Low Serum Cholesterol and Mortality
Abstract
Background Many studies have reported an association between a low or lowered blood total cholesterol (TC) level and subsequent nonatherosclerotic disease incidence or death. The question of whether low TC is a true risk factor or alternatively a consequence of occult disease at the time of TC measurement remains unsettled. To shed new light onto this problem, we analyzed TC change over a 6- year period (from exam 1 in 1965 through 1968 to exam 3 in 1971 through 1974) in relation to subsequent 16-year mortality in a cohort of Japanese American men.
Methods and Results The study was based on 5941 men 45 to 68 years of age without prior history of coronary heart disease, stroke, cancer, or gastrointestinal-liver disease at exam 1 who also participated in exam 3 of the Honolulu Heart Program. The association of TC change with mortality end points was investigated with two different approaches (continuous and categorical TC change) with standard survival analysis techniques. Falling TC level was accompanied by a subsequent increased risk of death caused by some cancers (hemopoietic, esophageal, and prostate), noncardiovascular noncancer causes (particularly liver disease), and all causes. The risk-factor–adjusted rate of all-cause mortality was 30% higher (relative risk, 1.30; 95% CI, 1.06 to 1.59) among persons with a decline from middle (180 to 239 mg/dL) to low (<180 mg/dL) TC than in persons remaining at a stable middle level. By contrast, there was no significant increase in all-cause mortality risk among cohort men with stable low TC levels. Nonillness mortality (deaths caused by trauma and suicide) was not related to either TC change or the average of TC levels in exams 1 and 3.
Conclusions These results add strength to the reverse-causality proposition that catabolic diseases cause TC to decrease.
A J- or U-shaped relation between blood total cholesterol (TC) level and all-cause mortality has been reported in several1234567891011121314151617181920 but not in other21222324252627282930313233 studies. The reports of inverse relations have caused some alarm and controversy about the implications of low TC in the general population.34353637
Three interpretations of this association are possible. One view is that a low TC level plays an etiologic role in a variety of nonatherosclerotic diseases (direct causality). Another possibility, however, is that the low TC–mortality relation is attributable to a hypocholesterolemic effect of disease in a preclinical stage (reverse causality). Finally, the association between low TC and mortality may be due to other unmeasured or unknown factors related to both low TC and disease (confounding).38394041 With a few exceptions,48 earlier studies on this topic have not adequately addressed the direct versus reverse causality problem, largely because most statistical analyses were based on a single measurement of TC. Consequently, inferences about temporality could not be made.
Previous studies in the Honolulu Heart Program (HHP) relating baseline TC with subsequent 9-year,7 13-year,8 and 17-year37 mortality demonstrated inverse relations of TC with hemorrhagic stroke, liver disease, chronic obstructive lung disease, and cancers of the esophagus, colon, liver, and hemopoietic system. The relations for cancer and benign liver disease were stronger in the first 5 years and showed flattening when deaths in the first 5 years of follow-up were deleted.
The aim of this study was to extend these preceding reports by assessing the relation of changes in TC level from 1965 through 1968 (exam 1) to 1971 through 1974 (exam 3) with subsequent 16-year mortality. The hypothesis of reverse causality, that low TC is a consequence of disease, would be compatible with a situation in which a declining TC level (resulting from catabolic diseases such as cancer or liver disease), not stable low level, would be related to subsequent mortality. Stated differently, a declining TC level would probably be caused by disease, whereas a stable low TC level would be the result of dietary, lifestyle, or genetic factors.
Methods
Study Population and Procedures
The HHP is a long-term, prospective study of cardiovascular disease among middle-aged men of Japanese ancestry who were residing on Oahu Island, Hawaii, in 1965. Details of the cohort recruitment and study design and procedures were given elsewhere.4243 In brief, a baseline comprehensive examination was carried out during 1965 through 1968 among 8006 study participants. Of those, 6860 men (85% of the original sample) completed the third examination cycle during 1971 through 1974. At both the baseline examination (exam 1) and the reexamination 6 years later (exam 3), detailed information was collected on sociodemographic characteristics, medical history, anthropometric measures, and smoking and alcohol consumption habits. For each subject, blood pressure was recorded three times (twice by a nurse and once by a physician) with a standard sphygmomanometer applied to the left arm of a seated subject, and the average reading was used for the analysis. Body mass index was computed as weight (in kilograms) over height (in meters squared). Laboratory tests included determination of TC, blood glucose 1 hour after a 50-g glucose oral challenge, and uric acid. Nonfasting TC was measured by the Auto Analyzer N24a method at both examinations,44 but the analyses were performed by different laboratories. During 1965 through 1968, frozen sera (−20°C) were shipped in dry ice to the US Public Health Service Heart Disease Control Program Laboratory in San Francisco, Calif; during 1971 through 1974, samples were processed at the Kuakini Medical Center in Honolulu, Hawaii. A physical activity index was calculated as a weighted sum of hours per day spent in five levels of activity, as used in the Framingham Study.45 A dietician made dietary assessments using the 24-hour recall method.46 Measures of diet and physical activity were available only at baseline.
For the present report, mortality follow-up through hospital surveillance, newspaper obituaries, and state health department records covered a 16-year period beginning in 1973 and continuing through the end of 1988 and is believed to be essentially complete. For each fatal event, a medical review panel assigned an underlying cause of death in light of relevant clinical records and coded it according to the International Classification of Diseases, eighth revision. Attrition in this cohort is known to be very small; a follow-up survey in 1984 found that only 1.3% of the men could not be located.8 To avoid the possible effect of disease at the initial examination on risk factor levels or blood lipids, the present analysis excluded men with documented history of coronary heart disease (n=325), stroke (n=113), cancer (n=81), gastrectomy (n=234), colectomy (n=36), hepatic cirrhosis (n=3), or premalignant intestinal disease (n=36) at baseline. Of the initial 8006 men, a total of 758 had one or more of the above medical conditions. Of the 7248 men free of documented prevalent disease, 256 were excluded from analysis because they died before 1973. Of the remaining 6992 men, 739 did not attend exam 3 and thus also were excluded. This left 6253 men, 312 of which had missing values on TC and study confounders. Thus, the final sample for multivariate analysis comprised 5941 men.
Data Analysis
Multiple linear regression was used to identify independent correlates of TC change between exams 1 and 3. Next, the association of changes in TC during a 6-year interval with subsequent 16-year mortality was examined by two different statistical approaches. The first method consisted of modeling mortality rate ratios as a function of continuous TC change adjusted for average TC level. We computed within-individual change in TC as TC in exam 3 minus TC in exam 1; then we used this continuous change as the main covariate of interest. To avoid misleading results that may have arisen by using the TC change alone or by entering initial TC as a covariable, we chose to include the average TC between exams 1 and 3 (but not the initial value). The problem with using the initial value is that it is typically correlated strongly and negatively with the change variable (as a result of regression to the mean), a circumstance that may lead to spurious results. Conversely, the mean TC value, which represents the value half-way between the two examinations, usually shows a very small correlation with the change.4748 The Cox proportional-hazards model4950 was used to estimate age-adjusted associations of categorical and continuous TC change with cause-specific mortality end points. The continuous TC change approach, less constrained by sample size, allowed us to examine individual causes of death. Multivariate analysis was carried out, controlling for potential confounding by baseline body mass index, blood pressure, blood glucose, uric acid, physical activity, percent calories from fat, changes in smoking and alcohol consumption status, and body mass index change between exams 1 and 3. Changes in smoking status between examinations were represented by dummy (dichotomous) variables for past smokers at baseline, quitters between examinations, and continuing smokers relative to men who never smoked. Similarly, changes in alcohol consumption status encompassed dummy variables for continuing abstainers, continuing heavy drinkers (>40 mL ethanol per day, top quintile, at both exams 1 and 3), and heavy drinkers becoming abstainers at exam 3 relative to other drinking groups combined.
To further elucidate the TC change–mortality associations, we also carried out survival analysis with categorical TC change. First, the study population was subdivided into groups according to patterns of TC change from baseline to the second cholesterol measurement 6 years later. At each examination, cholesterol levels in individuals were ranked as low (<180 mg/dL), middle (180 to 239 mg/dL), or high (≥240 mg/dL). Then we compared rates of broad mortality categories (cardiovascular disease, cancer, other causes, and all-cause mortality) across subgroups of the population defined by all possible patterns of TC change. There were three groups of stable TC level (high-high, middle-middle, and low-low), three groups of declining TC level (high-middle, high-low, and middle-low), and three groups of rising TC level (low-middle, low-high, and middle-high). The low cutoff point of 180 mg/dL for the low level (which represented the 13th percentile point of the baseline TC distribution in this population) was used for consistency with earlier investigations of low TC in relation to mortality in this same cohort.7837 The middle range of exam 1 TC (180 to 239 mg/dL) encompassed the zone of minimum death rate in this population,37 whereas the high cholesterol cutoff point of 240 mg/dL for the high level was selected so as to agree with the National Cholesterol Education Program “high-risk” category.51 This method allowed us to assess the two competing hypotheses (direct versus reverse causality) by comparing the associations of declining versus consistently low TC level with mortality outcomes. Relative risks were computed relative to the subgroup maintaining the middle TC level.
Results
Death was documented in 1370 of the 5941 eligible men during 16 years of follow-up (23% crude mortality rate). The number and causes of death were distributed as follows: 557 (41%) to cancer, 385 (28%) to cardiovascular causes, and 428 (31%) to noncardiovascular noncancer causes (see Table 3 for a detailed breakdown of causes of death).
Correlates of Serum Cholesterol Change
TC level fell substantially from exam 1 to exam 3 among participants developing coronary heart disease, gastrointestinal disease, and liver cirrhosis between 1969 and 1975 (Table 1). No important decline in TC was observed for incident cancer and lung disease during the same years. On the other hand, men with incident cerebrovascular disease experienced a modest rise in TC (Table 1).
Seven variables were found to relate significantly and independently to TC change: age (β=−0.23; SEM=0.06), initial TC (β=−0.39; SEM=0.009), body mass index change (β=3.31; SEM=0.26), initial uric acid level (β=−0.08; SEM=0.02), past smoking at exam 1 (β=−1.80; SEM=0.83), liver cirrhosis diagnosed between 1969 and 1975 (β=−14.30; SEM=6.3), and cerebrovascular disease diagnosed between 1969 and 1975 (β=6.65; SEM=3.50). The adjusted R2 of the model was .24, indicating that these seven variables explained about 24% of the variation of TC change. The direction of these associations demonstrated that TC tended to diminish from exam 1 to exam 3 with a decrease in body mass index, older age, higher uric acid concentrations, history of past smoking, and newly diagnosed liver cirrhosis and that it was strongly influenced by initial TC level.
Analysis Using Continuous TC Change
As expected, very little correlation existed between the TC change and average TC (r=−.025; P=.06). Computation of mean TC differences between exams 1 and 3 according to causes of death revealed that almost all causes were associated with a generally modest decline in TC (Table 2), except for large differences in nonmalignant liver disease (−17.5 mg/dL), hemopoietic cancer (−11.6 mg/dL), prostate cancer (−14.6 mg/dL), esophageal cancer (−22.3 mg/dL), rectal cancer (−10.5 mg/dL), and other circulatory deaths (−11.7 mg/dL). For decedents from nonillness mortality (deaths resulting from trauma and suicide), TC change was similar (−1.8 mg/dL) to that for survivors (−1.3 mg/dL). This also was the finding for nonhemorrhagic stroke (−1.7 mg/dL), with this value only slightly larger for decedents from hemorrhagic stroke (−3.0 mg/dL) and for lung cancer (−3.4 mg/dL).
Table 3 gives the multivariate-adjusted relative risks of mortality derived from the analysis with continuous TC change. A reduction of 32 mg/dL in TC (1 SD of TC change) was significantly associated with significant excess risk of hemopoietic, prostate, esophageal, and total cancer and with mortality resulting from nonmalignant liver disease and all causes combined. To compensate for potential confounding as a result of the inclusion of early mortality in the analysis (through TC-lowering effect of occult disease, leading to early death), we also estimated the associations of TC change and mean TC with late mortality (excluding deaths through year 5). After early deaths were excluded, the associations were generally maintained, except for some weakening of the relation between TC change and hemopoietic malignancies.
It should also be pointed out that after TC change and the other risk factors were taken into account and early deaths were excluded, the average TC between exams 1 and 3 (data not shown) was a direct determinant of fatal coronary heart disease (P=.0001) and an inverse correlate of total cancer (P=.02) and fatal hemorrhagic stroke (P=.01). Nonillness mortality (deaths resulting from trauma and suicide) was not related to either TC change or the average of TC levels in HHP exams 1 and 3.
Analysis Using Patterns of TC Change
Table 4 shows the distribution of the cohort by patterns of TC change. Men maintaining a stable middle level constituted by far the largest group (46% of the cohort). The two groups of primary interest—men with stable low TC levels and men with TC changes from middle to low—each represented about 6% of the cohort. Not surprisingly, very few subjects experienced TC changes of a large magnitude, eg, high to low or low to high.
Table 5 gives the results from the multivariate analysis performed with categorical variables for patterns of TC change. Relative to men remaining at a middle TC level, those with consistently high TC levels and those experiencing a rise from a low to a high level had a significantly elevated risk of cardiovascular mortality. However, only one cardiovascular death was observed in the low-to-high TC change category. For cancer mortality, subjects whose TC levels changed from middle to low showed a significantly increased risk. Likewise, a significant elevation of noncardiovascular noncancer mortality and all-cause mortality risk was seen among study subjects displaying a TC decline from a middle to a low level. Other patterns of TC change did not have significant relations with the risk of mortality. Notably, no significant relation was observed between all-cause mortality and a sustained low TC level, although there was some marginal excess risk of cancer and “other” deaths in this group. As Table 5 indicates, the middle-high subgroup had a slightly (nonsignificant) lower all-cause mortality than the middle-middle subgroup.
Discussion
The present report addresses the association of changes in TC during a 6-year period (from examinations done during 1965 through 1968 to examinations done during 1971 through 1974) with subsequent 16-year cause-specific mortality in a cohort of middle-aged Japanese American men. The fundamental question was whether and how strongly stable low TC or declining TC levels were associated with specific mortality outcomes. If only declining levels are associated with death by catabolic diseases (eg, cancer or liver disease), then the low TC–mortality association may be explained by reverse causality: disease-lowering TC. If, on the other hand, a stable low TC level is a significant risk factor for mortality, then the data would support the directionality of low TC to disease.
Observed changes in TC in the HHP cohort are considered to be spontaneous because only 1.7% of the decedents and 1.4% of the cohort survivors in 1988 were receiving lipid-lowering treatment at entry into the study (G. Wergowske, MD, personal communication, January 1995). No attempt was made to exclude these individuals.
Our results suggested that age, initial TC level, change in body mass index, former smoking at exam 1, baseline uric acid levels, and newly diagnosed liver cirrhosis were crucial factors related to TC change. A decrease in TC with increasing age has been reported in several investigations of older populations5253 and may be due to decreased energy expenditure with concomitant reduced energy intake. For this cohort, the average decline was about 3.2 mg/dL per decade after adjustment for lifestyle and biological factors. The observed relation between initial levels and subsequent change is probably the result of the presence of regression to the mean, with initial extreme values converging toward middle values at exam 3. A direct relation between TC and weight was reported in cohort1254 and case-control studies.5556 Moreover, in one case-control study, the authors concluded that a decrease in weight and TC were related to the progression of malignant disease.55
We observed a significant risk of cancer, noncardiovascular noncancer, and all-cause mortality in men whose TC levels changed to low (<180 mg/dL) from the middle (180 to 239 mg/dL) level at baseline. The above finding for cancer mortality agrees with results from the Multiple Risk Factor Intervention Trial,4 in which participants who died of cancer experienced within 1 year of death a fall from baseline in serum cholesterol level that was 22.7 mg/dL (0.59 mmol/L) greater than that of surviving participants of the same randomization group and smoking status. These results also support longitudinal data published by Pekkanen et al57 from the Seven Countries Study in which a decline in TC also tended to be associated with death caused by cancer among the older cohort. A decreasing TC also was associated with elevated overall mortality and cardiovascular mortality in both men and women in the 30-year follow-up of the Framingham Study.58
The current results with TC change as a continuous variable noted a relation between TC decline and all-cause mortality that was largely the result of associations of falling TC with cancers of the hemopoietic system, esophagus, and prostate and with nonmalignant liver disease. The association between hemopoietic malignancies and low TC was reported in several large prospective studies.4913192228 Alterations in plasma lipids and lipoprotein fractions were demonstrated in patients with acute leukemia and non-Hodgkin’s lymphoma that were related to the degree of underlying tumor burden and to the presence of bone marrow involvement.59 Several studies clearly demonstrated that high receptor-mediated uptake and degradation of LDL by the leukemic cells may cause hypocholesterolemia in acute myelogenous leukemia.606162 Moreover, Knekt et al17 noted a significant inverse relation between TC and prostate cancer that persisted after exclusion of the first 4 years of follow-up. An association between low TC and prostate cancer was found in three other cohorts.286364 Recent work in patients with metastatic carcinoma of the prostate showed an enhanced fractional elimination rate of plasma LDL associated with the tumor, which explained the low LDL levels, a situation similar to that in patients with myeloproliferative disorders.65
Deaths caused by nonmalignant liver disease in the HHP cohort consisted almost exclusively of alcoholic cirrhosis of the liver, a condition known to alter lipid metabolism, causing a decrease in lipoprotein production.66
Interestingly, coronary heart disease mortality was unrelated to cholesterol change but had a direct significant association with the mean TC level. This is in agreement with significant associations of serum cholesterol with the 10-year incidence of coronary heart disease,67 sudden cardiac death within 18 years,68 and autopsy-determined atherosclerosis in the coronaries and aortas in this same cohort.69
Mean cholesterol was an inverse determinant of fatal hemorrhagic stroke. Very low serum cholesterol concentrations have been hypothesized to be a causative factor in the development of intracranial aneurysms, lesions that are liable to rupture and cause intracerebral bleeding. The proposed mechanism would involve a weakening effect of the vascular architecture by inadequate cholesterol content in the membranes of the endothelial cells of small intracranial vessels.7071 Contrary to this notion, however, persons with hypobetalipoproteinemia and children (who sustain very low TC levels for years) do not have hemorrhagic strokes. In any case, the absolute number of hemorrhagic stroke deaths in persons with very low TC (<160 mg/dL) was very small (n=5), so this association may not be important from a public health perspective.
The inverse relation between TC change and hemopoietic cancer may be explained by the cholesterol-lowering activity of neoplastic cells in the asymptomatic stage. Further evidence for reverse causality is provided by studies showing that low TC reverts to usual concentrations with successful chemotherapy and remission of disease.72 Moreover, the association between cancers of the hemopoietic system and TC decrease as a continuous variable diminished substantially when early deaths were eliminated from the analysis. On the other hand, the association of a decrease in TC and cancers of the prostate, esophagus, and all sites combined did not weaken when the first 5 years were excluded. It has been proposed that this long-term association may be due to prolongation of survival by treatment.73
The observed slope of TC was fitted across only two measures, either of which could be atypical for the subject. Thus, it is important to remember that the observed associations between TC change and mortality could be biased by measurement error. In the case of multiple exposure groups (as is the case in the analysis with categorical TC change), the direction of bias is less predictable and depends on not only the misclassification rates but also the distribution of subjects across exposure levels.74 In the case of continuous TC change with adjustment for multiple confounders, measurement error in TC change could attenuate, inflate, or leave the coefficient magnitude unaltered.75
The key finding of this report is that spontaneously falling TC levels were associated with increased risk of nonmalignant liver disease, total cancer, and most noticeably, cancers of the esophagus, prostate, and hemopoietic systems. By contrast, after confounders were controlled for, a stable low TC level was not associated with significantly increased mortality risk, although some marginal risk existed owing to an association of very low TC with fatal hemorrhagic stroke.
In conclusion, the results of this longitudinal analysis, together with existing coherent biological plausibility of reverse causality, support the hypothesis that low TC, independent of the risk factors considered in this study, appears to be a manifestation of tumor activity or the consequence of chronic liver disease. Future research should concentrate on the temporal association of serial TC measurements with disease and the potential etiologic role of chronically low TC in the pathogenesis of intracerebral hemorrhage.
Reprint requests to James H. Dwyer, PhD, Institute for Prevention Research, USC School of Medicine, 1540 Alcazar St, CHP 205, Los Angeles, CA 90033.
| Prevalent Medical Condition (1969 Through 1975) | No. | Serum Cholesterol, mg/dL | ||
|---|---|---|---|---|
| At Exam 1 (1965 Through 1968) | At Exam 3 (1971 Through 1974) | Mean Difference1 | ||
| Coronary heart disease | 180 | 229 (39) | 219 (38) | −10.1 (34.8) |
| Cerebrovascular disease | 54 | 216 (34) | 220 (39) | 4.1 (35.6) |
| Cancer | 89 | 219 (36) | 217 (40) | −2.1 (30.7) |
| Gastrointestinal disease2 | 150 | 216 (37) | 208 (33) | −7.8 (34.3) |
| Liver cirrhosis | 17 | 239 (48) | 212 (48) | −27.5 (65.9) |
| Lung disease3 | 407 | 215 (38) | 213 (35) | −1.8 (33.2) |
| Any medical condition4 | 899 | 219 (38) | 213 (35) | −4.6 (34.8) |
| Cause of Death (ICD-8) | No. of Deaths (%) | Serum Cholesterol, mg/dL | ||
|---|---|---|---|---|
| At Exam 1 (1965 Through 1968) | At Exam 3 (1971 Through 1974) | Mean Difference1 | ||
| Coronary heart disease (410-414) | 190 (13.8) | 231 (38) | 226 (39) | −4.7 (36.8) |
| Hemorrhagic stroke (430-431) | 46 (3.3) | 206 (41) | 203 (41) | −3.0 (37.8) |
| Nonhemorrhagic stroke (432-438) | 77 (5.6) | 221 (44) | 220 (35) | −1.7 (34.0) |
| Other circulatory (394-404; 421-429; 440-458) | 72 (5.2) | 220 (36) | 208 (32) | −11.7 (30.5) |
| All cardiovascular (394-458) | 385 (28.1) | 224 (40) | 219 (38) | −5.2 (35.3) |
| Lung cancer (162) | 148 (10.8) | 215 (36) | 211 (36) | −3.4 (32.9) |
| Stomach cancer (151) | 103 (7.5) | 218 (35) | 210 (34) | −7.1 (28.2) |
| Colon cancer (153) | 53 (3.8) | 209 (36) | 210 (31) | 1.3 (25.1) |
| Rectal cancer (154) | 24 (1.7) | 223 (30) | 212 (32) | −10.5 (27.6) |
| Hemopoietic cancer (200-207) | 42 (3.0) | 213 (45) | 202 (40) | −11.6 (41.4) |
| Prostate cancer (185) | 34 (2.4) | 221 (38) | 207 (39) | −14.6 (41.6) |
| Esophageal cancer (150) | 21 (1.5) | 218 (52) | 196 (33) | −22.3 (52.4) |
| Other cancers | 131 (9.5) | 215 (34) | 211 (36) | −3.5 (30.3) |
| Total cancer (141-207) | 557 (40.6) | 215 (37) | 209 (36) | −5.6 (33.1) |
| COPD (491-492) | 47 (3.4) | 214 (38) | 209 (29) | −4.3 (29.7) |
| Nonmalignant liver disease (570-573) | 23 (1.6) | 219 (34) | 202 (35) | −17.5 (33.1) |
| Nonillness mortality (800-957) | 52 (3.7) | 221 (41) | 219 (35) | −1.8 (28.2) |
| Miscellaneous and unknown | 306 (22.3) | 217 (40) | 213 (39) | −4.3 (34.6) |
| All causes | 1370 (100) | 218 (39) | 213 (37) | −5.2 (33.8) |
| Cohort survivors in 19882 | … | 218 (36) | 217 (35) | −1.3 (31.3) |
| Cause of Death (ICD-8) | Complete Follow-up (1973 Through 1988) | Excluding the First 5 Years of Follow-up (1978 Through 1988) | ||
|---|---|---|---|---|
| No. of Deaths | Risk Factor–Adjusted RR (95% CI) | No. of Deaths | Risk Factor–Adjusted RR (95% CI) | |
| Coronary heart disease (410-414) | 190 | 0.98 (0.86, 1.12) | 153 | 0.95 (0.82, 1.10) |
| Hemorrhagic stroke (430-431) | 46 | 1.02 (0.75, 1.39) | 32 | 1.12 (0.77, 1.64) |
| Nonhemorrhagic stroke (432-438) | 77 | 0.95 (0.76, 1.20) | 68 | 0.94 (0.73, 1.21) |
| Other circulatory (394-404, 421-429, 440-458) | 72 | 1.22 (0.97, 1.53) | 58 | 1.08 (0.83, 1.42) |
| All cardiovascular (394-458) | 385 | 1.03 (0.93, 1.14) | 311 | 0.99 (0.89, 1.11) |
| Lung cancer (162) | 148 | 1.11 (0.93, 1.32) | 130 | 1.13 (0.94, 1.36) |
| Stomach cancer (151) | 103 | 1.20 (0.99, 1.46) | 87 | 1.17 (0.95, 1.46) |
| Colon cancer (153) | 53 | 0.84 (0.64, 1.12) | 43 | 0.91 (0.67, 1.24) |
| Rectal cancer (154) | 24 | 1.14 (0.81, 1.59) | 16 | 1.16 (0.76, 1.75) |
| Hemopoietic cancer (200-207) | 42 | 1.40 (1.05, 1.87) | 33 | 1.28 (0.92, 1.79) |
| Prostate cancer (185) | 34 | 1.38 (0.99, 1.93) | 31 | 1.48 (1.04, 2.10) |
| Esophageal cancer (150) | 21 | 1.88 (1.30, 2.72) | 16 | 1.96 (1.31, 2.91) |
| Other cancers | 131 | 1.05 (0.89, 1.24) | 120 | 1.04 (0.86, 1.26) |
| Total cancer (141-207) | 557 | 1.14 (1.05, 1.24) | 466 | 1.14 (1.04, 1.26) |
| COPD (491-492) | 47 | 1.02 (0.74, 1.42) | 39 | 1.08 (0.76, 1.55) |
| Nonmalignant liver disease (570-573) | 23 | 1.56 (1.10, 2.21) | 20 | 1.59 (1.08, 2.35) |
| Nonillness mortality (800-957) | 52 | 0.94 (0.71, 1.25) | 32 | 0.93 (0.64, 1.35) |
| Miscellaneous and unknown | 306 | 1.01 (0.89, 1.13) | 266 | 1.04 (0.92, 1.18) |
| Noncardiovascular* and noncancer combined | 428 | 1.03 (0.93, 1.14) | 357 | 1.06 (0.95, 1.18) |
| All causes | 1370 | 1.07 (1.02, 1.13) | 1134 | 1.07 (1.01, 1.14) |
| Patterns of Cholesterol Change1 | Serum Cholesterol, mg/dL | ||||
|---|---|---|---|---|---|
| No. Men | At Exam 1 (1965 Through 1968) | At Exam 3 (1971 Through 1974) | Mean Difference2 | ||
| n | % | ||||
| Stable | |||||
| Low/low | 376 | 6.3 | 158 (16) | 162 (15) | 3.4 (17.7) |
| Middle/middle | 2740 | 46.1 | 210 (15) | 208 (16) | −2.6 (19.0) |
| High/high | 855 | 14.4 | 269 (27) | 269 (24) | −0.9 (27.6) |
| Declining | |||||
| High/middle | 608 | 10.2 | 259 (20) | 219 (15) | −39.6 (25.8) |
| Middle/low | 386 | 6.5 | 201 (14) | 166 (14) | −35.4 (21.0) |
| High/low | 32 | 0.5 | 277 (38) | 162 (17) | −115.3 (42.5) |
| Rising | |||||
| Low/middle | 380 | 6.4 | 166 (12) | 197 (15) | 30.2 (19.6) |
| Middle/high | 547 | 9.2 | 220 (14) | 259 (17) | 39.0 (22.7) |
| Low/high | 17 | 0.3 | 169 (13) | 265 (26) | 95.4 (25.1) |
| Pattern of Serum Cholesterol Change1 | Cardiovascular (n=385) | Cancer (n=557) | Other (n=428) | Total (n=1370) | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| n | RR2 | 95% CI | n | RR | 95% CI | n | RR | 95% CI | n | RR | 95% CI | |
| Stable | ||||||||||||
| Low/low | 45 | 0.79 | 0.49, 1.26 | 20 | 1.18 | 0.85, 1.63 | 33 | 1.17 | 0.80, 1.71 | 98 | 1.07 | 0.86, 1.33 |
| Middle/middle (reference) | 248 | 1.00 | 168 | 1.00 | 176 | 1.00 | 592 | 1.00 | ||||
| High/high | 69 | 1.39 | 1.05, 1.84 | 71 | 0.97 | 0.74, 1.27 | 61 | 1.16 | 0.87, 1.56 | 201 | 1.15 | 0.98, 1.35 |
| Declining | ||||||||||||
| High/middle | 60 | 1.20 | 0.86, 1.68 | 45 | 1.19 | 0.89, 1.58 | 43 | 1.11 | 0.79, 1.55 | 148 | 1.16 | 0.97, 1.40 |
| Middle/low | 48 | 0.81 | 0.51, 1.28 | 21 | 1.37 | 1.00, 1.88 | 48 | 1.65 | 1.19, 2.29 | 117 | 1.30 | 1.06, 1.59 |
| High/low | 5 | 1.86 | 0.67, 5.17 | 4 | 2.09 | 0.85, 5.14 | 2 | 1.18 | 0.28, 4.83 | 11 | 1.76 | 0.96, 3.22 |
| Rising | ||||||||||||
| Low/middle | 42 | 1.02 | 0.64, 1.61 | 21 | 1.14 | 0.82, 1.59 | 31 | 1.30 | 0.88, 1.93 | 94 | 1.16 | 0.93, 1.44 |
| Middle/high | 39 | 1.03 | 0.70, 1.51 | 32 | 0.80 | 0.57, 1.12 | 33 | 1.02 | 0.70, 1.48 | 104 | 0.92 | 0.75, 1.14 |
| Low/high | 1 | 4.57 | 1.43, 14.5 | 3 | 0.67 | 0.09, 4.81 | 1 | 1.41 | 0.19, 10.1 | 5 | 1.77 | 0.73, 4.29 |
Presented at the Society for Epidemiologic Research 27th Annual Meeting, Miami, Fla, June 15-18, 1994.
The Honolulu Heart Program is supported by research contracts (NO1-HC-02901 and NO1-HV-02901) from the NHLBI, NIH, Bethesda, Md. This work was part of Dr Iribarren’s doctoral dissertation at the University of Southern California School of Medicine, Los Angeles.
References
- 1 Sorlie PD, Feinleib M. The serum cholesterol-cancer relationship: an analysis of time trends in the Framingham study. J Natl Cancer Inst.1982; 69:989-996. MedlineGoogle Scholar
- 2 Garcia-Palmieri MR, Sorlie PD, Costas R, Havlik RJ. An apparent inverse relationship between serum cholesterol and cancer mortality in Puerto Rico. Am J Epidemiol.1981; 114:29-40. CrossrefMedlineGoogle Scholar
- 3 Kozarevic D, McGee D, Vojvodic N, Gordon T, Racic Z, Zukel W, Dawber T. Serum cholesterol and mortality: the Yugoslavia Cardiovascular Disease Study. Am J Epidemiol.1981; 114:21-28. CrossrefMedlineGoogle Scholar
- 4 Sherwin RW, Wentworth DN, Cutler JA, Hulley SB, Kuller LH, Stamler J. Serum cholesterol levels and cancer mortality in 361,622 men screened for the Multiple Risk Factor Intervention Trial. JAMA.1987; 257:943-948. CrossrefMedlineGoogle Scholar
- 5 Salmond CE, Beaglehole R, Prior IAM. Are low cholesterol values associated with excess mortality? BMJ.1985; 290:422-424. CrossrefMedlineGoogle Scholar
- 6 Isles CG, Hole DJ, Gillis CR, Hawthorne VM, Lever AF. Plasma cholesterol, coronary heart disease, and cancer in the Renfrew and Paisley Survey. BMJ.1989; 298:920-924. CrossrefMedlineGoogle Scholar
- 7 Kagan A, McGee DL, Yano K, Rhoads GG, Nomura A. Serum cholesterol and mortality in a Japanese-American population: the Honolulu Heart Program. Am J Epidemiol.1981; 114:11-20. CrossrefMedlineGoogle Scholar
- 8 Stemmermann GN, Chyou PH, Kagan A, Nomura AM, Yano K. Serum cholesterol and mortality among Japanese-American men: the Honolulu (Hawaii) Heart Program. Arch Intern Med.1991; 151:969-972. CrossrefMedlineGoogle Scholar
- 9 Cambien F, Ducimetre P, Richard J. Total serum cholesterol and cancer mortality in a middle-aged male population. Am J Epidemiol.1980; 112:388-394. CrossrefMedlineGoogle Scholar
- 10 Kark JD, Smith AH, Hames CG. The relationship of serum cholesterol to the incidence of cancer in Evans County, Georgia. J Chronic Dis.1979; 33:311-322. Google Scholar
- 11 Keys A, Aravanis C, Blackburn H, Buzina R, Dontas AS, Fidanza F, Karvonen MJ, Menotti A, Nedeljkovic S, Punsar S, Toshima H. Serum cholesterol and cancer mortality in the Seven Countries Study. Am J Epidemiol.1985; 121:870-883. CrossrefMedlineGoogle Scholar
- 12 Davey Smith G, Shipley MJ, Marmot MG, Rose G. Plasma cholesterol concentration and mortality: the Whitehall Study. JAMA.1992; 267:70-76. CrossrefMedlineGoogle Scholar
- 13 Schatzkin AS, Hoover RN, Taylor PR, Ziegler RG, Carter CL, Albanes D, Larson DB, Licitra LM. Site-specific analysis of total serum cholesterol and incident cancer in the National Health and Nutrition Examination Survey I Epidemiologic Follow-up Study. Cancer Res.1988; 48:452-458. MedlineGoogle Scholar
- 14 Morris DL, Borhani NO, Fitzsimons E, Hardy RJ, Hawkins CM, Kraus JF, LaBarthe DR, Mastbaum L, Payne GH. Serum cholesterol and cancer in the Hypertension Detection and Follow-up Program. Cancer.1983; 52:1754-1759. CrossrefMedlineGoogle Scholar
- 15 Cowan LD, O’Connell DL, Criqui MH, Barrett-Connor E, Bush TL, Wallace RB. Cancer mortality and lipid and lipoprotein levels: the Lipid Research Clinics Program Mortality Follow-up Study. Am J Epidemiol.1990; 131:468-482. CrossrefMedlineGoogle Scholar
- 16 International Collaborative Group. Circulating cholesterol level and risk of death from cancer in men aged 40 to 69 years: experience of an International Collaborative Group. JAMA.1982; 248:2853-2859. CrossrefMedlineGoogle Scholar
- 17 Knekt P, Reunanen A, Aromaa A, Heliövaara M, Hakulinen T, Hakama M. Serum cholesterol and risk of cancer in a cohort of 39,000 men and women. J Clin Epidemiol.1988; 41:519-530. CrossrefMedlineGoogle Scholar
- 18 Törnberg SA, Holm LE, Carstensen JM, Eklund GA. Cancer incidence and cancer mortality in relation to serum cholesterol. J Natl Cancer Inst.1989; 81:1917-1921. CrossrefMedlineGoogle Scholar
- 19 Neaton JD, Blackburn H, Jacobs D, Kuller L, Lee DJ, Sherwin R, Shih J, Stamler J, Wentworth D. Serum cholesterol level and mortality findings for men screened in the Multiple Risk Factor Intervention Trial. Arch Intern Med.1992; 152:1490-1500. CrossrefMedlineGoogle Scholar
- 20 Schuit AJ, Van Dijk CE, Dekker JM, Schouten EG, Kok FJ. Inverse association between serum total cholesterol and cancer mortality in Dutch civil servants. Am J Epidemiol.1993; 137:966-976. CrossrefMedlineGoogle Scholar
- 21 Iso H, Naito Y, Kitamura A, Sato S, Kiyama M, Takayama Y, Iida M, Shimamoto T, Sankai T, Komachi Y. Serum total cholesterol and mortality in a Japanese population. J Clin Epidemiol.1994; 47:961-969. CrossrefMedlineGoogle Scholar
- 22 Dyer AR, Stamler J, Paul O, Shekelle RB, Schoenberger JA, Berkson DM, Lepper M, Collette P, Shekelle S, Lindberg HA. Serum cholesterol and risk of death from cancer and other causes in three Chicago epidemiological studies. J Chronic Dis.1981; 34:249-260. CrossrefMedlineGoogle Scholar
- 23 Thomas CB, Duszynski KR, Shaffer J. Cholesterol levels in young adulthood and subsequent cancer: a preliminary note. Johns Hopkins Med J.1982; 150:89-94. MedlineGoogle Scholar
- 24 Kromhout D, Bosschieter EB, Drijver M, Coulander CL. Serum cholesterol and 25-year incidence of and mortality from myocardial infarction and cancer. Arch Intern Med.1988; 148:1051-1055. CrossrefMedlineGoogle Scholar
- 25 Westlund K, Nicolaysen R. Ten-year mortality and morbidity related to serum cholesterol. Scand J Clin Lab Invest Suppl. 1972;30(suppl 127):2-24. Google Scholar
- 26 Shaper AG, Phillips AN, Pocock SJ. Plasma cholesterol, coronary heart disease and cancer. BMJ.1989; 298:1381. Google Scholar
- 27 Salonen JT. Risk of cancer and death in relation to serum cholesterol: a longitudinal study in an eastern Finnish population with high overall cholesterol levels. Am J Epidemiol.1982; 116:622-630. CrossrefMedlineGoogle Scholar
- 28 Hiatt RA, Fireman BH. Serum cholesterol and the incidence of cancer in a large cohort. J Chronic Dis.1986; 39:861-870. CrossrefMedlineGoogle Scholar
- 29 Yaari S, Goldbourt U, Even-Zohar S, Neufeld HN. Associations of serum high density lipoprotein and total cholesterol with total, cardiovascular, and cancer mortality in a 7-year prospective study of 10,000 men. Lancet.1981; 1:1011-1015. CrossrefMedlineGoogle Scholar
- 30 Baptiste MS, Nasca PC, Doyle JT, Rothenberg RR, MacCubbin PA, Mettlin C, Metzger BB, Carlton KA. Cholesterol and cancer in a population of male civil service workers. Int J Epidemiol.1992; 21:16-21. CrossrefMedlineGoogle Scholar
- 31 Peto R, Boreham J, Chen J, Li J, Campbell TC, Brun T. Plasma cholesterol, coronary heart disease and cancer. BMJ.1989; 298:1249. Google Scholar
- 32 Wald NJ, Thompson SG, Law MR, Densem JW, Bailey A. Serum cholesterol and subsequent risk of cancer: results from the BUPA study. Br J Cancer.1989; 59:936. CrossrefMedlineGoogle Scholar
- 33 Klag MJ, Ford DE, Mead LA, He J, Whelton PK, Liang KY, Levine DM. Serum cholesterol in young men and subsequent cardiovascular disease. N Engl J Med.1993; 328:313-318. CrossrefMedlineGoogle Scholar
- 34 Heady JA, Morris JN, Oliver MF. WHO clofibrate/cholesterol trial: clarifications. Lancet.1992; 340:1405-1406. Letter. CrossrefMedlineGoogle Scholar
- 35 Jacobs DR Jr, Blackburn H, Higgins M, Reed D, Iso H, McMillan G, Neaton J, Nelson J, Potter J, Rifkind B, Rossouw J, Shekelle R, Yusuf S. Report of the Conference on Low Blood Cholesterol: mortality associations. Circulation.1992; 86:1046-1060. CrossrefMedlineGoogle Scholar
- 36 Hulley SB, Walsh JMB, Newman TB. Health policy on blood cholesterol: time to change directions. Circulation.1992; 86:1026-1029. Editorial. CrossrefMedlineGoogle Scholar
- 37 Frank JW, Reed DM, Grove JS, Benfante R. Will lowering population levels of serum cholesterol affect total mortality? Expectations from the Honolulu Heart Program. J Clin Epidemiol.1992; 45:333-346. CrossrefMedlineGoogle Scholar
- 38 Stamler J, Stamler R, Brown V, Gotto AM, Greenland P, Grundy S, Hegsted M, Luepker RV, Neaton JD, Steinberg D, Stone N, Van Horn L, Wissler RW. Serum cholesterol: doing the right thing. Circulation.1993; 88:1954-1960. Editorial. CrossrefMedlineGoogle Scholar
- 39 Iribarren C, Reed DM, Burchfiel CM, Dwyer JH. Serum total cholesterol and mortality: confounding factors and risk modification in Japanese-American men. JAMA.1995; 273:1926-1932. CrossrefMedlineGoogle Scholar
- 40 Manolio TA, Ettinger WH, Tracy RP, Kuller LH, Borhani NO, Lynch JC, Fried LP. Epidemiology of low cholesterol in older adults: the Cardiovascular Health Study. Circulation.1993; 87:728-737. CrossrefMedlineGoogle Scholar
- 41 Law MR, Thompson SG, Wald NJ. Assessing possible hazards of reducing serum cholesterol. BMJ.1994; 308:373-379. CrossrefMedlineGoogle Scholar
- 42 Worth RM, Kagan A. Ascertainment of men of Japanese ancestry in Hawaii through World War II selective service registration. J Chronic Dis.1970; 23:389-397. CrossrefMedlineGoogle Scholar
- 43 Belsky JL, Kagan A, Syme DL. Epidemiologic studies of coronary heart disease and stroke in Japanese men living in Japan, Hawaii and California: research plan. Hiroshima, Japan; 1971. Atomic Bomb Casualty Commission technical report; 12-71. Google Scholar
- 44 Technicon Instruments Corporation. Technicon Auto-Analyzer Methodology. Tarrytown, NY: Technicon Corp; 1965. Google Scholar
- 45 Kannel WB, Sorlie P. Some benefits of physical activity: the Framingham Study. Arch Intern Med.1979; 129:857-861. CrossrefGoogle Scholar
- 46 Tillotson JT, Kato H, Nichaman MZ, Miller DC, Gay ML, Johnson KG. Epidemiology of coronary heart disease and stroke in Japanese men living in Japan, Hawaii and California: methodology for comparison of diet. Am J Clin Nutr.1973; 26:177-184. CrossrefMedlineGoogle Scholar
- 47 Cain KC, Kronmal RA, Kosinski AS. Analyzing the relationship between change in a risk factor and risk of disease. Stat Med.1992; 11:783-797. CrossrefMedlineGoogle Scholar
- 48 Oldham PD. A note on the analysis of repeated measurements on the same subject. J Chronic Dis.1962; 15:969-977. CrossrefMedlineGoogle Scholar
- 49 Cox DR. Regression models and life tables (with discussion). J R Stat Soc (B).1972; 34:187-220. Google Scholar
- 50 SAS Institute Inc. The PHREG procedure. In: SSAS/STAT Software Version 6.04. Cary, NC: SAS Institute Inc; 1987. Google Scholar
- 51 National Cholesterol Education Program. Second Report of the Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel II). Bethesda, Md: NHLBI; 1993. NIH publication 93-3095. Google Scholar
- 52 Hershcopf RJ, Elahi D, Andres R, Baldwin HL, Raizes GS, Schocken DD, Tobin JD. Longitudinal changes in serum cholesterol in man: an epidemiologic search for an etiology. J Chronic Dis.1982; 35:101-114. CrossrefMedlineGoogle Scholar
- 53 Keys A, Mickelsen O, Miller EO, Hayes ER, Todd RL. The concentration of cholesterol in the serum of normal men and its relation to age. J Clin Invest.1950; 29:1347-1353. CrossrefMedlineGoogle Scholar
- 54 Kritchevski SB, Wilcosky TC, Morris DL, Truong KN, Tyroler HA. Changes in plasma lipid and lipoprotein cholesterol and weight prior to the diagnosis of cancer. Cancer Res.1991; 51:3198-3203. MedlineGoogle Scholar
- 55 Wallace RB, Rosl C, Burmeister LF. Cancer incidence in humans: relationship to plasma lipids and relative weight. J Natl Cancer Inst.1982; 68:915-918. MedlineGoogle Scholar
- 56 Rudman D, Mattson DE, Hoskote S, Nagraj S, Caindec N, Rudman IW, Jackson DL. Antecedents of death in the men of a Veterans Administration nursing home. J Am Geriatr Soc.1987; 35:496-502. CrossrefMedlineGoogle Scholar
- 57 Pekkanen J, Nissinen A, Vartiainen E, Salonen JT, Punsar S, Karvonen MJ. Changes in serum cholesterol level and mortality: a 30-year follow-up: the Finnish Cohorts of the Seven Countries Study. Am J Epidemiol.1994; 139:155-165. CrossrefMedlineGoogle Scholar
- 58 Anderson KM, Castelli WP, Levy D. Cholesterol and mortality: 30 years of follow-up from the Framingham Study. JAMA.1987; 257:2176-2180. CrossrefMedlineGoogle Scholar
- 59 Spielberg RJ, Schaeffer EJ, Magrath IT, Edwards BK. Plasma lipid alterations in leukemia and lymphoma. Am J Med.1982; 72:775-782. CrossrefMedlineGoogle Scholar
- 60 Budd D, Ginsberg H. Hypocholesterolemia and acute myelogenous leukemia: association between activity and plasma low-density lipoprotein cholesterol concentrations. Cancer.1986; 58:1361-1365. CrossrefMedlineGoogle Scholar
- 61 Vitols S, Björkholm M, Gahrton G, Peterson C. Hypocholesterolemia in malignancy due to elevated low-density-lipoprotein-receptor activity in tumour cells: evidence from studies in patients with leukaemia. Lancet.1985; 2:1150-1153. CrossrefMedlineGoogle Scholar
- 62 Ho YK, Smith RG, Brown MS, Goldstein JL. Low-density lipoprotein (LDL) receptor activity in human acute myelogenous leukemia cells. Blood.1978; 52:1099-1114. MedlineGoogle Scholar
- 63 Wingard DL, Criqui MH, Holdbook MJ, Barret-Connor E. Plasma cholesterol and cancer morbidity and mortality in an adult community. J Chronic Dis.1984; 37:401-406. CrossrefMedlineGoogle Scholar
- 64 Morris DL, Borhani NO, Fitzsimons E, Hardy RJ, Hawkins CM, Kraus JF, LaBarthe DR, Mastbaum L, Payne GH. Serum cholesterol and cancer in the Hypertension Detection and Follow-up Program. Cancer. 1983;52;1754-1759. Google Scholar
- 65 Henriksson P, Ericsson S, Stege R, Eriksson M, Rudling M, Berlund L, Angelin B. Hypocholesterolemia and increased elimination of low-density lipoproteins in metastatic cancer of the prostate. Lancet.1989; 2:1178-1180. MedlineGoogle Scholar
- 66 Reitz RC. The effects of ethanol ingestion on lipid metabolism. Prog Lipid Res.1979; 18:87-115. CrossrefMedlineGoogle Scholar
- 67 Yano K, Reed DM, McGee D. Ten-year incidence of coronary heart disease in the Honolulu Heart Program: relationship to biologic and lifestyle characteristics. Am J Epidemiol.1984; 119:653-666. CrossrefMedlineGoogle Scholar
- 68 Kagan A, Yano K, Reed DM, McLean C. Predictors of sudden death among Hawaiian Japanese men. Am J Epidemiol.1989; 130:268-277. CrossrefMedlineGoogle Scholar
- 69 Reed DM, McLean C, Hayashi T. Predictors of atherosclerosis in the Honolulu Heart Program, I: biologic, dietary and lifestyle characteristics. Am J Epidemiol.1987; 126:214-225. CrossrefMedlineGoogle Scholar
- 70 Iso H, Jacobs DR, Wentworth D, Neaton JD, Cohen JD. Serum cholesterol levels and six-year mortality from stroke in 350,977 men screened for the Multiple Risk Factor Intervention Trial. N Engl J Med.1989; 320:904-910. CrossrefMedlineGoogle Scholar
- 71 Yano K, Reed DM, MacLean CJ. Serum cholesterol and hemorrhagic stroke in the Honolulu Heart Program. Stroke.1989; 20:1460-1465. CrossrefMedlineGoogle Scholar
- 72 Venkatanarayanan S, Nagarajan B. Association between tumor status and serum lipoprotein cholesterol in hemopoietic malignancy. Biochem Int.1988; 17:499-507. MedlineGoogle Scholar
- 73 Law MR, Thompson SG. Low serum cholesterol and the risk of cancer: an analysis of the published prospective studies. Cancer Causes Control.1991; 2:253-261. CrossrefMedlineGoogle Scholar
- 74 Birkett NJ. Effect of nondifferential misclassification on estimates of odds ratios with multiple levels of exposure. Am J Epidemiol.1992; 136:356-366. CrossrefMedlineGoogle Scholar
- 75 Thomas D, Stram D, Dwyer JH. Exposure measurement error: influence on exposure-disease relationships and methods of correction. Annu Rev Publ Health.1993; 14:69-93.CrossrefMedlineGoogle Scholar
