Article Text

Download PDFPDF

Epidemiology
Does cigarette smoking exacerbate the effect of total cholesterol and high-density lipoprotein cholesterol on the risk of cardiovascular diseases?
  1. K Nakamura1,2,
  2. F Barzi2,
  3. R Huxley2,
  4. T-H Lam3,
  5. I Suh4,
  6. J Woo5,
  7. H C Kim4,
  8. V L Feigin6,
  9. D Gu7,
  10. M Woodward2,8
  1. 1
    Department of Epidemiology and Public Health, Kanazawa Medical University, Uchinada, Japan
  2. 2
    Nutrition and Lifestyle Division, The George Institute for International Health, Sydney, Australia
  3. 3
    Department of Community Medicine, University of Hong Kong, Hong Kong, People’s Republic of China
  4. 4
    Department of Preventive Medicine, Yonsei University College of Medicine, Seoul, Korea
  5. 5
    Division of Geriatrics, Department of Medicine and Therapeutics, Chinese University of Hong Kong, Hong Kong, People’s Republic of China
  6. 6
    National Research Centre for Stroke, Applied Neurosciences and Neurorehabilitation, Auckland University of Technology, Auckland, New Zealand
  7. 7
    Cardiovascular Institute and Fu Wai Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, People’s Republic of China
  8. 8
    Department of Medicine, Mount Sinai School of Medicine, New York University, New York, USA
  1. Dr K Nakamura, Department of Epidemiology and Public Health, Kanazawa Medical University, 1-1 Daigaku, Uchinada, Ishikawa 920-0293, Japan; knaka{at}kanazawa-med.ac.jp

Abstract

Objective: To explore whether an interaction between smoking and serum total cholesterol (TC) and/or decreased levels of serum high-density lipoprotein cholesterol (HDLC) exists for any major subtype of cardiovascular disease.

Design: An individual participant overview of 34 cohort studies.

Setting: The Asia-Pacific region.

Participants: People aged ⩾20 years without a particular condition or risk factor.

Mean outcome measures: Hazard ratios (HRs) and 95% confidence intervals (CIs) for both TC and HDLC by smoking status were estimated using Cox proportional hazard models adjusted for age and systolic blood pressure and stratified by study and sex.

Results: During follow-up (median 4.0 years), 3298 coronary heart disease (CHD) and 4318 stroke events were recorded. For CHD, the HR (95% CI) for an additional 1.06 mmol/l increment in TC was greater in current smokers than in non-smokers: 1.54 (1.43 to 1.66) versus 1.38 (1.30 to 1.47); p = 0.02. Similarly, the HR (95% CI) for an additional 0.40 mmol/l decrement in HDLC was greater in current smokers than in non-smokers: 1.67 (1.35 to 2.07) versus 1.28 (1.10 to 1.49); p = 0.04. The positive association of TC with ischaemic stroke, and the negative association of TC with haemorrhagic stroke, were broadly similar for current smokers and non-smokers. Similarly, the risks of both the subtypes of stroke remained broadly unchanged as HDLC decreased in both current smokers and non-smokers.

Conclusions: Smoking exacerbated the effects of both TC and HDLC on CHD, although no interaction between smoking and TC or HDLC existed for either of the subtypes of stroke.

Statistics from Altmetric.com

Cigarette smoking and raised levels of serum total cholesterol (TC) are the second and third most common causes of death in the world, respectively,1 largely owing to their major aetiological role in coronary heart disease (CHD) and ischaemic stroke (IS).28 In contrast to TC, serum high-density lipoprotein cholesterol (HDLC), an important subfraction of TC, is cardioprotective and high levels of HDLC have been associated with reduced risk of CHD and IS.7 9

Previous studies have suggested that the impact of TC or low-density lipoprotein cholesterol on cardiovascular risk may be exacerbated by smoking, although these findings have not been consistently reported.4 1016 Furthermore, there is even less certainty about if, and how, smoking may affect HDLC and its cardioprotective properties.11 13 16 17 Most previous studies have been restricted in their analysis by insufficient data to reliably examine these questions, and the data that do exist largely relate to Western populations with few studies conducted among Asian cohorts. This is of importance given the considerable differences in smoking prevalence and lipid characteristics between Western and Asian populations. Whereas smoking levels tend to be low, and on the decline, in most Western countries, the opposite is true for large parts of Asia; China is a case in point, where over 60% of men are current smokers.1820 By contrast, mean levels of TC tend to be higher in Caucasians than in Asians.2022 Clearly, then, even a small potentiation of the effect between lipids and cardiovascular risk by smoking would translate into big effects, given the high prevalence of smoking among men in Asia and the sheer magnitude of its population.

Hence, we examined the joint effects of smoking and both TC and HDLC using the large database of the Asia Pacific Cohort Studies Collaboration (APCSC); an individual participant data overview of prospective cohort studies in the Asia-Pacific region.

MATERIALS AND METHODS

Participating studies

Details of APCSC are described elsewhere.23 24 Briefly, APCSC includes 44 pre-existing cohort studies which had at least 5000 person-years of follow-up and recorded age, sex and blood pressure at baseline and vital status at the end of the follow-up. Studies were excluded from APCSC if enrolment was dependent upon having a particular condition or risk factor. Additionally, for analyses in this report only people aged ⩾20 years at study entry with information on TC, HDLC and smoking status were included.

Measurement of baseline variables

TC was measured from serum samples and these were obtained while fasting in the majority of participants. HDLC was also similarly measured in some studies. As each study in APCSC was initiated over a long period of time (1966–94), the methods and instruments used for measuring lipids could vary.8 9 Smoking status (current cigarette smoker/not) was self-reported at the time when the participant entered the study. Cohorts were classified as Asian if the participants were recruited from mainland China, Hong Kong, Japan, Korea, Singapore, Taiwan or Thailand, and as ANZ if the participants were from Australia or New Zealand. This classification largely represented a split by ethnicity into Asians and Caucasians.

Outcomes

All studies reported deaths by underlying cause; a subset of studies also reported non-fatal cardiovascular events.24 Outcomes were classified (recoded where necessary) according to the ninth revision of the International Classification of Diseases (ICD-9). Outcomes used in this report, including fatal and non-fatal events, were CHD (ICD-9: 410–414), HS (intracerebral haemorrhage) (431.0–432.9) and IS (433.0–434.9). Since most studies identified events using record linkage, verification of pathological types of stroke was not routinely reported. All data provided to the Secretariat were checked for completeness and consistency and recoded when necessary to maximise comparability across cohorts. Summary reports were referred back to principal investigators of each collaborating study for review and confirmation.

Statistical methods

Subjects were classified according to approximately equal fourths of TC and HDLC as measured at baseline. The range of TC across the quarters was ⩽4.3, 4–4.9, 5.0–5.6 and ⩾5.7 mmol/l, respectively, and the range of HDLC was ⩽1.0, 1.1–1.3, 1.4–1.5 and ⩾1.6 mmol/l, respectively. For each variable the hazard ratios (HRs) for each subtype of cardiovascular disease (CVD) were estimated using a Cox proportional hazard model with corresponding 95% confidence intervals (CIs) derived using floating absolute risks,25 with the baseline group being those in the bottom quarter for TC (or upper quarter for HDLC). The log-linearity of the associations of TC or HDLC with each subtype of CVD8 9 was investigated, fitting a Cox model with a continuous variable defining the fourths. HRs and 95% CIs for the outcomes of interest were estimated for a one standard deviation (SD) increase (1.06 mmol/l) in TC and for a 1 SD decrease (0.40 mmol/l) in HDLC by smoking status (current smoking or not) using Cox proportional hazard regression models adjusted by age and systolic blood pressure and stratified by study and sex.26 Results were adjusted for regression dilution bias.25 27 Repeat measurements of TC and HDLC taken after a median of 3 years were available for 18 844 and 9072 participants, respectively. Using a linear mixed regression model that accounted for the heterogeneity of variance between studies and within-subject correlation,28 regression dilution coefficients were estimated to be 1.7 for TC and 1.5 for HDLC. The interaction effect between each lipid variable and smoking status was assessed using likelihood ratio tests, comparing the models with main effects only with the models that included the interaction term.25 In addition to analyses of the overall APCSC, predefined subgroup analyses were performed by sex, region (Asia and ANZ) and age at risk (<65 and ⩾65 years).23 The statistical analysis was performed using SAS version 9.1 (SAS Institute Cary NC).

RESULTS

Characteristics of the study population

A total of 34 cohorts (77% of all studies in APCSC), involving 352 993 participants (76% Asians; 41% female) with a mean age of 47 years, had information on both TC and smoking status (table 1). Overall, 34% of participants were smokers at baseline, but smoking prevalence differed by sex and region: in Asia, 59% of men and 3% of women were smokers, compared with 23% and 14%, respectively, in ANZ. Smokers and non-smokers had the same mean age and TC (46 years; 4.92 mmol/l) in Asia, but in ANZ, current smokers were, on average, younger and had a higher TC than non-smokers: 46 years versus 51 years; 5.67 mmol/l versus 5.57 mmol/l, respectively. Of the 34 cohorts included in the TC analyses, 33 (18 in Asia), involving 76 974 participants, additionally had information on HDLC. Smokers had a lower HDLC than non-smokers in both regions: 1.30 mmol/l versus 1.38 mmol/l in Asia; 1.29 mmol/l versus 1.37 mmol/l in ANZ, respectively.

Table 1 Study population characteristics by smoking status at baseline

Cardiovascular outcomes

In total, there were 2 468 537 person-years of follow-up. The median follow-up was 4.0 years (4.0 years for current smokers and 4.9 years for non-smokers), but it was shorter in Asia (4.0 years for both current smokers and non-smokers) than in ANZ (8.4 years for both current smokers and non-smokers) (table 2). During follow-up, 3298 CHD (1044 in Asia) and 4318 stroke (2976 in Asia) fatal and non-fatal events were recorded. Diagnosis of ischaemic or haemorrhagic stroke was documented by CT/MRI/autopsy investigations in 51% of fatal and 61% of non-fatal strokes. Of these documented stroke events, 1007 (846 in Asia) were classified as ischaemic and 1501 (1183 in Asia) as haemorrhagic.

Table 2 Cardiovascular events by smoking status

Associations between TC, HDLC and CHD by smoking status

The HR for CHD increased log-linearly as TC increased in both current smokers and non-smokers (p values for log-linear trend <0.001) (fig 1). The positive association between TC and CHD was significantly stronger among participants who smoked than among non-smokers: the HRs (95% CIs) for CHD in the highest quarter of TC, compared with the lowest, were 2.02 (1.85 to 2.20) for current smokers and 1.65 (1.55 to 1.77) for non-smokers; p value for interaction  =  0.11. On a continuous scale, a 1 SD increment in TC (1.06 mmol/l) was associated with an approximately 50% increased risk of CHD among smokers compared with 40% in non-smokers (fig 2): HR = 1.54 (95% CI 1.43 to 1.66) versus HR = 1.38 (1.30 to 1.47); p value for interaction  =  0.02. This difference was seen within most of the subgroups previously defined by sex, region and age, but was only significant for men (fig 2). Further investigation of the individual studies suggests that there is no consistent sex difference in the interaction between smoking, TC and CHD and that the lack of a significant findings among women may have been due to the smaller number of events in this subgroup (data not shown).

Figure 1

Associations of (A) usual serum total cholesterol (TC) and (B) high-density lipoprotein cholesterol (HDLC) with coronary heart disease (CHD) by smoking status. The hazard ratios (HRs) for the lowest quarter of TC and the highest quarter of HDLC are fixed at 1.0, separately for current smokers and non-smokers. Analyses are adjusted by age and systolic blood pressure and stratified by study and sex. The vertical lines show 95% confidence intervals (CIs). The p values shown are for the test of interaction between each lipid variable and smoking status. The dashed and continuous lines represent current smokers and non-smokers, respectively. CHD, coronary heart disease.

Figure 2

Hazard ratios (HRs) associated with a one standard deviation increase in usual serum total cholesterol (1.06 mmol/l) for coronary heart disease (CHD), ischaemic stroke (IS) and haemorrhagic stroke (HS) in current smokers and non-smokers, by sex, region, age and overall. Analyses are adjusted by age and systolic blood pressure and stratified by study and sex. The horizontal lines (or widths of diamonds for overall results) show 95% confidence intervals (CIs). The p values shown are for the test of interaction between serum total cholesterol and smoking status. The dashed and continuous lines represent current smokers and non-smokers, respectively. ANZ, Australia and New Zealand.

The HR for CHD increased log-linearly as HDLC decreased in both current smokers and non-smokers (p values <0.001) (fig 1). Similarly, there was evidence to suggest that the inverse relationship between HDLC and CHD was stronger among smokers than among non-smokers, such that the HRs (95% CIs) for CHD in the lowest quarter of TC, compared with the highest, were 2.10 (1.72 to 2.57) for current smokers and 1.82 (1.35 to 2.44) for non-smokers; p value for interaction  =  0.43. On a continuous scale, a 1 SD decrement in HDLC (0.40 mmol/l) was associated with an approximately 70% increased risk of CHD in smokers, compared with 30% in non-smokers (fig 3): HR = 1.67 (95% CI 1.35 to 2.07) versus HR = 1.28 (1.10 to 1.49); p value for interaction  =  0.04. This effect was apparent in most of the predefined subgroups but was significant only in men and for people aged <65 years (fig 3).

Figure 3

Hazard ratios (HRs) associated with a one standard deviation decrease in usual serum high-density lipoprotein cholesterol (0.40 mmol/l) for coronary heart disease (CHD), ischaemic stroke (IS) and haemorrhagic stroke (HS), in current smokers and non-smokers, by sex, region, age and overall. Analyses are adjusted by age and systolic blood pressure and stratified by study and sex. The horizontal lines (or widths of diamonds for overall results) show 95% confidence intervals (CIs). The p values shown are for the test of interaction between serum high-density lipoprotein cholesterol and smoking status. The dashed and continuous lines represent current smokers and non-smokers, respectively. ANZ, Australia and New Zealand.

Sensitivity analyses

A series of sensitivity analyses were conducted to determine the robustness of these estimates. Analyses that excluded in turn, the Busselton study, which is the most influential study in APCSC, and the Korean Medical Insurance Corporation (KMIC) study, which is the largest study, gave similar HRs for both TC and HDLC as in the main analyses. For TC, analyses without the Busselton study show an HR (95% CI) for CHD associated with a 1 SD increase of 1.40 (1.28 to 1.52) for non-smokers and 1.55 (1.41 to 1.71) for smokers (p for interaction  =  0.08), whereas analyses without KMIC show an HR (95% CI) of 1.37 (1.28 to 1.46) for non-smokers and 1.50 (1.39 to 1.63) for smokers (p for interaction  =  0.06). For HDLC, the corresponding results without Busselton were 1.33 (1.12 to 1.59) and 1.77 (1.39 to 2.24) (p for interaction  =  0.05); KMIC did not have information on HDLC.

In the subsample of 100 575 people for whom information was available on TC and the use of antihypertensive drugs, 12% were recorded as using antihypertensive drugs, the HRs for CHD associated with a 1 SD increase in TC did not differ materially whether or not adjustment for use of antihypertensive drugs was made: the HR (95% CI) without the adjustment was 1.34 (1.24 to 1.44) for non-smokers and 1.58 (1.42 to 1.75) for smokers (p for interaction  =  0.01), and with the adjustment was 1.35 (1.26 to 1.46) for non-smokers and 1.58 (1.42 to 1.75) for smokers (p for interaction  =  0.02). Similarly in the subsample of 30 114 people with information on HDLC and the use of antihypertensive drugs, 8% were recorded as using antihypertensive drugs, the corresponding results without the adjustment were 1.32 (1.06 to 1.64) and 1.33 (0.95 to 1.87) (p for interaction  =  0.96), and with adjustment 1.26 (1.01 to 1.57) and 1.30 (0.93 to 1.83) (p for interaction  =  0.86). Similarly, no important differences were seen when performing similar sensitivity analyses for cholesterol-lowering drugs (possible use of which was recorded for only 5% of participants).

Some studies, involving 215 515 people (61%), in APCSC recorded non-fatal, as well as fatal, CHD events. In these pooled studies, the HRs were similar, both for TC and HDLC, when fatal events only or all events were analysed. For TC, the HRs (95% CIs) for a 1 SD increase were 1.42 (1.29 to 1.56) and 1.35 (1.26 to 1.46), respectively (p value for interaction  =  0.44); and for HDLC, 1.86 (1.44 to 2.39) and 1.31 (1.09 to 1.58), respectively (p value for interaction  =  0.03).

Associations between TC, HDLC and stroke by smoking status

The HR for IS increased log-linearly with higher levels of TC in both current smokers and non-smokers (p<0.001) (data not shown). There was no evidence to suggest that its positive association with TC was affected by smoking status (fig 2). The direction of the relationship between TC and HS was opposite to that for IS (ie, inverse) and largely similar in smokers and non-smokers (fig 2); however, there was evidence of log-linearity of association for non-smokers (p = 0.002) but not for current smokers (p = 0.18) (data not shown). There was also no indication that smoking modified the effects of TC on IS or HS risk among any of the prespecified subgroups.

There was no evidence of a significant association between HDLC and the risk of either IS and HS in both current smokers and non-smokers (data not shown). The risk for both IS and HS due to decreased levels of HDLC was similar for current smokers and non-smokers among any of the predefined subgroups, as well as overall (fig 3), so that there was no suggestion of any interactions.

DISCUSSION

Findings from this large study in the Asia-Pacific region suggest that cigarette smoking potentiates the harmful effects of TC and reduces the cardioprotective properties of HDLC, increasing the risk of CHD. For TC, although the effect on CHD risk associated with smoking was shown to be relatively small (greater risk among smokers by about 10 percentage points), the widespread popularity of smoking across large parts of Asia (up to 70% in some areas)1820 would suggest that the impact of TC on subsequent risk of CHD in the region may have been underestimated. Furthermore, many countries in Asia have, or are, undergoing a substantial shift in their traditional dietary habits owing to economic growth and increasing exposure to the West, and in particular the “Western” diet, largely characterised by its high fat content. As a result, the mean population level of TC in Asian populations has been steadily rising over the past couple of decades, particularly in urban areas.29 30 Our findings, which are supported by some (but not all) earlier observations,1014 therefore suggest that current projections as to the direct burden of CHD due to TC and smoking in the Asia-Pacific region may well have to be revised upwards to take into account the potential interaction between smoking and lipids.

One notable, very large, study conflicts with the current findings, in that no interaction between smoking and TC on CHD risk was reported.15 Most of the data upon which that analysis was based, however, were derived from Western populations, as opposed to the predominately Asian cohorts in our study. Further, the previous study did not examine the interaction between smoking and HDLC on coronary risk nor did it report on whether there was any evidence of an interaction between smoking and lipids on cerebrovascular disease.

Although CHD and IS share certain risk factors, including blood pressure, TC, and smoking, we have no evidence to indicate that the strong and direct association between TC (and HDLC) and IS is amplified in smokers as is the case for CHD, a result that is largely in agreement with previous reports.12 17 The reasons why the interaction may be specific to CHD are unclear but may be due to the different pathophysiology of stroke and coronary disease. Unlike CHD, which occurs owing to occlusion of the coronary arteries from atherosclerosis and thrombosis, the causes of IS are diverse and include narrowing/thrombosis of extracranial carotid arteries (largely due to atherosclerosis of these arteries), cardioembolism, small artery diseases (largely due to raised blood pressure and amyloid angiopathy) and other causes (eg, blood disease, inflammation, etc).31 Similarly, our findings did not suggest that the inverse association between TC and HS is modified by smoking, which does not appear to have been reported previously. The current findings indicate that combined interventions aimed at normalising the lipid profile and quitting smoking would have a greater than expected effect at reducing the burden of CHD and might be important in countries of the Asia-Pacific region where the smoking habit is still popular1820 and the lipid profile is worsening.29 30 Unfortunately, however, smoking cessation is relatively rare in Asia compared with the West.1820 32 In China, for example, only 4% and <1% of the male and female population are former smokers compared with 63% and 4% who currently smoke, respectively.19 A lack of understanding about the health hazards associated with smoking, especially for CHD,19 has been cited as one of the explanations for its continuing popularity. Similarly, the general level of awareness as to the role of lipids in CVD could be low in Asia, because among people with hypercholesterolaemia, the proportion of those who are aware of the condition is low in some Asian countries.21 Ideally, it would be preferable for large-scale, population-wide interventions to target both lipid levels and smoking; in resource-poor settings, this represents an enormous challenge.

Despite important strengths of the study with its large amount of data and its wide range of TC values (due to the different distributions of TC values between Asia and ANZ8), on which to explore the interaction between lipids and smoking on each subtype of CVD, there are several limitations in APCSC that merit discussion. First, APCSC has incomplete data on lipid methodology, which was not available from all studies. However, as the method of lipid measurement is associated with study centre, any potential confounding due to differences in methods would have been adjusted for when the analyses were stratified by study. Second, APCSC records a limited set of potential risk factors for CVD, thus excluding some which may, conceivably, explain the observed interaction for CHD. Third, as we have no information about changes in smoking status during follow-up, we are unable to examine what impact, if any, duration of smoking (overall exposure to smoking) has on the relationship between lipid profile and subtypes of CVD. Fourth, information on HDLC as well as on former smoking status was available in limited studies. Finally, only about 50–60% of stroke events were classified as ischaemic or haemorrhagic in origin on the basis of CT/MRI/autopsy findings, which could lead to potential bias from misclassification of stroke subtype.

In summary, smoking exacerbates the association between TC and the risk of CHD and mitigates the protective effects of HDLC on coronary risk; hence, the likely benefits of interventions that target one or both of these risk factors are likely to be greater than previously expected.

Appendix

The Asia Pacific Cohort Studies Collaboration

Executive committee

M Woodward (Chair), X Fang, DF Gu, R Huxley, Y Imai, TH Lam, WH Pan, A Rodgers, I Suh, HC Kim, H Ueshima

Participating studies and principal collaborators

Aito Town: A Okayama, H Ueshima, H Maegawa; Akabane: N Aoki, M Nakamura, N Kubo, T Yamada; Anzhen 02: ZS Wu; Anzhen: CH Yao, ZS Wu; Australian Longitudinal Study of Aging: Mary Luszcz; Australian National Heart Foundation: TA Welborn; Beijing Aging: Z Tang; Beijing Steelworkers: LS Liu, JX Xie; Blood Donors’ Health: R Norton, S Ameratunga, S MacMahon, G Whitlock; Busselton: MW Knuiman; Canberra-Queanbeyan: H Christensen; Capital Iron and Steel Company Hospital Cohort (CISCH): J Zhou, XH Yu; Capital Iron and Steel Company: XG Wu; Civil Service Workers: A Tamakoshi; CVDFACTS: WH Pan; Electricity Generating Authority of Thailand (EGAT): P Sritara; East Beijing: ZL Wu, LQ Chen, GL Shan; Fangshan Farmers: DF Gu, XF Duan; Fletcher Challenge: S MacMahon, R Norton, G Whitlock, R Jackson; Guangzhou: YH Li; Guangzhou Occupational: TH Lam, CQ Jiang; Hisayama: Y Kiyohara, H Arima, M Iida; Hong Kong: J Woo, SC Ho; Huashan: Z Hong, MS Huang, B Zhou; Kinmen: JL Fuh; Kounan Town: H Ueshima, Y Kita, SR Choudhury; Korean Medical Insurance Corporation: I Suh, SH Jee, IS Kim; Melbourne Cohort: G Giles; Miyama: T Hashimoto, K Sakata; Newcastle: A Dobson; Ohasama: Y Imai, T Ohkubo, A Hozawa; Perth: K Jamrozik, M Hobbs, R Broadhurst; Saitama: K Nakachi; Seven Cities: XH Fang, SC Li, QD Yang; Shanghai Factory Workers: ZM Chen; Shibata: H Tanaka; Shigaraki: Y Kita, A Nozaki, H Ueshima; Shirakawa: H Horibe, Y Matsutani, M Kagaya; Singapore Heart: K Hughes, J Lee; Singapore 92: D Heng, SK Chew; Six Cohorts: BF Zhou, HY Zhang; Tanno/Soubetsu: K Shimamoto, S Saitoh; Tianjin: ZZ Li, HY Zhang; Western Australian AAA Screenees: P Norman, K Jamrozik; Xi'an: Y He, TH Lam; Yunnan: SX Yao.

REFERENCES

Footnotes

  • Funding: This project has received support from a National Health and Medical Research Council of Australia programme grant (358395) and an unrestricted educational grant from Pfizer Inc. This study has been partially supported by a grant from the Korea Health 21 R&D Project, Ministry of Health and Welfare, Republic of Korea (grant no A040152).

  • Competing interests: MW has received honoraria from Pfizer Inc to speak on material related to this paper at conferences.

  • Ethics approval: Ethics committee approval obtained.

  • The sponsors had no influence on the design, analysis, or interpretation of results and took no part in the writing of this paper.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.