M. and Glueck, C.J., Lipid and lipoprotein distributions in octo- and nonagenarians, Metabolism, 28 (1979) 51. 4 Alvarez. C.. Orejas. A., Gonzalez, S., Diaz. R. and Colomo. L.F.. Reference intervals for serum lipids, lipoproteins and apoproteins in the elderly, Clin. Chem., 30 (1984) 404. 5 Kannel. W.B.. Castelli, W.P. and Gordon. T., Cholesterol in the prediction of atherosclerotic disease - New perspectives based on the Framingham Study, Ann. Intern. Med., 90 (1979) 85. 6 Allain. CC., Poon, L.S., Chan, C.S.G.. et al., Enzymatic determination of total serum cholesterol, Clin. Chem.. 20 (1974) 470. 7 Sampson, E.J., Demers, L.M. and Krieg, A.F.. Faster enzymatic procedure for serum triglycerides. Clin. Chem.. 21 (1975) 1983. 8 Albers. J.J., Warnick. G.R.. Wiebe, D.. et al.. Multi-laboratory comparison of three heparin-manganese precipitation procedures for estimation of cholesterol in high-density lipoprotein, Clin. Chem.. 24 (1978) 853. 9 Steele. B.W.. Koehler, D.F.. Azar. M.M.. et al., Enzymatic determinations of cholesterol in high-density lipoprotein fractions prepared by a precipitation technique. Clin. C’hem.. 22 (1976) 96. 10 Friedenwald, W.T.. Levy, R.J. and Fredrickson, D.S., Estimation of the concentration of low-density lipoprotein cholesterol in plasma without use of preparative ultracentrifuge, Clin. Chem.. 18 (1972) 499. 11 Gidez. L.I.. Miller. G.J.. Burstein, M., Stagle. S. and Eder. H.A.. Separation and quantitation of sub-classes of human plasma high density lipoprotein by a simple precipitation procedure, J. Lipid Res.. 23 (1982) 1206.

12 Mantel, N. and Haenszel, W., Statistical aspects of the analysis of data from retrospective studies of disease. J. Nat. Cancer Inst., 22 (1959) 719. 13 Mantel, N., Chi-square tests with one-degree of freedom Extensions of the Mantel-Haenzsel procedure. J. Amer. Stat. Ass., 58 (1963) 690. 14 Miller, N.E., Hammett. F., Salt&i. S., et al.. Relation of angiographically defined coronary artery disease to plasma lipoprotein sub-fractions and apolipoproteins. Brit. Med. J.. 282 (1981) 1741. 15 Ballantyne, F.C.. Clark, R.S., Simpson. H.S. and Ballantyne. D.. High density and low density lipoprotein subfractions in survivors of myocardial infarction and in control subjects, Metabolism, 31 (1982) 433. 16 Woldow, A., Hyperlipidemia and its significance in the aged population, J. Amer. Geriat. Sot.. 23 (1975) 407. 17 Vermeulen, A.. Rubens. R. and Verdonck, L., Testosterone secretion and metabolism in male senescence. J. Clin. Endocrinol. Metab.. 34 (1972) 730. 18 Pirke. K.M. and Doerr, P., Age-related changes and interrelationships between plasma testosterone, estradiol and testosterone-estradiol binding globulin in normal adult males, Acta Endocrinol.. 74 (1973) 792. 19 Heller. R.F., Miller, N.E.. Lewis, B.. et al.. Associations between sex hormones, thyroid hormones and lipoproteins, Clin. Sci.. 61 (1981) 649. 20 Gutai, J.. LaPorte. R.. Kuller. L.. Dai. W.. Falvo-Gerard. L. and Caggiula, A., Plasma testosterone, high density lipoprotein cholesterol and other lipoprotein fractions, Amer. J. Cardiol., 48 (1981) 897. 21 Anderson, D.W., HDL cholesterol -- The variable components (Letter

to the Editor),

Lancet,

i: (1978) 819.