Background:

The Beta-Carotene and Retinol Efficacy Trial (CARET) was terminated 21 months ahead of schedule due to an excess of lung cancers. Deaths from cardiovascular disease also increased (relative risk=1.26 (95% confidence interval (CI) 0.99–1.61)) in the group assigned to a combination of 30 mg beta-carotene and 25 000 IU retinyl palmitate (vitamin A) daily. The basis for increased cardiovascular mortality is unexplained.

Design:

We analyzed data on serum lipids, available for 1474 CARET Vanguard participants who were enrolled in the two CARET pilot studies and transitioned to the Vanguard study. Total cholesterol and triglycerides were measured 2 months prior to, 4 and 12 months following randomization, and annually thereafter for up to 7 y.

Intervention:

In the asbestos-exposed pilot (N=816), participants were assigned to beta-carotene and retinol or to placebo; in the smokers pilot (N=1029), participants were assigned to beta-carotene, retinol, a combination, or placebo.

Results:

Serum cholesterol showed a decline over time in both arms; serum triglycerides had a continuous decline over time in the placebo arm, but an initial increase that persisted in the active arm. Both serum cholesterol concentrations (P<0.0003) and serum triglycerides (P<0.0001) were significantly higher in the participants receiving vitamin A and/or a combination of vitamin A and beta-carotene (n=863) as compared to the placebo group (n=611). Those in this active intervention group had an average cholesterol concentration 5.3 mg/dl (0.137 mmol/l) higher than those in the placebo arm.

Conclusion:

The differences in cholesterol and triglyceride concentrations between the groups following randomization may account in part for the unexpected excess in cardiovascular deaths seen in the active intervention arm of CARET.

Sponsorship:

Supported by grants U01 CA63673, U01 CA63674, U01 CA47989, U01 CA48200, U01 CA48203, U01 CA48196, and U01 CA52596 from the National Cancer Institute, USA.

Introduction

The Beta-Carotene and Retinol Efficacy Trial (CARET) is a multicenter randomized lung cancer chemoprevention trial, initiated with pilot studies in smoker (N=1029) and asbestos-exposed (N=816) populations in 1985, and expanded 10-fold (N=18 314) in 1989 (Goodman et al, 1993; Thornquist et al, 1993; Omenn et al, 1993, 1994b). In January 1996, the intervention was terminated 21 months ahead of schedule due to the unexpected findings that the active treatment group, assigned to take 30 mg beta-carotene and 25 000 IU retinyl palmitate (vitamin A) daily, had 28% higher lung cancer incidence (95% confidence interval (CI), 1.04–1.57), 46% higher mortality rate from lung cancer (95% CI, 1.07–2.00), and 17% increased overall mortality rate compared to the placebo group (95% CI, 1.03–1.33) (Omenn et al, 1996). Deaths from cardiovascular disease (ICD categories: 390–459 798) were increased by 26% (95% CI, 0.99–1.61) in the active treatment group compared with the placebo group. The increased mortality from cardiovascular disease is unexplained.

Synthetic retinoids have been well documented to affect the lipid profile, causing elevations in both triglycerides and cholesterol (Zech et al, 1983; Bershad et al, 1985; Vahlquist et al, 1985; Marsden, 1986; Tangrea et al, 1993). Data for high dose vitamin A (300 000 IU per day) show a similar effect on both triglycerides and cholesterol (Murray et al, 1983; Infante et al, 1991; Pastorino et al, 1991). However, data for moderate dose vitamin A are scant. One of the authors has reported that the group receiving 25 000 IU retinol per day in a large randomized skin cancer chemoprevention trial had significantly higher cholesterol and triglycerides concentrations compared to the placebo group (Cartmel et al, 1999).

In contrast to vitamin A, several studies have reported no effect of beta-carotene on serum lipids (Nierenberg et al, 1991; Ringer et al, 1991; Hughes et al, 1994; Van Poppel et al, 1994; Ribaya-Mercado et al, 1995). In a large randomized trial in which a combination of antioxidants (vitamin C, vitamin E and beta-carotene) was compared to placebo, small but significant differences in both total cholesterol and triglycerides were observed at trial end. The active treatment arm had significantly higher concentrations of total cholesterol (mean (s.e.): 4.89 (0.017) vs 4.74 (0.017) mmol/l) and triglycerides (mean (s.e.): 2.13 (0.020) vs 1.92 (0.018) mmol/l); however, active treatment had no effect on fatal or nonfatal cardiovascular events (Heart Protection Study Collaborative Group, 2002). In other chemoprevention trials that have used beta-carotene alone or in combination with other antioxidants, no statistically significant excess of cardiovascular deaths has been observed in the beta-carotene arm (The Alpha-Tocopherol, Beta-Carotene Prevention Study Group, 1994; Greenberg et al, 1996; Hennekens et al, 1996; Green et al, 1999; Lee et al, 1999). A meta-analysis of such chemoprevention trials showed a slightly increased risk of cardiovascular disease mortality associated with beta-carotene (OR=1.1, 95% CI 1.03–1.17). This meta-analysis included CARET, in which beta-carotene was used in combination with vitamin A in the active intervention arm (Vivekananthan et al, 2003). CARET was the only trial of the six included in the meta-analysis for which deaths due to cardiovascular disease were statistically significantly higher in the active intervention arm, which included beta-carotene. The statistically significant relative difference in event rates for cardiovascular deaths in CARET (2.4% active arm vs 1.7% placebo arm) included in the meta-analysis was considerably higher than any of the other five trials included (eg ATBC 5.3% active arm vs 4.9% placebo arm), none of which was statistically significant within the individual trial setting.

A prior report of the Pilot Study and Vanguard CARET participants with 72 months follow-up described a statistically significant higher triglyceride concentration in the group receiving vitamin A and or combination of vitamin A and beta-carotene compared to the placebo group (Omenn et al, 1994a); no data were presented for cholesterol. In this report we focus on the presentation of serum cholesterol data for the CARET Pilot study and Vanguard participants; however, for completeness we include a re-analysis of the triglyceride data previously presented (Omenn et al, 1994a). As in the prior report (Omenn et al, 1994a) to help focus on the possible effect of vitamin A, those who received only beta-carotene (n=256) during the heavy smoker pilot study have been excluded from this ancillary analysis.

Methods

Study population

CARET is a double-blind randomized trial in which 18 314 participants at elevated risk for lung cancer received either 30 mg beta-carotene and 25 000 IU retinyl palmitate (vitamin A) or a placebo (Omenn et al, 1996). This report includes data from the pilot study/Vanguard CARET participants, whose enrollment began in 1985, the only group of participants within CARET who had routine triglyceride and cholesterol measurements. Briefly, in one pilot study, 816 male asbestos-exposed participants were randomized to receive 15 mg beta-carotene and 25 000 IU retinol/day, or placebo (Omenn et al, 1993). In the companion pilot study, 539 male and 490 female smokers were randomized to one of four equal arms: placebo, 30 mg beta-carotene/day, 25 000 IU retinol/day, 30 mg beta-carotene and 25 000 IU retinol/day (Goodman et al, 1993). Between July 1, 1988, and June 30, 1989, 98% of the asbestos-exposed pilot participants and 95% of the smoker pilot participants who had been active throughout the trial became the CARET Vanguard Cohort (N=1456) (Thornquist et al, 1993; Omenn et al, 1996; Bowen et al, 1999). Those receiving the placebo continued to do so; the remaining participants received the combination of 30 mg beta-carotene and 25 000 IU retinyl palmitate (vitamin A)/day.

Data collection

Baseline and demographic data were collected at enrollment (Goodman et al, 1993; Omenn et al, 1993). Information on new cancers and vital status of participants was collected at least annually (Omenn et al, 1994b). Mortality from cardiovascular disease (ICD categories: 390–459, 798) was determined by a review of medical records and death certificates by the CARET Endpoints Committee.

Collection and analysis of blood samples

Cholesterol and triglycerides were measured in a standard commercial automated multitest analysis. HDL and LDL were not assessed. Serum samples were obtained at enrollment (2 months prior to randomization in a pilot study), 4 months and 12 months after randomization, and annually thereafter. Participants were not required to fast prior to blood collection, but information on the number of hours prior to the blood draw the participant last ate was recorded. Adjustment for time since last ate was made in the multivariate model (see Statistical Analysis below). Participants included in this report had a baseline cholesterol/triglyceride assessment (2 months prior to randomization) and at least one other measurement postrandomization. Data are only included for the active intervention period for each participant, that is for the period when the participant was taking the study capsules.

Statistical analysis

To help focus on the possible effect of vitamin A on serum lipids, we excluded the group of participants assigned to receive beta-carotene alone in the smokers pilot study (n=256).

The t-test was used to compare unadjusted cholesterol and triglyceride concentrations between the intervention arms at each time point. A natural log transformation was conducted to minimize the right skewing of the triglyceride data. To test for treatment differences in cholesterol and triglyceride levels and the rate of change of the cholesterol and triglyceride over time, a random coefficients analysis of covariance model was used (Brown & Prescott, 1999) that adjusted for baseline pretreatment lipid concentrations.

Briefly, we fit fixed effects for time, treatment group and the time by treatment group interaction, which represented the average rate of change, the average intercept for each treatment group, and the extent to which treatments differed with respect to the average rate of change, respectively. Random effects were also included for intercept and time, which allowed for these parameters to vary randomly between patients essentially permitting a separate regression line to be fitted for each subject. This allowed for the estimation of the average effect of the treatment on the individual trajectories of cholesterol and triglyceride measurements with individuals. Model validation was accomplished by comparing Akaike's criteria (Brown & Prescott, 1999) for alternative models including interactions, quadratic effects, and other covariance structures.

The models were adjusted for baseline concentration of cholesterol or triglycerides, baseline age (continuous), sex (dichotomous), BMI (continuous), smoking status (dichotomous), and fasting status (dichotomous). SAS/STAT (ver 8.0; SAS Institute Inc., Cary, NC, USA) was used for the statistical analyses.

Results

Demographics and risk factors

Demographic and risk factor information for the participants included in the analysis is shown in Table 1. With a few exceptions, the placebo and vitamin A/beta-carotene groups were very similar. There was a somewhat larger proportion of females in the vitamin A/beta-carotene arm due to the gender eligibility criteria for the pilot studies (Table 1). Body mass index was slightly but significantly lower in the vitamin A/beta-carotene group as compared to the placebo group. Smoking behavior prior to and during the trial was similar between the two groups, with 51.3% of the vitamin A/beta-carotene group and 54.3% of the placebo group current smokers at enrollment and 12.2 and 10.5% quitting smoking during the trial in the respective groups. At enrollment, 12.6% of those taking vitamin A or vitamin A and beta-carotene reported having fasted for 8 h prior to the blood draw compared to 13.1% of those in the placebo group. The comparable percentages for the subsequent visits ranged from 21.9 to 36.0% for the group receiving vitamin A and from 18.5 to 38.4% for those in the placebo group.

Table 1 - Baseline characteristics of participants from smokers pilot (excluding beta-carotene only arm) and asbestos-exposed pilot.

Full table

Cholesterol

The mean cholesterol concentrations were comparable between groups at baseline but thereafter cholesterol concentrations were higher in the vitamin A/beta-carotene group compared to the placebo group, with the difference reaching statistical significance at the 4, 12, 36 and 60 months time points (Figure 1), although cholesterol concentrations declined notably over the trial in both groups. Females had significantly higher serum cholesterol concentrations than males; at baseline, females in these two groups had a mean cholesterol concentration of 243.8 mg/dl (95% CI, 237.7–249.9) (6.31 mmol/dl) compared to 237.4 mg/dl (95% CI, 233.8–241.0) (6.15 mmol/dl) in the males. However, once on intervention, both males and females assigned to the vitamin A/beta-carotene arm had higher concentrations of cholesterol than their counterparts in the placebo arm (data not shown).

Figure 1.

Unadjusted mean cholesterol concentrations in the vitamin A/beta-carotene and placebo arms of the CARET Vanguard group (95% CI). Cholesterol conversion factor to provide values mmol/l is 0.0259.

Full figure and legend (14K)

The results of the analysis of covariance show a statistically significant effect of the intervention (P=0.0003), with those in the vitamin A/beta-carotene arm having an average cholesterol concentration 5.3 mg/dl (s.e. 1.4) (0.14 mmol/dL) higher that those in the placebo arm. However, the time intervention term was not significant, indicating that there was no further change in the difference between the two intervention groups over the remainder of the study after the 4-month post-randomization time point. The magnitude of the difference between the arms was 7.3 mg/dl (s.e. 1.9) (0.19 mmol/dl) in the asbestos-exposed cohort and 2.0 mg/dl (s.e. 2.1) (0.05 mmol/dl) in the heavy smoker cohort; no significant difference in cardiovascular mortality was seen between the asbestos-exposed and heavy smoker cohorts. Overall, total cholesterol concentrations for the placebo group and active group combined decreased significantly over the course of the trial at a rate of 2.7 mg/dl (s.e. 0.37) (0.07 mmol/dl) per year (P<0.0001).

Triglycerides

The geometric mean triglyceride concentrations were comparable between groups at baseline, but thereafter triglyceride concentrations were higher in the vitamin A/beta-carotene group with the difference reaching statistical significance at 4 months and persisting thereafter (Figure 2). Triglyceride concentrations showed no progressive increase. In contrast to cholesterol concentrations, male subjects had significantly higher serum triglyceride concentrations compared to female subjects; at baseline the geometric mean concentration was 183.1 mg/dl (2.07 mmol/dl) in male subjects and 151.4 mg/dl (1.71 mmol/dl) in female subjects. However, for both male and female subjects, the geometric mean triglyceride concentrations were comparable between the vitamin A/beta-carotene and placebo arms at baseline, but thereafter triglyceride concentrations were higher in the vitamin A/beta-carotene group (data not shown). In both male and female subjects, triglyceride concentrations were elevated above baseline at the 4-month time point and remained elevated for males through month 60 and for female subjects throughout the study period.

Figure 2.

Unadjusted geometric mean triglyceride concentrations in the vitamin A/beta-carotene and placebo arms of the CARET Vanguard group (95% CI). Triglyceride conversion factor to provide values mmol/l is 0.0113.

Full figure and legend (14K)

The results of the analysis of covariance show a statistically significant association with the intervention, with those in the vitamin A/beta-carotene group arm having significantly higher triglyceride concentrations (unadjusted mean difference during the intervention period=16 mg/dl (0.18 mmol/dl)) than those in the placebo arm (P<0.0001). As with cholesterol concentrations, overall, serum triglyceride concentrations decreased significantly over time at a rate of 4.9 mg/dl (0.06 mmol/dl) per year in the placebo group (P<0.0001). The time intervention term was not statistically significant, indicating that there was no further change in the difference between the two intervention groups over the remainder of the study after the 4-month postrandomization time point.

Discussion

The findings presented here show that those participants receiving vitamin A alone (239 heavy smoker participants for a mean duration of 2 y prior to the transition to combination of vitamin A/beta-carotene) or vitamin A and beta-carotene combined (624 participants (388 asbestos-exposed and 236 heavy smoker participants) throughout the study period), had consistently higher concentrations of cholesterol than those in the placebo group (611 participants), although the serum cholesterol concentrations in both groups declined over time. The mean difference between groups during the intervention period was 5.3 mg/dl (0.14 mmol/l). The difference in cholesterol between the vitamin A/beta-carotene and placebo groups was larger for those enrolled in the asbestos-exposed pilot (7.3 mg/dl (0.19 mmol/l)) than those in the smokers pilot (2.0 mg/dl (0.05 mmol/l)). Interestingly, the difference between treatment arms for cardiovascular mortality was RR 1.60 (CI, 0.83–3.10) in the asbestos-exposed pilot vs RR=0.87 (CI, 0.47–1.61) (unpublished data) in the smokers pilot (including the beta-carotene-only arm excluded in this analysis); of course, the small numbers and overlapping confidence intervals limit the interpretation of these results. In addition, our results show that those receiving vitamin A alone or in combination with beta-carotene have significantly higher triglyceride concentrations. The difference between the groups remains significantly elevated over the course of the trial compared to those receiving the placebo.

We did not have data on use of lipid-lowering drugs for the Vanguard study population and thus could not adjust for lipid-lowering medication. However, limited data are available from a different subgroup of the asbestos-exposed efficacy study population. Redlich et al (1999) reported that a very small subgroup (n=52) of CARET participants from the New Haven Study Center receiving vitamin A and beta-carotene was twice as likely to start taking lipid-lowering drugs during the trial as was the placebo group (28 vs 13%). We found a similar pattern when all participants at the New Haven Study Center (n=1024) were included. Of those enrolled at the New Haven Study Center, significantly more participants in the group taking vitamin A/beta-carotene reported starting to use lipid-lowering medications during the course of the trial as compared to the placebo group (47 (9.6%) vs 23(4.7%); P<0.03). Another 41 subjects (21 vitamin A/beta-carotene vs 20 placebo) who were on lipid-lowering agents at enrollment were excluded from the drug use comparison. The greater use of lipid-lowering medication in this limited group of active participants suggests that the intervention may cause hyperlipidemia, and that the increased use of lipid-lowering agents could mask a larger effect of the intervention on lipid concentrations than we measured.

The reason for the unexpected increase in cardiovascular mortality observed in CARET remains unknown, but our data suggest that higher cholesterol and triglyceride concentrations in those on active intervention than the more rapidly declining values among those on placebo may contribute to this finding. There is a substantial body of evidence that synthetic retinoids and high dose vitamin A alter lipid metabolism by increasing cholesterol and/or triglycerides (Infante et al, 1991; Pastorino et al, 1991; Tangrea et al, 1993) and that such changes increase cardiovascular disease risk (Grundy, 1986; Betteridge & Morrell, 1998). Although the differences between the vitamin A/beta-carotene and placebo arms were relatively modest and the triglyceride differential was clearly nonprogressive (Omenn et al, 1994a), evidence from combined cohort studies shows that even a small change in the population mean cholesterol concentration has a substantial effect on coronary heart disease. Law et al (1994a) reported that, for men from several cohort studies, a log-linear association best described the relation between risk of ischemic heart disease and serum cholesterol concentration. The authors found that within the serum cholesterol range of 154–309 mg/dl (3.99–8.00 mmol/dl), an absolute difference in cholesterol of 23 mg/dl (0.60 mmol/dl) was associated with a 17–38% difference in ischemic heart disease in the 10 studies reported. From randomized trials, a 23 mg/dl (0.60 mmol/dl) reduction was associated with a 25% reduction in risk of ischemic heart disease after 5 y (Law et al, 1994a). Similarly, a higher total serum cholesterol of 23 mg/dl (0.60 mmol/dl) was associated with a 27% increase in mortality from ischemic heart disease in the BUPA study (Law et al, 1994b). In a meta-analysis involving 17 population-based studies of triglyceride concentrations and cardiovascular disease, Austin et al (1998) reported that, after adjustment for HDL cholesterol and other risk factors, an 89 mg/dl (1 mmol/l) increase in fasting triglycerides was associated with 14 and 37% higher incident cardiovascular disease for men and women, respectively; both fatal and nonfatal events were included in the analysis. So, on this basis and extrapolating very cautiously to the total CARET population, the lipid differences seen between arms might be associated with an approximately 6–10% difference in CV mortality rates in the full CARET population, a modest part of the observed difference. This estimate assumes that increase in risk of CV mortality is linear with increasing concentrations of each lipid and leaves aside the complications of analyzing quite different proportions of asbestos-exposed and heavy smoker participants in this substudy compared with the total CARET population or with the Vanguard or Efficacy subgroups.

In contrast to retinoids, beta-carotene has not been reported to affect cholesterol concentrations, with the exception of a single report of combined treatment with vitamins C and E in the MRC/BHF Heart Protection Study; during follow-up cholesterol was 3.2% and triglycerides 10.9% higher in the active treatment arm (Heart Protection Study Collaborative Group, 2002). As beta-carotene is an antioxidant, among many other actions, it was hypothesized to help prevent cardiovascular disease (Witztum, 1994). However, a meta-analysis of trials of beta-carotene alone or in combination with other antioxidants showed the active treatment groups to have a significantly increased mortality rate for cardiovascular death (Vivekananthan et al, 2003). The statistically significant relative difference in event rates for cardiovascular deaths in CARET (2.4% active arm vs 1.7% placebo arm) included in the meta-analysis was considerably higher than any of the other five trials included (eg ATBC 5.3% active arm vs 4.9% placebo arm), none of which was statistically significant within the individual trial setting. The largest study included in the meta-analysis is the Alpha Tocopherol and Beta-Carotene Study (The Alpha-Tocopherol, Beta-Carotene Prevention Study Group, 1994) in which 29 133 men were randomized to receive alpha-tocopherol (50 mg), beta-carotene (20 mg), both agents or placebo daily; the mortality rate for ischemic heart disease was somewhat higher in the group that received beta-carotene (77.1 per 10 000 person years) compared to those not receiving beta-carotene (68.9 per 10 000 person years); P=0.08. Later analysis of the ATBC data attributed the increase in mortality rate from coronary heart disease in the arms containing beta-carotene to the group of men who had had a previous myocardial infarction; men with previous myocardial infarction who received vitamin E also had a nonsignificant increase in mortality rate (Rapola et al, 1997). No effect on mortality from coronary heart disease was observed in those men with no prior history of myocardial infarction (Virtamo et al, 1998). In the Physicians' Health Trial, Hennekens et al (1996) reported no effect of beta-carotene on incidence of cardiovascular events or deaths from cardiovascular disease in the 22 071 participants who received the intervention for up to 12 y. Although the authors of the MRC/BHF Heart Protection Study reported significantly higher concentrations of triglycerides and cholesterol in those receiving beta-carotene, vitamins C and E, they reported no effect of the intervention on vascular events (OR=1.00) or mortality from vascular causes (OR=1.05) in that high-risk population of 20 536 participants (Heart Protection Study Collaborative Group, 2002). Notably, the mean cholesterol concentration in the MRC/BHF placebo population was considerably lower than in CARET (183 mg/dl (4.74 mmol/dl) vs 239 mg/dl (6.19 mmol/dl)).

The findings of this study are comparable to those reported by one of the authors and her colleagues for a similarly designed skin cancer prevention study using the same dose of vitamin A (Cartmel et al, 1999). Cartmel et al reported cholesterol concentrations to be significantly higher in the vitamin A group, with those in the vitamin A group having a geometric mean of 5.57 mmol/l (214 mg/dl) after 49 months and the placebo group a geometric mean of 5.42 mmol/l (209 mg/dl). As in our study, Cartmel et al reported a significant increase in triglycerides in the group receiving vitamin A and a decline in cholesterol concentrations over time in the placebo arm. The decrease in cholesterol concentrations over time seen in both studies could be explained in part by secular trends such as changes in the health habits of the participants or an increase in usage of medications to lower cholesterol. The study reported here spanned the years 1985–1992, a period when extensive cholesterol interventions that began in the middle 1980s appear to have created favorable cholesterol-related changes at the community level (Frank et al, 1992). In addition, some of the decline in cholesterol may reflect the aging of the population; there are several reports of declining cholesterol concentration starting in men and women in their 60s (Newschaffer et al, 1992; Wallace & Colsher, 1992; Weijenberg et al, 1996; Ferrara et al, 1997; Manolio et al, 2004). Manolio et al (2004) reported a mean decline of 6.3 mg/dl (0.16 mmol/l) over a 5-y period in a cohort of men and women aged 65 and over, similar to a 6 mg/dl (0.16 mmol/l) decline over a 4-year period reported in elderly men (Weijenberg et al, 1996). The mean age at enrollment for the study reported here was 59.

In addition to affecting cholesterol and triglyceride concentrations, it is possible that the intervention also altered the LDL/HDL ratio so as to increase cardiovascular risk. Cartmel et al (1999) suggested from their results that moderate concentrations of vitamin A may have an adverse effect on the LDL/HDL ratio: the LDL/HDL ratio increased slightly from 3.01.08 to 3.31.18 (NS) in the retinol group and from 3.11.17 to 3.21.14 (NS) in the placebo group in conjunction with a relative lowering of HDL and a relative increase in LDL in the retinol compared to the placebo group. Lowering of HDL and an increase in LDL has also been reported with use of some synthetic retinoids (Vahlquist et al, 1985; Marsden, 1986; Tangrea et al, 1993); such a change in LDL/HDL ratio increases risk of ischemic heart disease.

A further limitation of this study is the lack of required fasting prior to blood draw. Postprandial effects on triglyceride concentrations are well documented, although such effects appear to be of less importance for assessment of cholesterol concentration (Cooper et al, 1992). Adjustment was made in the analysis for fasting status and there was no significant difference between the study arms for fasting. However, as postprandial triglyceride concentrations do not follow a linear change with time, the statistical adjustment for fasting status may not completely correct for the different fasting status' of participants at different timepoints, and therefore nondifferential error may have been introduced because many patients did not fast prior to blood collection.

Although the study design has several limitations, we consider the results of interest to the scientific community as it seems unlikely that other large randomized trials of vitamin A and beta-carotene will be conducted to allow further investigation of this topic.

One cautionary note for future chemoprevention trials that the CARET results highlight is that small adverse effects, significant at the population level, may be missed in much smaller studies that use biomarkers as the endpoint of interest. Although such studies may help researchers select the most appropriate intervention agents to explore, full-scale chemoprevention studies will need to be conducted to assess not only the effectiveness of the agent(s) on reducing the cancer of interest but also to uncover unanticipated adverse effects.

In summary, cholesterol and triglycerides showed a continuous decline in CARET participants in the placebo arm. For those on the active intervention arm, cholesterol concentration declined less rapidly than in the placebo arm and triglycerides were elevated compared to baseline. These combined effects of the vitamin A/beta-carotene combination on lipids may account for some of the increase observed for cardiovascular mortality in the vitamin A/beta-carotene arm in CARET. Further investigation of the effects of vitamin A on HDL and LDL concentrations is warranted, especially as vitamin A supplements and retinoids remain in widespread use despite the CARET findings.

References

1. Austin MA, Hokanso JE & Edwards KL. (1998) Hypertriglyceridemia as a cardiovascular risk factor. Am. J. Cardiol. 81: 7B−12B. | Article | PubMed | ChemPort |
2. Bershad S, Rubinstein A, Paterniti JR, Le NA, Poliak SC, Heller B, Ginsberg HN, Fleishmajer R & Brown WV. (1985) Changes in plasma lipids and lipoproteins during isotretinoin therapy for acne. N. Engl. J. Med. 313: 981−985. | PubMed | ISI | ChemPort |
3. Betteridge DJ & Morrell JM. (1998) Clinicians' Guide to Lipids and Coronary Heart Disease London: Chapman and Hall Medical.
4. Bowen D, Cartmel B, Barnett M, Goodman GE & Omenn GS. (1999) Predictors of participant retention in two chemoprevention feasibility trials. Ann. Behav. Med. 21: 210−215. | PubMed | ChemPort |
5. Brown H & Prescott R. (1999) Applied Mixed Models in Medicine Chichester, England: John Wiley and Sons.
6. Cartmel B, Moon TE & Levine N. (1999) The effects of long term intake of retinol on selected clinical and laboratory parameters. Am. J. Clin. Nutr. 69: 937−943. | PubMed | ChemPort |
7. Cooper GR, Myers GL, Smith SJ & Schlant RC. (1992) Blood lipid measurements. Variations and practical utility. JAMA 267: 1652−1660. | Article | PubMed | ChemPort |
8. Ferrara A, Barrett-Connor E & Shan J. (1997) Total, LDL, and HDL cholesterol decrease with age in older men and women. The Rancho Bernardo Study 1984−1994. Circulation 96: 37−43. | PubMed | ChemPort |
9. Frank E, Winkleby MA, Fortmann SP, Rockhill B & Farquhar JW. (1992) Improved cholesterol-related knowledge and behavior and plasma cholesterol levels in adults during the 1980s. JAMA 268: 1566−1572. | Article | PubMed | ChemPort |
10. Goodman GE, Omenn GS, Thornquist MD, Lund B, Metch B & Gylys_Colwell I. (1993) The Carotene and Retinol Efficacy Trial (CARET) to prevent lung cancer in high-risk populations: pilot study with cigarette smokers. Cancer Epidemiol. Biomarkers Prev. 2: 389−396. | PubMed | ChemPort |
11. Green A, Williams G, Neale R, Hart V, Leslie D, Parsons P, Marks GC, Gaffney P, Battistutta D, Frost C, Lang C & Russell A. (1999) Daily sunscreen application and beta-carotene supplementation in prevention of basal-cell and squamous-cell carcinomas of the skin; a randomized controlled trial. Lancet 354: 723−729. | Article | PubMed | ChemPort |
12. Greenberg ER, Baron JA, Karagas MR, Stukel TA, Nierenberg DW, Stevens MM, Mandel JS & Haile RW. (1996) Mortality associated with low plasma concentration of beta carotene and the effect of oral supplementation. JAMA 275: 699−703. | Article | PubMed | ChemPort |
13. Grundy SM. (1986) Cholesterol and coronary heart disease. A new era. JAMA 256: 2849−2858. | Article | PubMed | ChemPort |
14. Heart Protection Study Collaborative Group. (2002) MRC/BHF Heart Protection Study of antioxidant vitamin supplementation in 20 536 high-risk individuals: a randomized placebo-controlled trial. Lancet 360: 23−33. | ISI |
15. Hennekens CH, Buring JE, Manson JE, Stampfer M, Rosner B, Cook NR, Belanger C, LaMotte F, Gaziano JM, Ridker PM, Willett W & Peto R. (1996) Lack of effect of long-term supplementation with beta-carotene on the incidence of malignant neoplasms and cardiovascular disease. New. Engl. J. Med. 334: 1145−1149. | Article | PubMed | ChemPort |
16. Hughes GS, Jr, Ringer TV, Francom SF, Means LK & DeLoof MJ. (1994) Lack of effects of beta-carotene on lipids and sex steroid hormones in hyperlipidemics. Am. J. Med. Sci. 308: 16−22. | PubMed |
17. Infante M, Pastorino U, Chiesa G, Bera E, Pisani P, Valente M & Raviasi G. (1991) Laboratory evaluation during high-dose vitamin A administration: a randomized study of lung cancer patients after surgical resection. J. Cancer Res. Clin. Oncol. 117: 156−162. | Article | PubMed | ChemPort |
18. Law MR, Wald NJ & Thompson SG. (1994a) By how much and how quickly does reduction in serum cholesterol concentration lower risk of ischaemic heart disease? BMJ 308: 367−373. | PubMed | ChemPort |
19. Law MR, Wald NJ, Wu T, Hackshaw A & Bailey A. (1994b) Systematic underestimation of association between serum cholesterol concentrations and ischaemic heart disease in observation studies: data from the BUPA study. BMJ 308: 363−366. | ChemPort |
20. Lee I, Cook NR, Manson JE, Buring JE & Hennekens CH. (1999) Beta-carotene supplementation and incidence of cancer and cardiovascular disease: the Women's Health Study. J. Natl. Cancer Inst. 91: 2102−2106. | Article | PubMed | ChemPort |
21. Manolio TA, Cushman M, Gottdiener JS, Dobs A, Kuller LH & Kronmal RA. (2004) Predictors of falling cholesterol levels in older adults: The Cardiovascular Health Study. Ann. Epidemiol. 14: 325−331. | Article | PubMed |
22. Marsden J. (1986) Hyperlipidaemia due to isotretinoin and etretinate: possible mechanisms and consequences. Br. J. Dermatol. 114: 401−407. | PubMed | ChemPort |
23. Murray JC, Gilor RS & Lazarus GS. (1983) Serum triglyceride elevation following high-dose vitamin A treatment for pityriasis rubra pilaris. Arch. Dermatol. 119: 675−676. | Article | PubMed | ChemPort |
24. Newschaffer C, Bush TL & Hale WE. (1992) Aging and total cholesterol levels: cohort, period and survivorship effects. Am. J. Epidemiol. 136: 23−34. | PubMed | ChemPort |
25. Nierenberg DW, Bayrd GT & Stukel TA. (1991) Lack of effect of chronic administration of oral beta-carotene on serum cholesterol and triglyceride concentrations. Am. J. Clin. Nutr. 53: 652−654. | PubMed | ChemPort |
26. Omenn GS, Goodman G, Thornquist M, Rosenstock LS, Barnhart S, Gylys-Colwell I, Metch B & Lund B. (1993) The Carotene and Retinol Efficacy Trial (CARET) to prevent lung cancer in high risk populations: pilot study with asbestos-exposed workers. Cancer Epidemiol. Biomarkers Prev. 2: 381−387. | PubMed | ChemPort |
27. Omenn GS, Goodman GE, Thornquist MD, Balmes J, Cullen MR, Glass A, Keogh JP, Meyskens FL, Valanis B, Williams JH, Barnhart S & Hammar S. (1996) Effects of a combination of beta-carotene and vitamin A on lung cancer and cardiovascular disease. N. Engl. J. Med. 334: 1150−1155. | Article | PubMed | ISI | ChemPort |
28. Omenn GS, Goodman GE, Thornquist M & Brunzell JD. (1994a) Long term vitamin A does not produce clinically significant hypertriglyceridemia: results from CARET, the Beta-Carotene and Retinol Efficacy Trial. Cancer Epidemiol. Biomarkers Prev. 3: 711−713. | ChemPort |
29. Omenn GS, Goodman GE, Thornquist M, Grizzle J, Rosenstock L, Barnhart S, Balmes J, Cherniack MG, Cullen MR & Glass A. (1994b) The -Carotene and Retinol Efficacy Trial (CARET) for chemoprevention of lung cancer in high-risk populations: smokers and asbestos-exposed workers. Cancer Res. 54: 2038−2043.
30. Pastorino U, Chiesa G, Infante M, Soresi E, Clerici M, Valente M, Bellon PA & Ravasi G. (1991) Safety of high-dose vitamin A. Oncology 48: 131−137. | PubMed | ChemPort |
31. Rapola JM, Virtamo J, Ripatti S, Huttunen JK, Albanes D, Taylor PR & Heinonene OP. (1997) Randomized trial of alpha-tocopherol and beta-carotene supplements on incidence of major coronary events in men with previous myocardial infarction. Lancet 349: 1715−1720. | Article | PubMed | ChemPort |
32. Redlich CA, Chung JS, Cullen MR, Blaner WS, Van Bennekum AM & Berglund L. (1999) Effect of long-term beta-carotene and vitamin A on serum cholesterol and triglyceride levels among participants in the Carotene and Retinol Efficacy Trial (CARET). Atherosclerosis 143: 427−434. | Article | PubMed | ChemPort |
33. Ribaya-Mercado JD, Ordovas JM & Russell RM. (1995) Effect of beta-carotene supplementation of the concentrations and distribution of carotenoids, vitamin E, vitamin A and cholesterol in plasma lipoprotein and non-lipoprotein fractions in healthy older women. J. Am. Coll. Nutr. 14: 614−620. | PubMed | ChemPort |
34. Ringer TV, DeLoof MJ, Winterrowd GE, Francom SF, Gaylor SK, Ryan JA, Sanders ME & Hughes GS. (1991) Beta-carotene's effects on serum lipoproteins and immunologic indices in humans. Am. J. Clin. Nutr. 53: 688−694. | PubMed | ChemPort |
35. Tangrea JA, Adrianza ME, Helxel WE, Taylor PR, Hartman AM, Peck GL & Edwards BK. (1993) Clinical and laboratory adverse effects associated with long-term, low-dose isotretinoin: incidence and risk factors. Cancer Epidemiol. Biomarkers Prev. 2: 375−380. | PubMed | ChemPort |
36. The Alpha-Tocopherol, Beta-Carotene Prevention Study Group. (1994) The effect of vitamin E and beta-carotene on the incidence of lung cancer and other cancer in male smokers. New Engl. J. Med. 330: 1029−1035. | PubMed |
37. Thornquist MD, Omenn GS, Goodman GE, Grizzle JE, Rosenstock L, Barnhart S, Anderson GL, Hammar S, Balmes J, Cherniack M, Cone J, Cullen MR, Glass A, Keogh JP, Meyskens F, Valanis B & Williams JH. (1993) Statistical design and monitoring of the Carotene and Retinol Efficacy Trial (CARET). Cont. Clin. Trials 14: 308−324. | Article | ChemPort |
38. Vahlquist C, Michaëlsson G, Vahlquist A & Vessby B. (1985) A sequential comparison of etretinate (Tigason®) with special regard to their effects on serum lipoproteins. Br. J. Dermatol. 112: 69−76. | PubMed | ChemPort |
39. Van Poppel G, Hospers H, Buytenhek R & Princen HM. (1994) No effects of beta-carotene supplementation on plasma lipoproteins in healthy smokers. Am. J. Clin. Nutr. 60: 730−734. | PubMed | ChemPort |
40. Virtamo J, Rapola JM, Ripatti S, Heinonen OP, Taylor PR, Albanes D & Huttunen JK. (1998) Effect of vitamin E and beta carotene on the incidence of primary nonfatal myocardial infarction and fatal coronary heart disease. Archives Int. Med. 158: 668−675. | Article | ChemPort |
41. Vivekananthan DP, Penn MS, Sapp SK, Hsu A & Topol EJ. (2003) Use of antioxidant vitamins for the prevention of cardiovascular disease: meta-analysis of randomized trials. Lancet 361: 2017−2023. | Article | PubMed | ChemPort |
42. Wallace RB & Colsher PL. (1992) Blood lipid distributions in older persons: prevalence and correlates of hyperlipidemia. Ann. Epidemiol. 2: 15−21. | PubMed | ChemPort |
43. Weijenberg MP, Feskens EJ & Kromhout D. (1996) Age-related changes in total and high-density -lipoprotein cholesterol in elderly Dutch men. Am. J. Public Health. 86: 798−803. | PubMed | ChemPort |
44. Witztum JL. (1994) The oxidation hypothesis of atherosclerosis. Lancet 344: 793−795. | Article | PubMed | ISI | ChemPort |
45. Zech LA, Gross EF, Peck GL & Brewer HB. (1983) Changes in plasma cholesterol and triglyceride levels after treatment with oral isotretinoin. A prospective study. Arch. Dermatol. 119: 987−993. | Article | PubMed | ChemPort |

Acknowledgements

We thank all the CARET study participants and CARET study staff.