Elderly People With Low Body Weight May Have Subtle Low-grade Inflammation

Authors


(keinaka@josai.ac.jp)

Abstract

Low-grade inflammation, which plays important roles in the development of fatal diseases, is commonly observed in obese people. However, this has not been evaluated in lean people, who have relatively increased mortality risk compared with people of normal weight. Here, we elucidate the association between systemic low-grade inflammation and low body weight, with particular emphasis on aging. We examined the relationship between circulating C-reactive protein (CRP) and BMI in a cross-sectional study of 2,675 apparently healthy adults who had undergone a medical check-up. Overall, subjects with low BMI (<21.0 kg/m2, n = 585) showed a favorable cardiovascular profile without being undernourished. In the elderly (≥55 years old), logarithmic CRP (LogCRP) showed a sigmoid curve against BMI with a base at BMI 21.0–22.9 kg/m2, but not against waist circumference (WC), even in nonsmokers. In contrast, in middle-aged people, LogCRP showed an almost linear relationship with both BMI and WC. LogCRP levels in elderly nonsmokers with low BMI, but not normal or high BMI, were significantly higher than those in middle-aged with corresponding BMI (P < 0.05). After adjustment for age, sex, smoking status, and weight change over the past 2 years, the adjusted means of LogCRP still had a similar sigmoid curve against BMI in the elderly. These results suggest that elderly people with low body weight may have subtle low-grade inflammation irrespective of a favorable cardiovascular risk, which remains to be confirmed in further studies.

Introduction

Many large prospective studies have provided evidence that both underweight and obesity are associated with increased all-cause mortality with a U- or J-shaped association curve (1,2,3,4,5,6,7,8,9). Clinical and molecular biological studies of obesity have been, or are being, carried out worldwide over recent decades. Such studies have explored the pathogenesis of obesity and obesity-related diseases such as insulin resistance, type 2 diabetes, and cardiovascular disease (CVD), along with detailed investigation of the underlying cellular and molecular mechanisms and integrated clinical concepts such as metabolic syndrome (10).

In contrast, the mechanisms underlying the relationship between low BMI, including underweight, and increased mortality are poorly understood, although latent diseases, malnutrition, smoking, aging, or weight loss, or a combination of these have been considered to be involved in the disease etiology (2,3,4,5,6,8). In addition, an attenuated association in elderly people between obesity and increased mortality as well as fatal diseases has been recently proposed (1,3,5,6,7,8,9), suggesting that advancing age substantially interferes with the association between obesity and the development and progression of fatal disease.

Meanwhile, chronic inflammation is also thought to play a key role in the development and progression of diseases such as CVD and cancer (11,12,13). Furthermore, advancing age generally aggravates the systemic low-grade inflammatory status (14). Even subtle systemic inflammation can be detected using high sensitive measurement of circulating C-reactive protein (CRP), which is also associated with central obesity, type 2 diabetes, cardiovascular event, and even cancer risk (11,12,13,15,16).

Although numerous cross-sectional studies have shown significant associations between obesity, i.e., high BMI, and an increased circulating CRP level in the general healthy population as well as in patients with obesity-related diseases (15,16), there are no reports evaluating the relationship between low BMI and the circulating CRP level.

In this study, we examined the relationship between low BMI and circulating CRP level in a population-based cross-sectional study. As extensive data on the association between anthropometric measurements and mortality as well as fatal diseases have been evaluated mostly using BMI instead of waist circumference (WC) (1,2,3,4,5,6,7,9), we addressed this issue from the viewpoint of BMI. Nevertheless, for the comparison of BMI with WC, we also stratified subjects according to WC, which better reflects visceral adiposity and is also associated with metabolic abnormalities (17,18). Furthermore, as recent weight change is closely associated with chronic inflammation, coexisting diseases, and nutrition (19,20,21), we took into consideration changes in body weight over the past 1–2 years.

Methods and Procedures

Subjects

This study is based on composite research that was conducted to elucidate lifestyle-related diseases in collaboration with Josai University, Sakado, Japan and Social Insurance Omiya General Hospital, Saitama, Japan (starting in 2006 and ongoing). The protocol was approved by the Ethics Committee of Josai University and the committee of the Hospital. We randomly recruited apparently healthy subjects aged 30–80 years old from those who underwent a detailed medical check-up at the Omiya General Hospital. The medical check-up is voluntary and consists of tests such as hemogram, urine test including urinary sediments, occult blood test with stool samples, stomach X-rays, abdominal echo, serum tumor makers, serum rheumatoid factor, and, if necessary, chest/brain computed tomography, in addition to the annual regular check-ups. The recruited subjects gave informed consent. They had no self-reported medical history of diseases such as CVD and cancer and responded to a questionnaire about their lifestyle characteristics. Asymptomatic stable subjects with cardiovascular risks, such as hypertension, hyperlipidemia, hyperuricemia, or diabetes, were included, whether they were treated with medication or not. Individuals treated with an HMGCoA reductase inhibitor were excluded. Subjects with stable well-controlled coexisting diseases such as gastritis, gallbladder/renal stones, collagen diseases including inactive rheumatism, allergic diseases, and orthopedic diseases, were included. Subjects with CRP levels >10.0 mg/l and those suspected of having cancer by several tests were excluded. Consequently, 2,675 subjects were included in this study and had complete data for the cross-sectional analysis.

Anthropometric and laboratory measurements

Blood tests, anthropometric tests, and blood pressure tests were carried out after overnight fasting. Height and weight were measured objectively (cm and kg) to one decimal place using a digital electronic scale, while wearing light clothing without shoes. WC was measured (cm) to one decimal place at the height of the navel level at the end of a slight expiration. BMI was calculated as weight (in kg) divided by height (in m2). Subjects were divided into six BMI categories (≤18.9, 19.0–20.9, 21.0–22.9, 23.0–24.9, 25.0–26.9, ≥27.0 kg/m2). We defined low BMI as a BMI <21.0 kg/m2, normal BMI as 21.0–24.9 kg/m2, and high BMI as ≥25.0 kg/m2. As the percentage of obese subjects (BMI ≥30.0 kg/m2) was small (3.7%), the overweight subjects with a BMI of 27.0–30.0 (8.4%) and obese subjects were grouped together. We also checked the body weight at 1 and 2 years before the study, which had been objectively measured previously and recorded on the subject's medical sheet. For the evaluation of central obesity, subjects were also subdivided into categories according to WC (≤72.5, 72.6–77.5, 77.6–82.5, 82.6–87.5, 87.6–92.5, ≥92.6 cm), each of which is approximately equivalent to each category of BMI based on a linear regression equation obtained from current results (y = 2.42 × x + 26.36; y: WC; x: BMI; adjusted R2 = 0.74). Low-density lipoprotein-cholesterol concentration was assessed by direct measurement (Kyowa Medex, Tokyo, Japan). High-density lipoprotein-cholesterol and triglyceride were enzymatically measured using an autoanalyzer. Uric acid and fasting plasma glucose were also measured using standard methods. Plasma CRP concentrations were determined with an automated high-sensitivity immunoassay (CRP-HSII, Wako, Osaka) with a detection level of 0.10 mg/l and run-to-run coefficient of variation <3.0%.

Data analysis

Data are expressed as mean ± s.d. As the distribution of CRP levels and weight changes were highly skewed, they were expressed as the geometric mean and median, respectively. Log10-transformed values of CRP (LogCRP) were used for the parametric analysis, i.e., unpaired t-test for the difference between the middle-aged and the elderly. We also divided subjects into two categories: middle-aged (<55 years old) and elderly (≥55 years old), because 55 years old is equivalent to midlife and women are in a postmenopause state. Clinically important annual weight loss was defined as a loss of ≥4% of body weight (19,20).

Significant differences in the prevalence of men, current smoking, performing regular exercise, and clinically important weight loss were examined by χ2-test. LogCRP was tested using a two-way analysis of covariance, with BMI category as one factor and age group as a second factor with Bonferroni adjustments. Estimated marginal means of LogCRP were calculated using generalized linear models after adjustment for confounders.

Statistical analysis was performed with statistical software packages (SPSS, version 16.0; Chicago, IL) and Statview 5.0 (SAS Institute, Cary, NC). All P values were two-sided, and a P value <0.05 was considered statistically significant.

Results

The clinical characteristics of the study participants according to BMI categories are presented in Tables 1 and 2. Overall, the proportion of subjects with low BMI was relatively high (n = 585, 21.9%) compared with Western populations (1,2,3,4,6,8,9), with good cardiovascular risk profiles and without apparent undernourishment, even among the elderly subjects. While all variables, except high-density lipoprotein-cholesterol and age in the elderly subjects, were elevated across the increased BMI categories in both age groups, high-density lipoprotein-cholesterol in both groups and age in the elderly were reduced across the BMI categories (P for all trends <0.0001). Prevalence of current smokers was positively associated with increasing BMI categories only in the middle-aged (P < 0.01, χ2-test). Although not statistically significant, the proportion of current smokers was lowest in the category of elderly subjects with BMI 21.0–22.9 kg/m2.

Table 1.  Characteristics of the middle-aged participants
inline image
Table 2.  Characteristics of the elderly participants
inline image

Of note, subjects with low BMI, especially the elderly subjects, lost weight in the past 1–2 years, while normal weight, overweight, and obese subjects gained weight. The proportion of subjects with clinically important annual weight loss was inversely associated with increasing BMI categories in both middle-aged and elderly subjects (P < 0.01 and P < 0.001, respectively; χ2-test). The number of subjects with previously diagnosed coexisting diseases, except cardiovascular risks, was very low in both the middle-aged and in the elderly groups (n = 37 and 43, respectively), without a significant difference in prevalence between BMI categories in both age groups (data not shown).

Figure 1 shows LogCRP levels in nonsmoking subjects according to BMI or WC categories. In the elderly subjects (n = 948), LogCRP showed a sigmoid curve, rather than a J-shaped curve, against the six BMI categories with a base at 21.0–22.9 kg/m2, a plausible reference category of BMI (Figure 1a). In the low BMI categories, LogCRP levels in the elderly subjects were significantly higher than for the corresponding middle-aged subjects (both P < 0.05). However, there was no statistical difference in LogCRP between the low BMI categories and the reference BMI category. When subjects were divided into six categories according to WC, corresponding to six BMI categories, LogCRP showed a slight tendency toward a J-shaped curve against WC in the elderly subjects (Figure 1b). In contrast, in the middle-aged subjects, LogCRP showed an almost linear relationship with the BMI and WC categories.

Figure 1.

The relationship between LogCRP (CRP, mg/l), BMI categories, and waist circumference (WC) categories in nonsmokers. Open squares and closed circles represent data for the middle-aged and elderly subjects, respectively. (a) Subjects are stratified by BMI. Sample numbers in each BMI category are 76, 176, 259, 232, 152, and 145 for the middle-aged group and 55, 140, 248, 255, 138, and 79 for the elderly group according to increasing BMI categories, respectively. (b) Subjects are stratified by WC. Sample numbers in each BMI category are 167, 164, 207, 215, 143, and 144 for the middle-aged group and 101, 98, 184, 232, 167, and 133 for the elderly group according to increasing WC categories, respectively. Bars represent standard error. *P < 0.05, elderly vs. middle-aged subjects in the corresponding BMI group.

Figure 2 shows the estimated marginal means of LogCRP after adjustment for age, sex, smoking status, and weight changes over the past 2 years, because longer-term weight change influences the relationship between BMI and CRP (19). The adjusted means of LogCRP also showed a sigmoid curve, rather than a J-shaped curve, against BMI categories in the elderly subjects, and a small J-shaped curve with a different base in the middle-aged subjects.

Figure 2.

The relationship between adjusted means of LogCRP (CRP, mg/l) and BMI categories. Estimated marginal means of LogCRP were adjusted for age, sex, smoking status, and weight change over the past 2 years. Open squares and closed circles represent data for the middle-aged (n = 832) and elderly (n = 719) subjects, respectively. Sample numbers in BMI categories are presented in Tables 1 and 2. Bars represent standard error.

Discussion

There is mounting concern worldwide over the increasing prevalence of obesity and the resulting public health implications. Nevertheless, better understanding of the risk associated with low BMI is also of scientific interest. To date, the underlying mechanism for the relationship between low BMI and increased mortality risk has been poorly studied. In this study, of several clinical variables examined, the circulating CRP levels only showed a sigmoid curve association with BMI, particularly for the elderly subjects, suggesting the presence of subtle low-grade inflammation in the lean elderly subjects, irrespective of good cardiovascular profiles and the absence of obvious undernourishment. When only subjects in the low BMI categories were considered, LogCRP levels in the elderly were significantly higher than in the corresponding middle-aged subjects, implying an interference of aging.

The shape of the curve for the elderly subjects appeared to be sigmoid rather than J-shaped. Then, the association between BMI and circulating CRP might be attenuated in the highest BMI category of the elderly. However, such a curve, especially the part corresponding to high BMI, might be due to the relatively small number of subjects with severe obesity in this study. Further studies including a larger number of elderly subjects with high BMI will better reveal the shape of the curve against BMI.

Meanwhile, CRP is mainly synthesized by liver and secreted into the circulation, suggesting a considerable effect of liver function on the circulating CRP level. We also examined the relationship between circulating CRP levels and liver function, i.e., plasma hepatic enzyme (data not shown). Of note, the levels of aspartate transaminase, alanine transaminase, and γ-glutamyltranspeptidase appear to show small J-shaped curves against BMI categories in the elderly as well, but with different bottoms, i.e., BMI 23.0–24.9 kg/m2. Unfortunately, because of the limited data currently available (total n = 493; data collection ongoing), we could not clearly reveal the shape of the curve, which remains to be elucidated.

With regard to the clinical relevance of these findings, although high levels of circulating CRP per se are not necessarily directly involved in the pathogenesis of critical diseases such as CVD (22,23,24), persisting inflammatory status is commonly associated with fatal diseases (11,12,13,15,16). Recent studies have demonstrated that, in a general population of Japanese, even slightly elevated CRP, e.g., around 0.6 mg/l, may be more strongly associated with CVD events and coronary heart disease risk than lower CRP levels, e.g., 0.2–0.4 mg/l (25,26), which is substantially lower compared with Western populations (23,24). Therefore, the current results suggest that lean elderly subjects are at increased risk for CVD compared with normal weight subjects, but at lower risk compared with obese elderly people. Consequently, in the elderly, low body weight as well as high body weight may be associated with systemic inflammation via the different mechanisms presented in many clinical and basic studies. However, because this was a cross-sectional study, the cause-effect relationship as well as plausible mechanisms between mild CRP elevation and low body weight could not be determined and remains unknown. Furthermore, there may be racial differences in CRP cutoff levels for increased CVD risk as well as in the relationship with BMI.

Smoking is associated with a lower BMI (27,28). However, in the current study, the percentage of smokers was positively associated with increasing BMI categories in the middle-aged, whereas in the elderly this was lowest in the reference BMI category. In addition, when we considered nonsmokers only, LogCRP showed a sigmoid curve, suggesting that current smoking is unlikely to interfere with the sigmoid association of CRP with BMI.

Consistent with a previous report and review (3,19), subjects with low BMI tended to lose weight, particularly amongst elderly people, while overweight and obese persons tended to gain weight. The weight loss observed in the lean elderly people may be unintentional, but this was not confirmed by interviews or questionnaires. Unintentional weight loss appears to be associated with systemic inflammation and mortality (19,20,21), which mostly originates from pre-existing diseases, malnutrition, or advancing age (19). Thus, the change in body weight over recent years is a critical factor that distorts the association between BMI and CRP level. However, even after adjustment for confounders including weight change over the previous 2 years, the sigmoid CRP curve remained in the elderly subjects. This implies that weight loss itself cannot explain the shape of the curve and that low BMI as a result of long-term weight loss could affect systemic inflammatory status to a greater extent than weight loss.

With regard to nutrition, chronic loss of appetite is often observed in underweight persons and is accompanied not only with a lack of sufficient energy but also a minimum requirement of anti-inflammatory or antioxidative nutrition such as vitamin E, C, β-carotene, and trace elements (19), which may result in decreased tolerance to systemic inflammation, especially with advancing age. Unfortunately, we did not examine specific nutritional variables and food intake or use of dietary supplements; thus the impact of nutrition remains to be addressed.

In this study, we evaluated the effect of BMI on circulating CRP in comparison with WC. Indeed, BMI is moderately-to-strongly correlated with WC in a population study, but it is not always true on an individual basis. Considering that LogCRP levels stratified by WC did not clearly show a sigmoid or J-shaped curve, WC rather than BMI may be a more appropriate predictor for circulating CRP levels. Nevertheless, the important finding from the study is that low BMI, but not low WC, could better differentiate the elderly people with subtle low-grade inflammation from other healthy lean elderly people, probably because BMI is related not only to body fat mass but also to lean body mass including skeletal muscle, bone, and visceral organs. Iannuzzi-Sucich et al. (29) reported that BMI is a strong predictor of skeletal muscle mass in older women and men. The current findings might be related to the pathogenesis of frailty or Sarcopenia (29,30,31), which are characterized by unintentional weight loss, muscle loss, and systemic inflammation. Taken together, the assessment of physique, particularly for elderly people, with anthropometric measurements including BMI, is still needed.

Given that obesity as a risk for high mortality in the elderly is still debatable (1,3,5,6,7,8,9), elucidation of the mechanism(s) underlying the association between low BMI and fatal diseases is essential. This is particularly true in aging societies, such as developed countries, to gain new insights into the relationship between anthropometric measurements, critical diseases, and ultimately mortality. Large prospective studies to determine the circulating CRP level and other inflammatory markers as well as detailed body composition and nutrition in subjects with low or high BMI would address these issues.

Limitations

Lean elderly subjects may be more susceptible to infections of, e.g., genitourinary and respiratory tract, often showing latent and mild progression, resulting in negative test during detailed check-ups. Furthermore, other specific conditions associated with elevated CRP, such as cancer and collagen disease, as primary causes of mortality cannot be entirely ruled out. Therefore, careful interpretation should be placed on the current results. Finally, other confounders such as liver function, race, alcohol consumption, and physical activity, also interfere with the association. In this context, a larger study is needed, by stratifying subjects according to these confounders, because statistical adjustment for confounders does not fully address the issue.

Conclusion

Among several cardiovascular risk factors, the circulating CRP level showed a sigmoid curve against BMI in elderly subjects even after adjustment for confounders. These findings suggest the presence of subtle low-grade inflammation in lean elderly persons irrespective of good cardiovascular profiles and the absence of obvious undernourishment. These findings need to be confirmed in further studies.

Acknowledgment

We thank Dr Terao, Department of Biostatistics, Josai University, for helping with statistical analysis.

Disclosure

The authors declared no conflict of interest.

Ancillary