|Year : 2019 | Volume
| Issue : 2 | Page : 129-134
12-year follow-up study of the c-reactive protein in Iranian middle-aged women: Isfahan cohort study
Alireza Afshari-Safavi1, Sayed Mohsen Hosseini2, Mohammad Talaei3, Hamidreza Roohafza4, Nizal Sarrafzadegan4, Masoumeh Sadeghi5
1 Student Research Committee, Department of Statistics and Epidemiology, School of Health, Isfahan University of Medical Science, Isfahan, Iran
2 Skin Diseases and Leishmaniasis Research Center; Department of Epidemiology and Biostatistics, School of Health, Isfahan University of Medical Sciences, Isfahan, Iran
3 Saw Swee Hock School of Public Health, National University of Singapore, Singapore
4 Isfahan Cardiovascular Research Center, Cardiovascular Research Institute, Isfahan University of Medical Sciences, Isfahan, Iran
5 Cardiac Rehabilitation Research Center, Cardiovascular Research Institute, Isfahan University of Medical Sciences, Isfahan, Iran
|Date of Web Publication||8-May-2019|
Sayed Mohsen Hosseini
Department of Epidemiology and Biostatistics, School of Health, Isfahan University of Medical Sciences, Hezarjerib Street, Isfahan
Source of Support: None, Conflict of Interest: None
Background: As an important inflammatory marker, study of C-reactive protein (CRP) changes over time may lead to better identification of risk factors of cardiovascular disease. We evaluated the relationship between CRP changes and lifestyle, metabolic syndrome (MS) and body mass index (BMI) in middle-aged women and explored potential bias from attrition. Materials and Methods: We studied 1234 participants in the Isfahan cohort study – a longitudinal population-based study of adults older than 35 years living in urban and rural areas of three counties in central Iran. Data were collected every 6 years since 2001 (3 points). Random effects model was used to evaluate the effects of behavioural risk factors and MS on CRP, with pattern mixture model to account for cohort attrition. Results: Mean CRP levels decreased over time (Est = −0.066, P < 0.001). MS (Est = −0.195, P < 0.001), BMI (Est = 0.022, P < 0.001), physical activity (Est = −0.009, P = 0.002) and history of smoking (Est = −0.399, P = 0.002) were independently associated with increases in CRP. Pattern mixture model showed that CRP decreased in participants with monotone measurement (Est = −0.032, P < 0.001), as well as intermittent measurement (Est = −0.022, P < 0.001), with no association in participants who responded at all points (Est = −0.015, P = 0.083). Conclusion: In this study, the rate of changes in CRP level in middle-aged women over time was higher in participants who were irregularly measured than those who measured continuously. MS, BMI and physical activity may be related to the CRP changes over time in middle-aged women.
Keywords: C-reactive protein, metabolic syndrome, middle-aged women, missing data, pattern mixture model
|How to cite this article:|
Afshari-Safavi A, Hosseini SM, Talaei M, Roohafza H, Sarrafzadegan N, Sadeghi M. 12-year follow-up study of the c-reactive protein in Iranian middle-aged women: Isfahan cohort study. Adv Hum Biol 2019;9:129-34
|How to cite this URL:|
Afshari-Safavi A, Hosseini SM, Talaei M, Roohafza H, Sarrafzadegan N, Sadeghi M. 12-year follow-up study of the c-reactive protein in Iranian middle-aged women: Isfahan cohort study. Adv Hum Biol [serial online] 2019 [cited 2020 Oct 22];9:129-34. Available from: https://www.aihbonline.com/text.asp?2019/9/2/129/257819
| Introduction|| |
C-reactive protein (CRP) is an acute phase reactant, produced by the liver. Elevated level of CRP, as an inflammatory biomarker, can be identified with blood tests within a few hours after tissue injury or other cause of inflammation. Inflammation of the arteries is a risk factor for cardiovascular disease. It has been associated to an increased risk of arterial disease, heart attack, heart disease and stroke. Even without any symptoms of disease, a high level of CRP may independently suggest an increased risk for nearly all degenerative disorders, including cancer, cardiovascular disease and diabetes. The chronic inflammation behind an elevated level of CRP may be influenced by inappropriate dietary patterns, lack of physical activity, smoking, genetics and environmental factors. In addition, CRP levels have also been reported to differ by gender. Existing evidence also proposed that CRP level might be more strongly correlated with metabolic syndrome (MS) in women than in men.
As women age and undergo menopause, they may experience a wide range of chronic diseases such as cardiovascular diseases. The reasons for the high prevalence of health-related problems in middle-aged women are multifactorial, and more studies can improve their quality of life in the old age. In a cross-sectional study, McDade et al. showed that waist circumference was the strongest anthropometric predictor of elevated CRP in Filipino women.
Several cross-sectional studies have demonstrated significant associations between CRP levels and other risk factors for cardiovascular diseases such as blood pressure, triglyceride levels, high-density lipoprotein cholesterol levels, low-density lipoprotein and total cholesterol.,,, However, there is little information to understand the relationship between longitudinal changes in CRP and other risk factors of cardiovascular disease, with genetic, environmental factors or lifestyle interventions. A longitudinal study conducted by Camhi et al. to assessing the effect of lifestyle intervention on CRP showed a significant decrease in CRP levels among women with MS, but not in men. However, Heilbronn et al. did not find a significant association between CRP levels with change in lipids or glucose in response to a low-fat diet weight loss program.
Although a longitudinal study may provide a relevant picture of patients' follow up, however, one of the issues with these studies is missing data. Participants experiencing deteriorating health because of progressive disease are more likely to discontinue the follow-up. As a result, more data are missing in these participants compared with those who continue the follow-up and lead to participants with various response patterns. A commonly used approach is analysis based on completers (who complete the follow-up survey) only and exclude the participants with missing data from the final analysis. However, often, the probability of missingness on a variable is related to the values of that variable itself, even after controlling for other variables. In this situation, the missing measurement itself is informative, and estimates obtained from complete case analysis may be biased. To address this, two different approaches to analysis of longitudinal data were introduced by Little and Rubin (1993, 1994): selection modelling and pattern mixture model (PMM). Recently, pattern mixture models have reached considerable interest due to some dominant features., Pattern mixture models can be applied to evaluate the contribution of missingness to the outcome variable by including missing data patterns as a predictor in model.
To assess the changes in CRP level over 12-year follow-up in Iranian middle-aged women and to test the association between behavioural risk factors of cardiovascular disease and MS with CRP changes, we present a linear mixed effect model (LMEM) and a PMM with multiple imputation approach as sensitivity analysis to check whether various response patterns lead to different results or not.
| Materials and Methods|| |
Participants and measurements
Participants were women aged 35–60 years participating in the Isfahan cohort study (ICS) – a longitudinal population-based study of adults aged 35 years old or more, living in urban and rural areas of three counties in central Iran. The ICS has been started since 17 years ago (2001) as a part of baseline study of Isfahan Healthy Heart Program, and each of the participants was followed every 6 years. All participants gave their informed consent for inclusion before they participated in the study. Ethical approval was obtained from the Ethics Committee of Isfahan Cardiovascular Research Centre, a World Health Organization collaborating centre. All participants were assessed for demographic characteristics, medical history, related lifestyle behaviours (including smoking, physical activity and dietary behaviour) and other risk factors. Smoking status (current smokers/nonsmokers), physical activity (mean hours per week) and dietary behaviour (according to the Global Dietary Index) were determined by self-report. Body mass index (BMI) was calculated as weight (kg) divided by height squared (m2).
The primary objective of this study was to determine the relationship between changes in CRP levels with BMI, MS and lifestyle behaviours including smoking, physical activity and nutritional habits in middle-aged women over the 12-year study time frame. CRP was measured using the immunoturbidimetry method with autoanalyser (Hitachi 902) with Pars Azmun kit (Iran). MS was calculated using adult treatment panel (ATP III) criteria over an average follow-up of 12 years. According to ATP III criteria, MS is defined as having at least three of the five criteria: (1) waist circumference >102 cm in men and >88 cm in women, (2) serum triglyceride >150 mg/dl, (3) high-density lipoprotein <40 mg/dl in men and <50 mg/dl in women, (4) fasting blood sugar >110 mg/dl and (5) blood pressure >130/85 mmHg.
All variables were measured at baseline in 2001 and at 6 and 12 years of follow-up. However, CRP test for some of the participants was missed on one or more measurement times. Information on CRP measurement was not fully recorded, and only if needed, this test was conducted. Hence, it is thought that patients with a worse health status were more likely to be measured for CRP. Accordingly, we classified participants into three groups based on the order of CRP measurements: continuous measurement (who measured at all times due to a worse health status), intermittent measurement (who did not have a stable health status) and monotone measurement (who first had a bad health status but then got better) [Table 1].
|Table 1: Pattern of the participant's C-reactive protein measurements in Isfahan cohort study|
Click here to view
We evaluated whether different patterns of missing data were associated with different relationships between CRP and behavioural risk factors of cardiovascular disease and MS.
We applied a LMEM with variance components' covariance structure for analysing the CRP levels, with time, BMI, MS, physical activity (activity), dietary behaviour (diet) and history of smoking (smoke) as fixed effects and individual subject as random effect. Because the distribution of CRP levels is asymmetric, the natural logarithm of CRP was used as the dependent variable. In this study, many CRP measurements contained missing values. The LMEM assumes that the missing data are missing at random meaning that the mechanism of the missingness is independent of the unobserved values. If the missingness is informative (if missingness is related to patient health status), then analysis may be biased. A PMM with multiple imputation approach was also used to analyse the CRP as sensitivity analyses, to adjust for bias that may potentially occur by patient attrition. Under this approach, different patterns of missing data are identified to explicitly model the missing data distribution.,
First, the missing data in intermittent and monotone patterns were imputed in multiple versions, assuming that dropouts are missing not at random. Then, imputed data were pooled with completers and a LMEM was fitted again. Mean changes from baseline were estimated and compared between different patterns. The LMEM and PMM were fitted in R software (version 3.3.3, The R Foundation for Statistical Computing,. Vienna, Austria) using the lme4 and mice packages. P < 0.05 was considered as statistically significant.
| Results|| |
In ICS, the overall percentage of missing observations for CRP was 24.6%. Only 27.5% of participants had available CRP data at all three waves (continuous measurement). The most common missing data pattern was intermittent (43.8%), and 28.7% were monotone.
Of 1234 participants included in the study, 238 (19.3%) had MS at baseline, increasing to 51.9% at 12 years. Mean CRP increased during the 1st 6-year period (1.84 ± 0.51) and then decreased during the 2nd 6-year period (2.12 ± 0.14). The total changes from the baseline in CRP level were 0.28 ± 0.65. More details on descriptive statistics of the study participants are presented in [Table 2] for each time point.
The estimated mean CRP level changes from baseline through 12 years with and without regard to the measurement patterns are shown in [Figure 1]. All measurement patterns showed a trend consistent with changes in CRP level in all individuals without imputation. Participants with continuous measurements were consistently lower than the other groups over the first 6 year.
|Figure 1: Estimated mean C-reactive protein changes from baseline over time. All, all participants regardless of the pattern of C-reactive protein measurement and without imputation. Continuous, who have been measured for C-reactive protein at all time. Monotone, participants with monotone measurement pattern and with imputation. Intermittent, participants with monotone measurement pattern and with imputation.|
Click here to view
The results from LMEM and PMM are displayed in [Table 3]. Findings from the LMEM analysis suggest that, except for dietary habits, all other predictors were independently associated with either positive or negative changes in CRP. Mean change from baseline in CRP level indicated significant decrease (−0.066, P < 0.001). An increase in physical activity was independently associated with decrease in CRP changes. Middle-aged women without MS had more decrease in CRP than those with MS (−0.195, P < 0.001). Furthermore, compared to the nonsmoker, participants with a history of smoking have experienced more increase in CRP by 0.399 points (P = 0.002).
|Table 3: Model estimates and standard error for random effects model and pattern mixture model|
Click here to view
After missing data imputation, random effect PMM has revealed the same outcomes. Imputed measurements had less decrease in CRP than unimputed measurements over time (−0.027, P = 0.018). In contrast, the relationship between dietary and history of smoking was not associated with CRP level changes.
Results from the LMEM for each of the patterns separately are summarised in [Table 4]. For continuous measurement, only BMI and dietary habit were significantly associated with change in CRP. One unit increase in BMI is associated with a 0.046 unit increase in mean changes of the natural logarithm of CRP. However, there was an inverse association between CRP level and dietary habits (−0.394, P = 0.025), which signifies that increase in global dietary index (GDI) score (higher GDI shows worse dietary habits) leads to decrease in CRP changes.
|Table 4: Parameter estimates of random effects model by adjusting on patterns of the participant's C-reactive protein measurements|
Click here to view
For participants who were in monotone pattern, the effect of BMI on CRP level changes was less than continuous pattern (0.018, P = 0.026). Compared to the continuous measurement pattern, the CRP level has changed over time in this group of patients (−0.032, P < 0.001).
Under intermittent pattern, which was the most common type of missingness pattern in this study, all predictors were associated with changes in CRP level.
| Discussion|| |
In this paper, the pattern mixture approach as a method for correction for informative missingness was studied. In studies with a lot of missing values, it is important to explore whether the missingness is informative or not. We applied this approach on data from ICS to study the longitudinal changes in CRP level among middle-aged women. We focused on lifestyle features, BMI and MS as risk factors of cardiovascular disease. We compared the results of the pattern mixture approach to the results of the commonly used random effects model.
Results of all participants showed that, on average, the CRP level decreased over time. Given the measurement pattern, we found the same results for participants with intermittent measurement, a slight decreasing in CRP level over time for participants with monotone measurement, but no significant association was seen with participants who completed all CRP measurements. The results of the LMEM under intermittent pattern were consistent with the findings of LMEM without imputation because the participants with intermittent measurement were about 75% of the total number of participants.
Under LMEM, we observed a significant difference in CRP between participants with and without MS. However, this relationship was not significant for participants with continuous measurement when LMEM was applied by adjusting on patterns of the participant's CRP measurements. Camhi et al. found that plasma CRP levels decreased significantly in response to a diet and a diet plus physical activity intervention in women with MS. In our study, mean change in CRP was greater in participants with MS than in non-MS. Au et al. also showed that high CRP levels are associated with MS.
BMI was significantly associated with change in CRP under both approaches. The mean change in CRP was doubled, respectively, from 0.022 and 0.018 in the intermittent and monotone up to 0.046 in the continuous. Data from the National Longitudinal Study of Adolescent Health demonstrated that changes in BMI during the transition from adolescence to adulthood were associated with greater odds of inflammation than current (age 27 years) or past (age 15 years) BMI alone. For schoolchildren, BMI was the most powerful predictor of serum concentrations of CRP. It may be an important factor to control body weight to prevent an increase in serum CRP in children and to help the prevention of chronic diseases. Others have found comparatively stronger associations between BMI gain or loss with CRP increase or decrease than current BMI.,
Except for participants with intermittent measurement, physical activity was not shown to be a significant predictor of CRP changes. The present study showed that an increase in physical activity may decrease the changes in CRP over time. Sadeghipour et al. reported no significant relation between CRP and physical activity. In a longitudinal study, the average changes of CRP were not associated to average current physical activity or to natural changes in current physical activity across the visits.
The association between dietary habits and CRP level was not significant under LMEM as well as PMM. However, this association was different across patterns of measurement. Among participants with continuous or intermittent measurement, a good dietary habits led to a decrease in CRP level, while no association was observed for monotone measurement. Boylan et al. observed that the high glycaemic load in dietary habits seems to increase CRP concentrations which, in turn, may lead to increased CVD risk. In a study in Korea, CRP changes related to the food group consumption patterns.
In this study, the results of the PMM with multiple imputations were consistent with the findings of the commonly used random effects model; however, the standard error estimates from PMM tended to be slightly larger for some predictors, implying that the results should be interpreted with caution. Post et al. showed that if the standard errors are larger for PMM, then the missing at random assumption within the patterns must be checked by sensitivity analysis.
This study showed that the CRP level in middle-aged women over follow-up time has been decreased. The rate of these changes was more in participants who were irregularly measured for CRP than those who measured at all time, that can be related to their health status. Dietary behaviour and BMI were the only predictors that influenced the CRP changes among participants with continuous measurement. Unlike other participants, changes in CRP over time were stable for participants who were measured at all times.
| Conclusion|| |
PMM adjusted the estimates to a certain degree and led to partly different results from the random effects model. Hence, this approach might be considered in the analysis of longitudinal data with large amount of informative missing data.
Financial support and sponsorship
This research has been supported by the Isfahan University of Medical Science (Grant number: 396596).
Conflicts of interest
There are no conflicts of interest.
| References|| |
Montecucco F, Mach F. New evidences for C-reactive protein (CRP) deposits in the arterial intima as a cardiovascular risk factor. Clin Interv Aging 2008;3:341-9.
Min JJ, Nam K, Kim TK, Kim HJ, Seo JH, Hwang HY, et al.
Relationship between early postoperative C-reactive protein elevation and long-term postoperative major adverse cardiovascular and cerebral events in patients undergoing off-pump coronary artery bypass graft surgery: A retrospective study. Br J Anaesth 2014;113:391-401.
Kopp HP, Kopp CW, Festa A, Krzyzanowska K, Kriwanek S, Minar E, et al.
Impact of weight loss on inflammatory proteins and their association with the insulin resistance syndrome in morbidly obese patients. Arterioscler Thromb Vasc Biol 2003;23:1042-7.
Young D, Camhi S, Wu T, Hagberg J, Stefanick M. Relationships among changes in C-reactive protein and cardiovascular disease risk factors with lifestyle interventions. Nutr Metab Cardiovasc Dis 2013;23:857-63.
Ye X, Yu Z, Li H, Franco OH, Liu Y, Lin X. Distributions of C-reactive protein and its association with metabolic syndrome in middle-aged and older chinese people. J Am Coll Cardiol 2007;49:1798-805.
Devaraj S, Singh U, Jialal I. Human C-reactive protein and the metabolic syndrome. Curr Opin Lipidol 2009;20:182-9.
Post WJ, Buijs C, Stolk RP, de Vries EG, le Cessie S. The analysis of longitudinal quality of life measures with informative drop-out: A pattern mixture approach. Qual Life Res 2010;19:137-48.
McDade TW, Rutherford JN, Adair L, Kuzawa C. Adiposity and pathogen exposure predict C-reactive protein in filipino women. J Nutr 2008;138:2442-7.
Kawamoto R, Kusunoki T, Abe M, Kohara K, Miki T. An association between body mass index and high-sensitivity C-reactive protein concentrations is influenced by age in community-dwelling persons. Ann Clin Biochem 2013;50:457-64.
Nash DT. Relationship of C-reactive protein, metabolic syndrome and diabetes mellitus: Potential role of statins. J Natl Med Assoc 2005;97:1600-7.
Ridker PM; JUPITER Study Group. Rosuvastatin in the primary prevention of cardiovascular disease among patients with low levels of low-density lipoprotein cholesterol and elevated high-sensitivity C-reactive protein: Rationale and design of the JUPITER trial. Circulation 2003;108:2292-7.
Ridker PM, Rifai N, Rose L, Buring JE, Cook NR. Comparison of C-reactive protein and low-density lipoprotein cholesterol levels in the prediction of first cardiovascular events. N Engl J Med 2002;347:1557-65.
Camhi SM, Stefanick ML, Ridker PM, Young DR. Changes in C-reactive protein from low-fat diet and/or physical activity in men and women with and without metabolic syndrome. Metabolism 2010;59:54-61.
Heilbronn LK, Noakes M, Clifton PM. Energy restriction and weight loss on very-low-fat diets reduce C-reactive protein concentrations in obese, healthy women. Arterioscler Thromb Vasc Biol 2001;21:968-70.
Little RJ, Rubin DB. Statistical Analysis with Missing Data. New York: John Wiley & Sons; 2014.
Carpenter J, Kenward M. Multiple Imputation and its Application. New York: John Wiley & Sons; 2012.
Kinnaird T, Anderson R, Ossei-Gerning N, Gallagher S, Large A, Strange J, et al.
Vascular access site and outcomes among 26,807 chronic total coronary occlusion angioplasty cases from the british cardiovascular interventions society national database. JACC Cardiovasc Interv 2017;10:635-44.
Lee MS, Lai CJ, Yang FY, Su HH, Yu HL, Wahlqvist ML. A global overall dietary index: ODI-R revised to emphasize quality over quantity. Asia Pac J Clin Nutr 2008;17 Suppl 1:82-6.
Sarrafzadegan N, Talaei M, Sadeghi M, Kelishadi R, Oveisgharan S, Mohammadifard N, et al.
The isfahan cohort study: Rationale, methods and main findings. J Hum Hypertens 2011;25:545-53.
Grundy SM, Brewer HB Jr., Cleeman JI, Smith SC Jr., Lenfant C; American Heart Association. Definition of metabolic syndrome: Report of the national heart, lung, and blood institute/American Heart Association conference on scientific issues related to definition. Circulation 2004;109:433-8.
Liew HP. A random-effects pattern-mixture model with application to comorbidity. Bull Sociol Methodol 2016;129:78-93.
Kenward MG. Pattern-Mixture Model. Chichester: Wiley Statsref: Statistics Reference Online; 2016.
Au B, Smith KJ, Gariépy G, Schmitz N. The longitudinal associations between C-reactive protein and depressive symptoms: Evidence from the english longitudinal study of ageing (ELSA). Int J Geriatr Psychiatry 2015;30:976-84.
Attard SM, Herring AH, Howard AG, Gordon-Larsen P. Longitudinal trajectories of BMI and cardiovascular disease risk: The national longitudinal study of adolescent health. Obesity (Silver Spring) 2013;21:2180-8.
Sadeghipour HR, Rahnama A, Salesi M, Rahnama N, Mojtahedi H. Relationship between C-reactive protein and physical fitness, physical activity, obesity and selected cardiovascular risk factors in schoolchildren. Int J Prev Med 2010;1:242-6.
Yatsuya H, Jeffery RW, Langer SL, Mitchell N, Flood AP, Welsh EM, et al.
Changes in C-reactive protein during weight loss and the association with changes in anthropometric variables in men and women: LIFE study. Int J Obes (Lond) 2011;35:684-91.
Fogarty AW, Glancy C, Jones S, Lewis SA, McKeever TM, Britton JR. A prospective study of weight change and systemic inflammation over 9 y. Am J Clin Nutr 2008;87:30-5.
Rawson ES, Freedson PS, Osganian SK, Matthews CE, Reed G, Ockene IS, et al.
Body mass index, but not physical activity, is associated with C-reactive protein. Med Sci Sports Exerc 2003;35:1160-6.
Boylan S, Welch A, Pikhart H, Malyutina S, Pajak A, Kubinova R, et al.
Dietary habits in three central and Eastern European countries: The HAPIEE study. BMC Public Health 2009;9:439.
Yeo R, Yoon SR, Kim OY. The association between food group consumption patterns and early metabolic syndrome risk in non-diabetic healthy people. Clin Nutr Res 2017;6:172-82.
[Table 1], [Table 2], [Table 3], [Table 4]