Academic Editors: Manuel Martínez-Sellés and Peter Kokkinos
Background: Evidence on statin use for primary prevention of
cardiovascular disease (CVD) in older people needs to be extended and updated,
aiming to provide further guidance for clinical practice. Methods:
PubMed, EMBASE, Cochrane Library and Web of Science were searched for eligible
observational studies comparing statin use vs. no-statin use for primary
prevention of CVD in older people (age
Cardiovascular disease (CVD) is a global burden, and more than 80% cases of
mortality occur in older population (age
It is well-established that statin use is recommended for secondary prevention
of CVD in older people as level A evidence, while considerable evidence for
primary prevention is insufficient [5, 6]. Currently, statin therapy for high CVD
risk people
Another meta-analysis reported reduced CVD risk in statin-use for secondary prevention over the primary prevention in older population, and the data are insufficient for the risk of onset diabetes [10]. The main limitations of former results include that they mainly focus on component outcomes (major vascular events) rather than specific outcomes (coronary heart disease (CHD), myocardial infraction (MI), stroke, etc.) [11]. Then, considering the strict inclusion criteria, older people were always omitted from clinical trials. Current results on the primary prevention for older populations were always from subgroup analyses, which is not enough [10, 12]. Worse more, evidence on this topic based on clinical trials was coupled with limited sample size of intended population in a short period of follow-up [12, 13]. To our point of view, the outcomes of interest like total CVD events were not reported in previous meta-analysis upon observational studies, which also lacked some key eligible studies [14]. Therefore, we could not have a comprehensive evaluation of the statin use for CVD primary prevention especially in older population. Observational studies in this scope may extend the current limited evidence with larger population and longer follow-up period. Herein, we conducted this meta-analysis based on observational studies to (1) investigate the CVD primary prevention via statin use in older population; (2) present the preventive association by age; (3) make updated clinical advice to high CVD risk population.
According to the Cochrane Handbook and the Meta-analysis of Observational Studies in Epidemiology (MOSE) Guidelines Checklist and Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Supplementary Table 1) [15, 16], this study was designed. The protocol is consistent with a previous study [14], and has been registered on the INPLASY website (https://inplasy.com/) with a reference ID: INPLASY2021120045 (doi: 10.37766/inplasy2021.12.0045) (Appendix File 1).
We reviewed Pubmed, EMBASE, Cochrane Library and Web of Science for related literatures from the inception to Sep. 15th, 2021. We used a combination of relevant keywords and Medical Subject Headings (MeSH) terms, including “Aging”, “Aged”, “elderly”, “Statin”, “atorvastatin”, “cardiovascular disease”, “cardiovascular events”, “coronary heart disease”, “myocardial infarction”, “stroke” and “observational study”. Detailed search strategy is given in Supplementary Table 2. No restrictions were applied on language. Reference lists of the retrieved literature were also searched manually.
All articles were screened in two-step methods. Two authors independently screened the studies’ titles and abstracts, then reviewed the full texts of potentially eligible studies. Any disagreements were resolved by another author who is exceptional in cardiology and evidence-based medicine from a discussion in a group panel.
The eligible criteria following PICOS principles are as follows.
Being limited to or including a subgroup of older people aged
Statin (atorvastatin, fluvastatin, lovastatin, pitavastatin, pravastatin, rosuvastatin, or simvastatin) use vs. no statin use for primary prevention.
Including at least one of the following outcomes: All-cause mortality, CVD mortality, CHD/MI, stroke or total CV events.
Only the most informative studies with longer follow-up (no less than one year considering the limited life expectancy of older people) could be included to avoid duplication. Clinical trials, reviews, case reports, conference abstracts, experimental studies, and studies without essential data were excluded.
Two independent authors performed data extraction following a prespecified protocol from eligible studies. The extracted information included characteristics of the eligible studies (year of publication, first author, study design, study location, follow-up period, etc.), characteristics of the populations (median age and sample size), and the characteristics of the program (systematic exposure, outcomes of endpoints, adjusted confounders, registration information, etc.). All risk estimates were evaluated in fully adjusted models. Intention-to-treat principles (ITT) were applied if available, and the primary authors would be contacted if there were missing data. However, analyses would still have been taken without these data if no response was received.
The primary outcomes included risk of all-cause mortality, CVD mortality, CHD/MI, stroke and total CV events, because they had most clinical significance and abundant useful data. Secondary outcomes included risk estimate on no diabetes mellitus (NODM) and cancer incidence. Detailed definitions about outcomes of interest have been summarized in Supplementary material 1. The data regarding older people who survived from the first age to a new age were reported by independent cohorts, respectively, and then the data could be deemed as being achieved from two different cohorts. The methods to avoid duplication have been addressed in the selection criteria part.
To evaluate the quality of included studies, we applied the Newcastle-Ottawa Scale (NOS) as previously, which has been validated to assess the quality of nonrandomized controlled trials in meta-analyses [17]. As for a 0–10 scale, each study was categorized as low (0–5), medium (6–7), and high (8–10) quality. Two authors performed a quality assessment on all of the included studies based on the method. In case of any disagreements, there would be a discussion between the two authors.
Afterwards, we used the Risk of Bias in Non-randomized Studies of Interventions (ROBINS-I) tool to make further risk estimates on the included studies [18]. This tool displays 7 items and classifies the risks of bias into low, moderate, serious, critical and unclear risks. The process was completed by two independent authors and there would be a discussion in case of any disagreements.
In this case, we applied the Grading of Recommendation Assessment, Development and Evaluation (GRADE) approach to identify the level rating of each outcome of interest as very low, low, moderate, or high quality [19]. The rating system follows 5 items: risk of bias, imprecision, inconsistency, indirectness, publication bias, large effect size, dose-response gradient and all residual confounding reducing an effect size [20, 21]. If there was one “serious” item, the evidence level could have been regarded as “low”; and if there was one “very serious”, the evidence level been “very low”.
Multivariable hazard ratio (HR) and the corresponding 95% confidence intervals
(95% CIs) for outcome of interests obtained from Cox-Hazard regression analysis
were mainly estimated with DerSimonian-Laird (D-L) random effects model, because
the assumptions involved accounted for the presence of within-study and
between-study heterogeneity. In order to provide the most comprehensive results,
both fixed- and random-effects models results were shown in the forest plots. The
adjusted relative risk (RR) and odd ratio (OR) in primary studies were
approximately considered as HR. Fully adjusted HRs and standard errors (SEs)
originating from the correspondence 95% CIs were logarithmically transformed to
stabilized variance, and the distribution was normalized. Between-study
heterogeneity was determined with the Cochran Q chi-square test and
I
In addition, a sensitivity analysis was performed by moving one study each turn to try to elaborate the causes of the heterogeneity. We would also conduct post subgroup analyses to ascertain the influence of other design and individual factors as follows: different categories on age, region, diabetic characteristics, hypertension status, study follow-up period and study design.
Publication bias was investigated by Egger’s linear regression tests at
p
Among 869 studies (846 from the main searched databases (PubMed = 486, EMBASE = 316, Cochrane Library = 22, Web of Science = 22) and 23 from other related literature), 803 studies were excluded after initial screening, and 20 studies were excluded after full consideration due to no required outcomes of final interest, overlapped outcomes, different types of statin plus other drugs, biased outcomes definition, etc. (Fig. 1).
The flow chart for study screening and selection.
A total of 12 observational studies incorporating 1,627,434 population were
eligible for this analysis [24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35]. Detailed characteristics were summarized in
Table 1 (Ref. [24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35]). All eligible studies involved
Study | Population | Follow-up period (y) | Groups | Mean Age (y) | Female (n/%) | BMI (kg/m |
TC† (mmol/L) | LDL-C† (mmol/L) | HDL-C† (mmol/L) | TG‡ (mmol/L) | Smoking status* (%) | Alcoholic Status* (%) | DM (%) | HTN (%) | Outcomes | Study Design |
Lemaitre et al. [24], 2002 (USA) | 7.3 | Stain used group (n = 251) | 71.1 (4.6) | 172 (68.5) | 49.4 | 5.83 (1.1) | 3.70 (1.1) | 1.39 (0.4) | 1.78 (1.0) | 9.6 | 49.4 | 21.9 | 48.20% | ①, ②, ③, ④, ⑤ | Prospective cohort study | |
Stain recommended group (n = 717) | 72.7 (5.6) | 478 (66.7) | 27.5 (4.5) | 6.70 (0.9) | 4.59 (0.7) | 1.31 (0.3) | 1.76 (0.7) | 14.6 | 45.3 | 20.5 | 48.1 | |||||
Diet recommended group (n =946) | 72.5 (5.3) | 600 (63.4) | 27.2 (5.0) | 5.92 (0.7) | 3.82 (0.5) | 1.37 (0.4) | 1.63 (0.7) | 13.9 | 48.9 | 20 | 43.7 | |||||
Alpérovitch et al. [25], 2015 (France) | 9.1 | Stain used group (n = 1007) | 73.1 (4.6) | 683 (67.8) | 25.8 (4.0) | 5.68 (0.9) | 3.40 (0.9) | 1.64 (0.4) | 1.27 (0.84–1.93)§ | 34.6 | 82.9 | 10.9 | 79.7 | ③, ④, ⑤ | Prospective cohort study | |
No stain use group (n = 5436) | 74.1 (5.6) | 3368 (62.0) | 25.4 (4.0) | 5.97 (1.0) | 3.78 (0.9) | 1.63 (0.4) | 1.14 (0.76 to 1.70)§ | 37.6 | 82.5 | 7.2 | 74.5 | |||||
Gitsels et al. [26], 2016 (UK) | 65, 70, 75 years people without prior CVD stratified by QRISK2 Score | 16–24 | Stain used group in QRISK |
65 | 833 (100) | 26.0 (4.0) | NA | NA | NA | NA | 10 | NA | 0 | NA | ① | Retrospective study |
No stain used group in QRISK |
65 | 39866 (100) | 26.0 (4.0) | NA | NA | NA | NA | 13 | NA | 0 | NA | |||||
Stain used group in QRISK |
70 | 3 (100) | 28.0 (6.0) | NA | NA | NA | NA | 0 | 75.6 | 0 | NA | |||||
No stain used group in QRISK |
70 | 322 (100) | 25.0 (4.0) | NA | NA | NA | NA | 23 | 0 | NA | ||||||
Stain used group in QRISK 10–19% (n = 6438) | 65 | 4381 (68) | 28.0 (5.0) | NA | NA | NA | NA | 34 | NA | 7 | NA | |||||
No stain used group in QRISK 10–19% (n = 116240) | 65 | 54094 (47) | 26.0 (4.0) | NA | NA | NA | NA | 41 | NA | 1 | NA | |||||
Stain used group in QRISK 10–19% (n = 10822) | 70 | 9928 (92) | 27.0 (5.0) | NA | NA | NA | NA | 21 | NA | 0 | NA | |||||
No stain used group in QRISK 10–19% (n = 108703) | 70 | 93010 (86) | 26.0 (5.0) | NA | NA | NA | NA | 22 | NA | 0 | NA | |||||
Stain used group in QRISK 10–19% (n = 661) | 75 | 661 (100) | 26.0 (4.0) | NA | NA | NA | NA | 5 | NA | 0 | NA | |||||
No stain used group in QRISK 10–19% (n = 13685) | 75 | 13684 (100) | 25.0 (4.0) | NA | NA | NA | NA | 6 | NA | 0 | NA | |||||
Stain used group in QRISK |
65 | 1742 (33) | 29.0 (5.0) | NA | NA | NA | NA | 64 | NA | 59 | NA | |||||
No stain used group in QRISK |
65 | 4532 (16) | 27.0 (5.0) | NA | NA | NA | NA | 76 | NA | 22 | NA | |||||
Stain used group in QRISK |
70 | 9570 (37) | 29.0 (5.0) | NA | NA | NA | NA | 56 | NA | 39 | NA | |||||
No stain used group in QRISK |
70 | 23626 (24) | 26.0 (4.0) | NA | NA | NA | NA | 59 | NA | 8 | NA | |||||
Stain used group in QRISK |
75 | 19566 (56) | 28.0 (5.0) | NA | NA | NA | NA | 44 | NA | 29 | NA | |||||
No stain used group in QRISK |
75 | 78799 (55) | 26.0 (4.0) | NA | NA | NA | NA | 41 | NA | 5 | NA | |||||
Orkaby et al. [27], 2017 (USA) | 7 | Stain used group (n = 1130) | 76.0 (4.5) | 0 (0) | 25.6 (3.1) | NA | NA | NA | NA | 51.8 | 85.1 | 13 | 73.8 | ①, ③, ④, ⑤ | Prospective cohort study | |
No stain use group (n = 1130) | 76.0 (4.6) | 0 (0) | 25.6 (3.2) | NA | NA | NA | NA | 53.8 | 85.9 | 13.1 | 75.3 | |||||
Ramos et al. [28], 2018 (Spain) | 5.6 | Stain used in 75–84 years, no DM group (n = 4802) | 78.8 (2.7) | 3126 (65.1) | 28.6 (4.6) | 6.1 (1.1) | 3.9 (1.0) | 1.5 (0.4) | 1.4 (0.7) | 13.5 | NA | NA | 65.7 | ①, ③, ④, ⑤, ⑥, ⑦ | Retrospective study | |
No stain used in 75–84 years, no DM group (n = 27114) | 79.1 (2.8) | 17028 (62.8) | 28.4 (4.6) | 5.4 (0.9) | 3.3 (0.7) | 1.5 (0.4) | 1.2 (0.5) | 12.4 | NA | NA | 57.3 | |||||
Stain used in |
88.5 (3.2) | 519 (69.8) | 27.1 (4.3) | 5.9 (1.2) | 3.7 (1.0) | 1.5 (0.4) | 1.4 (0.6) | 7.8 | NA | NA | 66.8 | |||||
No stain used in |
88.6 (3.2) | 4415 (69.8) | 27.6 (4.5) | 5.2 (0.9) | 3.1 (0.8) | 1.6 (0.4) | 1.2 (0.5) | 6.7 | NA | NA | 58.7 | |||||
Stain used in 75–84 years, DM group (n = 1756) | 78.8 (2.6) | 1076 (61.3) | 29.7 (4.7) | 5.8 (1.1) | 3.7 (0.9) | 1.4 (0.4) | 1.7 (0.8) | 15.4 | NA | NA | 78.4 | |||||
No stain used in 75–84 years, DM group (n = 4885) | 79.2 (2.8) | 2833 (58) | 29.4 (4.8) | 5.0 (0.8) | 3.0 (0.7) | 1.4 (0.4) | 1.4 (0.7) | 14.7 | NA | NA | 75.1 | |||||
Stain used in |
88.2 (2.8) | 135 (67.2) | 29.7 (4.7) | 5.8 (1.1) | 3.7 (0.9) | 1.4 (0.4) | 1.7 (0.8) | 15.4 | NA | NA | 78.4 | |||||
No stain used in |
88.2 (2.7) | 706 (68) | 29.4 (4.8) | 5.0 (0.8) | 3.0 (0.7) | 1.4 (0.4) | 1.4 (0.7) | 14.7 | NA | NA | 75.1 | |||||
Bezin et al. [29], 2019 (France) | 4.7 | Primary prevention without modifiable risk factors (n = 752) | 78 (76–81) | 540 (71.8) | NA | NA | NA | NA | NA | NA | NA | 0 | NA | ① | Retrospective study | |
Jun et al. [30], 2019 (South Korea) | NA | Cases (n = 11017) | 83.7 (3.2) | 6966 (66.4) | NA | NA | NA | NA | NA | NA | NA | 14.7 | 44.2 | ①, ③, ④, ⑤ | Nested case-control study | |
Controls (n = 55085) | 83.7 (3.2) | 34830 (63.2) | NA | NA | NA | NA | NA | NA | NA | 11.5 | 49.9 | |||||
Kim et al. [31], 2019 (South Korea) | 5.2 | Stain used group (n = 639) | 78 (76–80) | 413 (64.6) | 23.4 (22.2–25.8)§ | 4.46 (3.78–5.18)§ | 2.77 (2.20–3.45)§ | 1.17 (1.01–1.40)§ | 1.27 (0.94–1.73)§ | NA | NA | 32.6 | 95.6 | ①, ②, ③, ④, ⑥, ⑦ | Retrospective study | |
No stain use group (n = 639) | 78 (76–80) | 392 (61.3) | 23.3 (22.0–25.6)§ | 4.40 (3.86–5.16)§ | 2.77 (2.20–3.34)§ | 1.19 (0.98–1.42)§ | 1.23 (0.91–1.74)§ | NA | NA | 30.8 | 95.9 | |||||
Orkaby et al. [32], 2020 (USA) | 6.8 | Stain used group (n = 57178) | 81.2 (3.6) | 1544 (2.7) | 27.5 (4.3) | NA | NA | NA | NA | 70.9 | NA | 27 | 80.4 | ①, ②, ③, ④, ⑤ | Retrospective study | |
No stain use group (n = 326981) | 80.7 (4.0) | 8828 (2.7) | 26.7 (4.4) | NA | NA | NA | NA | 79.2 | NA | 13.1 | 66.2 | |||||
Rea et al. [33], 2020 (Italy) | 7 | Good clinical frailty group (n = 82782) | 73.0 (6.0) | 49249 (59.5) | NA | NA | NA | NA | NA | NA | NA | 13.5 | NA | ①, ② | Case-control study | |
Intermediate clinical frailty group (n = 175771) | 74.0 (6.0) | 96138 (54.7) | NA | NA | NA | NA | NA | NA | NA | 10.4 | NA | |||||
Poor clinical frailty group (n = 170483) | 76.0 (7.0) | 81331 (47.7) | NA | NA | NA | NA | NA | NA | NA | NA | NA | |||||
Very poor clinical frailty (n = 31424) | 76.0 (6.0) | 12801 (40.7) | NA | NA | NA | NA | NA | NA | NA | NA | NA | |||||
Zhou et al. [34], 2020 (Australia and USA) | 4.7 | Stain used group (n = 5629) | 74.2 (71.8–77.7) | 3413 (60.6) | NA | NA | NA | NA | NA | 45.4 | 75.6 | 19.6 | 82.4 | ①, ②, ③, ④, ⑤ | Retrospective study | |
No stain use group (n = 12467) | 74.2 (71.8–77.9) | 6727 (54.0) | NA | NA | NA | NA | NA | 44 | 78.3 | 6.1 | 70.8 | |||||
Lavie et al. [35], 2021 (Israel) | 5 | 76.9 (5.9) | 3699 (62) | NA | NA | NA | NA | NA | 26.5 | NA | NA | NA | ①, ⑤, ⑥ | Retrospective study | ||
Continuous data was presented as mean (standard deviation, SD); Dichotomous data
was presented as percentage. *Percentage of smoking and alcohol use was calculated by current + past use. †In TC, LDL-C, HDL-C, 1 mmol/L = 38.6 mg/dL. ‡In TG, 1 mmol/L = 86.8 mg/dL. §Data was presented as median (interquartile range). Abbreviations: BMI, body mass index; TC, total cholesterol; LDL-C, low density lipoprotein-cholesterol; HDL-C, low density lipoprotein-cholesterol; TG, triglyceride; DM, diabetes mellitus; HTN, hypertension; CVD, cardiovascular disease; NA, not available. Outcomes: ①, All-cause mortality; ②, CVD mortality; ③, Coronary heart disease (CHD)/Myocardial infraction (MI); ④, Stroke; ⑤, Total CV events; ⑥, New onsets on DM; ⑦, New onsets on cancer. |
Regarding the study quality by NOS, the average NOS score was 6.67. Among all 12 studies, there were 2 low quality studies [29, 30], 6 medium quality studies [26, 27, 31, 33, 34, 35], and 4 high quality studies [24, 25, 28, 32]. Limited by the life expectancy, most studies were lack of adequate follow-up periods (Supplementary Table 4). With ROBINS-I tool, there were 5 studies [28, 29, 31, 32, 33] of moderate overall bias and the others [24, 25, 26, 27, 30, 34, 35] were of serious overall bias (Supplementary Table 5).
Eleven studies on all-cause mortality showed that the risk was reduced by 46%
(HR: 0.54, 95% CI: 0.46–0.63; p
Forrest plots for the primary outcomes. CVD, cardiovascular disease; CHD/MI, coronary heart disease/myocardial infraction; HR, hazard ratio.
Five studies on CVD mortality displayed that the risk was reduced by 49% (HR:
0.51, 95% CI: 0.39–0.65; p
Eight studies on CHD/MI demonstrated that the risk was reduced by 17% (HR:
0.83, 95% CI: 0.69–1.00; p = 0.05) with significant heterogeneity
(I
Eight studies on stroke revealed that the risk was decreased by 21% (HR: 0.79,
95% CI: 0.68–0.92; p
As for Total CV events, there were 8 relevant studies. The risk was reduced by
25% (HR: 0.75, 95% CI: 0.66–0.85; p
Three studies on new onset of DM indicated that statin use had no significant
association with primary prevention on DM incidence (HR: 0.93, 95% CI:
0.75–1.16; p = 0.52). There was no significant heterogeneity
(I
Forrest plots for the secondary outcomes. HR, hazard ratio.
Two studies on cancer incidence illustrated that statin use had no significant
association with primary prevention on cancer incidence (HR: 0.98, 95% CI:
0.88–1.08; p = 0.66). There was no significant heterogeneity
(I
In the subgroup analyses, the reduced risk with statin use on all-cause
mortality primary prevention kept robust across all subgroups, including
Subgroup | No. of studies | HR (95% CI) on fixed-effects model | HR (95% CI) on random-effects model | Final pooled HR (95% CI) on fixed-effects model | Final pooled HR (95% CI) on random-effects model | I | |
All-cause mortality | |||||||
Age (y) | 11 | ||||||
3 | 0.44 (0.43–0.46) | 0.42 (0.32–0.55) | 0.66 (0.64–0.67) | 0.70 (0.60–0.80) | 95%, p | ||
3 | 0.58 (0.50–0.67) | 0.56 (0.44–0.71) | 61%, p = 0.08 | ||||
5 | 0.66 (0.64–0.67) | 0.70 (0.60–0.80) | 89%, p | ||||
1 | 0.85 (0.74–0.97) | 0.85 (0.74–0.97) | NA | ||||
Region | 11 | ||||||
North America | 4 | 0.65 (0.63–0.66) | 0.59 (0.48–0.72) | 0.51 (0.45–0.58) | 0.42 (0.28–0.62) | 72%, p = 0.01 | |
Europe | 4 | 0.50 (0.49–0.52) | 0.57 (0.45–0.73) | 98%, p | |||
Asia | 3 | 0.51 (0.45–0.58) | 0.42 (0.28–0.62) | 75%, p = 0.02 | |||
Diabetes proportion (%) | 9 | ||||||
4 | 0.64 (0.63–0.66) | 0.47 (0.35–0.65) | 0.60 (0.59–0.61) | 0.48 (0.40–0.59) | 87%, p | ||
5 | 0.46 (0.44–0.47) | 0.50 (0.40–0.64) | 97%, p | ||||
Hypertension proportion (%) | 6 | ||||||
4 | 0.65 (0.64–0.66) | 0.68 (0.59–0.79) | 0.65 (0.64–0.66) | 0.63 (0.55–0.73) | 85%, p | ||
2 | 0.53 (0.46–0.61) | 0.38 (0.16–0.88) | 80%, p | ||||
Follow-up (y) | 10 | ||||||
4 | 0.45 (0.43–0.46) | 0.44 (0.34–0.56) | 0.60 (0.60–0.61) | 0.54 (0.46–0.64) | 98%, p | ||
6 | 0.66 (0.64–0.67) | 0.68 (0.59–0.79) | 87%, p | ||||
Study design | 11 | ||||||
Prospective study | 2 | 0.52 (0.41–0.65) | 0.38 (0.15–0.96) | 0.60 (0.59–0.61) | 0.54 (0.46–0.63) | 86%, p | |
Retrospective study | 7 | 0.66 (0.65–0.67) | 0.70 (0.63–0.77) | 86%, p | |||
Nested case-control study | 1 | 0.55 (0.47–0.64) | 0.55 (0.47–0.64) | NA | |||
Case control study | 1 | 0.37 (0.36–0.39) | 0.35 (0.31–0.41) | 90%, p | |||
CHD/MI | |||||||
Age (y) | 8 | ||||||
2 | 0.99 (0.66–1.47) | 0.56 (0.11–2.93) | 0.97 (0.93–1.01) | 0.83 (0.69–1.00) | 85%, p | ||
2 | 0.66 (0.51–0.87) | 0.64 (0.30–1.35) | 87%, p | ||||
4 | 0.98 (0.94–1.02) | 0.91 (0.77–1.07) | 51%, p = 0.06 | ||||
1 | 0.87 (0.54–1.40) | 0.87 (0.54–1.40) | NA | ||||
Region | 8 | ||||||
North America | 4 | 0.97 (0.93–1.02) | 0.66 (0.41–1.06) | 0.97 (0.93–1.01) | 0.83 (0.69–1.00) | 86%, p | |
Europe | 2 | 0.87 (0.75–1.02) | 0.87 (0.71–1.08) | 30%, p = 0.22 | |||
Asia | 2 | 1.10 (0.75–1.02) | 0.57 (0.10–3.23) | 79%, p = 0.03 | |||
Diabetes proportion (%) | 7 | ||||||
4 | 0.99 (0.94–1.03) | 0.82 (0.53–1.25) | 0.98 (0.94–1.02) | 0.80 (0.59–1.08) | 73%, p = 0.01 | ||
3 | 0.78 (0.62–0.98) | 0.78 (0.44–1.40) | 85%, p | ||||
Hypertension proportion (%) | 8 | ||||||
6 | 0.97 (0.93–1.01) | 0.82 (0.67–1.00) | 0.97 (0.93–1.01) | 0.83 (0.69–1.00) | 71%, p | ||
2 | 1.05 (0.77–1.44) | 0.56 (0.11–2.89) | 86%, p | ||||
Follow-up (y) | 7 | ||||||
3 | 0.95 (0.73–1.24) | 0.81 (0.46–1.45) | 0.97 (0.93–1.01) | 0.79 (0.64–0.97) | 71%, p = 0.03 | ||
4 | 0.97 (0.93–1.01) | 0.75 (0.58–0.97) | 77%, p | ||||
Study design | 8 | ||||||
Prospective study | 3 | 0.95 (0.73–1.24) | 0.81 (0.46–1.45) | 0.97 (0.93–1.01) | 0.83 (0.69–1.00) | 71%, p = 0.03 | |
Retrospective study | 4 | 0.97 (0.93–1.01) | 0.75 (0.58–0.97) | 77%, p | |||
Nested case-control study | 1 | 1.18 (0.85–1.63) | 1.18 (0.85–1.63) | NA | |||
Stroke | |||||||
Age (y) | |||||||
2 | 0.58 (0.35–0.95) | 0.58 (0.35–0.95) | 0.95 (0.92–0.99) | 0.79 (0.68–0.92) | 0%, p = 0.97 | ||
2 | 0.64 (0.47–0.88) | 0.64 (0.47–0.88) | 0%, p = 0.79 | ||||
4 | 0.96 (0.93–0.99) | 0.85 (0.73–1.00) | 62%, p = 0.01 | ||||
1 | 1.02 (0.73–1.42) | 0.88 (0.44–1.76) | NA | ||||
Region | |||||||
North America | 4 | 0.97 (0.93–1.00) | 0.77 (0.56–1.05) | 0.95 (0.92–0.99) | 0.79 (0.68–0.92) | 60%, p = 0.06 | |
Europe | 2 | 0.87 (0.76–0.99) | 0.84 (0.68–1.03) | 42%, p = 0.14 | |||
Asia | 2 | 0.64 (0.49–0.83) | 0.64 (0.49–0.83) | 0%, p = 0.64 | |||
Diabetes proportion (%) | |||||||
4 | 0.96 (0.93–1.00) | 0.76 (0.53–1.07) | 0.96 (0.92–0.99) | 0.70 (0.53–0.91) | 72%, p = 0.01 | ||
3 | 0.63 (0.48–0.83) | 0.63 (0.48–0.83) | 0%, p = 0.91 | ||||
Hypertension proportion (%) | |||||||
6 | 0.96 (0.93–0.99) | 0.83 (0.72–0.97) | 0.95 (0.92–0.99) | 0.79 (0.68–0.92) | 53%, p = 0.03 | ||
2 | 0.64 (0.49–0.83) | 0.64 (0.49–0.83) | 0%, p = 0.84 | ||||
Follow-up (y) | |||||||
3 | 0.58 (0.40–0.86) | 0.58 (0.40–0.86) | 0.96 (0.92–0.99) | 0.83 (0.71–0.96) | 0%, p = 1.00 | ||
4 | 0.96 (0.93–1.00) | 0.88 (0.76–1.01) | 49%, p = 0.07 | ||||
Study design | |||||||
Prospective study | 3 | 0.58 (0.40–0.86) | 0.58 (0.40–0.86) | 0.95 (0.92–0.99) | 0.79 (0.68–0.92) | 0%, p = 1.00 | |
Retrospective study | 4 | 0.96 (0.93–1.00) | 0.88 (0.76–1.01) | 49%, p = 0.07 | |||
Nested case-control study | 1 | 0.65 (0.49–0.85) | 0.65 (0.49–0.85) | NA | |||
Total CV events | |||||||
Age (y) | 8 | ||||||
2 | 0.68 (0.49–0.93) | 0.53 (0.20–1.40) | 0.87 (0.85–0.89) | 0.75 (0.66–0.85) | 84%, p = 0.01 | ||
3 | 0.65 (0.55–0.77) | 0.65 (0.53–0.80) | 33%, p = 0.22 | ||||
3 | 0.88 (0.86–0.90) | 0.81 (0.71–0.93) | 78%, p | ||||
1 | 0.94 (0.71–1.25) | 0.94 (0.71–1.25) | NA | ||||
Region | 8 | ||||||
North America | 4 | 0.88 (0.86–0.90) | 0.67 (0.48–0.92) | 0.87 (0.85–0.89) | 0.75 (0.66–0.85) | 85%, p | |
Europe | 2 | 0.85 (0.77–0.95) | 0.84 (0.72–0.98) | 37%, p = 0.18 | |||
Asia | 2 | 0.68 (0.60–0.76) | 0.68 (0.60–0.76) | 0%, p = 0.49 | |||
Diabetes proportion (%) | 6 | ||||||
3 | 0.88 (0.86–0.90) | 0.70 (0.53–0.93) | 0.87 (0.85–0.89) | 0.71 (0.58–0.86) | 92%, p | ||
3 | 0.70 (0.59–0.83) | 0.71 (0.56–0.91) | 49%, p = 0.14 | ||||
Hypertension proportion (%) | 7 | ||||||
5 | 0.88 (0.86–0.90) | 0.81 (0.73–0.91) | 0.87 (0.85–0.89) | 0.76 (0.67–0.87) | 60%, p = 0.01 | ||
2 | 0.67 (0.59–0.75) | 0.49 (0.23–1.08) | 80%, p = 0.03 | ||||
Follow-up (y) | 7 | ||||||
3 | 0.74 (0.60–0.92) | 0.66 (0.42–1.03) | 0.88 (0.86–0.90) | 0.76 (0.67–0.87) | 71%, p = 0.03 | ||
4 | 0.88 (0.86–0.90) | 0.78 (0.68–0.90) | 73%, p | ||||
Study design | 8 | ||||||
Prospective study | 3 | 0.74 (0.60–0.92) | 0.66 (0.42–1.03) | 0.87 (0.85–0.89) | 0.75 (0.66–0.85) | 71%, p = 0.03 | |
Retrospective study | 4 | 0.88 (0.86–0.90) | 0.78 (0.68–0.90) | 73%, p | |||
Nested case-control study | 1 | 0.69 (0.61–0.77) | 0.69 (0.61–0.77) | NA | |||
*Ramos et al. [28] reported two groups about 85 in one study. †More than one groups about related data in one study. Abbreviations: HR, hazard ratio; 95% CI, 95% confidence interval; CHD/MI, coronary heart disease/myocardial infraction; CV events, cardiovascular events; NA, not available. Bold type, statistical significance. |
According to the GRADE approach, evidence for all-cause mortality and CVD mortality was rated as “very low”, and for CHD/MI, stroke, total CV events, DM incidence, and cancer incidence was rated as “low”. Details have been given in Table 3. We analyzed potential publication bias for all-cause mortality, including most eligible studies (11 studies), and no evidence of publication bias was found (Egger’s test p = 0.246). The effect estimate of all-cause mortality was visualized and improved by “trim-and-fill” method. After the trim-and-fill statistical process, the revised funnel plot seemed to be more symmetry (Fig. 4).
Outcomes | Risk of bias |
Inconsistency** | Indirectness | Imprecision† | Publication bias†† | Large effect | Dose response | Residual bias | Quality of evidence‡ |
All-cause mortality | Not serious | Very serious | Not serious | Not serious | Undetected | Undetected | Undetected | Undetected | |
Very low | |||||||||
CVD mortality | Serious | Very serious | Not serious | Not serious | Not available | Undetected | Undetected | Undetected | |
Very low | |||||||||
CHD/MI | Serious | Serious | Not serious | Serious | Not available | Undetected | Undetected | Undetected | |
Low | |||||||||
Stroke | Serious | Serious | Not serious | Not serious | Not available | Undetected | Undetected | Undetected | |
Low | |||||||||
Total CV events | Serious | Serious | Not serious | Not serious | Not available | Undetected | Undetected | Undetected | |
Low | |||||||||
DM incidence | Serious | Not serious | Not serious | Serious | Not available | Undetected | Undetected | Undetected | |
Low | |||||||||
Cancer incidence | Serious | Not serious | Not serious | Not serious | Not available | Undetected | Undetected | Undetected | |
Low | |||||||||
*Risk of bias of included studies were assessed by study number, NOS and
ROBINS-I tools. **Serious inconsistency indicated significant heterogeneity of 80% †Serious imprecision indicated the confidence intervals for pooled results were board (larger than 0.3). ††Publication bias was evaluated by Egger’s test, a p-value ‡If there was one “serious”, the evidence was “low” and if there was one “Very serious”, the evidence was “Very low”. Abbreviation: CVD, cardiovascular disease; CHD/MI, coronary heart disease/myocardial infraction; DM, diabetes mellitus. |
Funnel plot with fill-and-trim method. After trim-and-fill statistical process, the funnel plot seemed to be more symmetry.
By 2050, more than 45 million Americans will be 75 years or older, with a great proportional rate of 85 years and older people [3]. Evidence suggested that the incidence and prevalence of atherosclerotic cardiovascular disease (ASCVD) increases with age and keeps the leading cause of total mortality, disturbs the quality of life, and extends medical costs [9, 36]. Thus, proper management and care on those older populations are urgent. In our meta-analysis, it was found that statin use might be associated with a significant risk reduction on all-cause mortality, CVD mortality, CHD/MI, stroke and total CV events, and the reduced risks was 46%, 49%, 17%, 21% and 25%, respectively. Risk reduction in all-cause mortality keeps significant at higher ages regardless of diabetes as well as hypertension status. No significant association was found between statin use and diabetes incidence or cancer incidence. Briefly, there findings supported the positive correlation between statin use and CVD primary prevention in older population. Due to the observational nature, we still require further investigations to address the causality.
The beneficial role of statin use in all-cause mortality was consistent with the
results from former clinical trials, and statin preserved risk role of elevated
LDL-C in older people. A limitation of those trials was the limited sample size
in subgroups of
The aging people have a higher risk of drugs adverse events due to multiple
comorbidities, polypharmacy, and altered pharmacokinetics and pharmacodynamics.
The safety of statin in these people is a major of concern related to statin
therapy continuation. Many older people are companied with hypertension
especially in the high CVD risk populations. In current study, we revealed that
statin use links to reduced risks of all-cause mortality regardless of the
hypertension status, which implies that statin can be recommended to older people
suffering from mild CVD (less proportion of hypertension). In a meta-analysis of
more than 3 million older subjects, only 47.9% statin users were adherent to
therapy after one year of follow-up [42]. According to a current study, there was
no significant association between statin use and risk of DM or cancer incidence,
and such results were in line with the conclusion from previous randomized
controlled trials (RCTs) that investigated the primary prevention in older people
[43, 44, 45]. However, evidence that focuses on general mixed populations (including
both primary and secondary prevention) reported a 9% to 55% increased risk of
diabetes in statin-use participants compared with the no users [46]. Another
meta-analysis revealed that older statin-use participants were associated with
21% of decreased risk of T2DM compared with younger participants [47]. Based on
these results, statin-associated DM risk will be more obvious in people with
extremely high CVD risk such as extremely old people who have already suffered
from serious CVD, metabolic syndrome etc. [46, 48]. Older people are always
heterogeneous in many aspects (i.e., demographic characteristics, health and body
function). Unfortunately, these confounders are not well elaborated in RCTs
especially those with
When comparing with other similar studies, a recent meta-analysis incorporated
40 RCTs to investigate the efficacy and safety of statins for primary prevention
of CVD with 94,283 patients at different ages [50]. That study displayed that
statin use significantly reduced the risk of all-cause mortality (HR: 0.89, 95%
CI: 0.85–0.93) in the included populations [50]. However, no further data about
the elderly can be found. Another Bayesian analysis that calculated the available
data on older people (
Several limitations should be illustrated. Firstly, there is great heterogeneity among analyses on the primary outcomes, and the heterogeneity still exists in all-cause mortality by omitting high heterogeneous studies. We hypothesized that it might be caused by the inconsistent characteristics of older people in many aspects and the poor nature of observational studies. The results on CVD mortality, CHD/MI, stroke and total CV events are not significantly changed whether the studies of great heterogeneity were excluded or not. The second limitation is the poor quality of included observational studies whose average NOS was 6.67. The evidence on all-cause mortality and CVD mortality is evaluated as “very low”. Actually, there are only 4 high quality studies, final pooled results require more caution to be applied on clinical practice. Moreover, even though we found most of the results were robust, we performed sensitivity and subgroup analyses to try to find the source of heterogeneity. Thirdly, in terms of outcomes of interest, the definitions on CVD or CV events are various. We consistently pursue uniformed definitions on CVD and seek for individualized differences and commonalities among people, just as the guidelines’ requirement. It is suggested that further studies should be more precise on that. Finally, due to the nature of observational studies, we failed to draw strong causality, so we need to compare the results of meta-analysis based on observational studies and further RCTs with larger sample size and/or longer follow-up period. In that case, we will out forward more useful suggestions for the clinical duties and public health.
Statin use is useful for primary prevention for all-cause mortality, CVD mortality, CHD/MI, stroke and total CV events. The relevance keeps existing regardless of diabetes and hypertension status, and even older populations. Furthermore, no association was found for DM and cancer incidence. These findings supported that statin use is suitable for older people in primary prevention setting especially those with high CVD risk. Most importantly, considering the observational nature of evidence, more relevant trials should be conducted in older people.
All authors designed and conducted this study. HH—wrote the paper. HH, HZ and RY—helped design the study. HZ and RY—revised the statistical methodology. RY—assume primary responsibility for the final content. All authors read and approved the final manuscript.
The patients entered this study from included studies and had completed the informed consent of participants.
The authors thanked the primary authors who shared their original data in the publications and observational studies. The authors thanked researchers who established methodology on systematic review/meta-analysis and who created public R packages.
This research received no external funding.
The authors declare no conflict of interest.
See details in Appendix File 1.