Introduction

Body weight is one of the markers of nutritional status.[1] Most of the current studies usually use body mass index (BMI), which is defined as the weight (kg) divided by the square of the height (m2), to assess obesity or low body weight. The association between body weight or BMI and the risk of common chronic diseases has been widely reported.[2–5] Multiple disease states, such as cancer and cardiovascular disease, are associated with weight extremes and unexpected weight changes.[6] Most of the current studies confirmed the J-shaped or U-shaped correlation between BMI and the risk of mortality.[7,8] For example, obesity was confirmed to be related to metabolic and cardiovascular risks,[9] while some studies indicated that mortality in the elderly and chronic disease populations was consistently associated with a lower BMI.[10,11] However, body weight changes continuously throughout a lifetime. Many prospective studies only conducted one-time measurements of weight or BMI, and whether the dynamic weight change is an independent risk factor for cancer, stroke, cardiovascular diseases, or other chronic diseases remains controversial.[3,12,13]

According to a meta-analysis of 26 prospective studies with participants aged 40–65 years at baseline, weight gain and weight loss were both associated with increased mortality risk.[14] A similar conclusion was reached by another meta-analysis of 17 prospective studies among adults over 60 years old.[15] However, this association seems to differ by weight status. Some studies showed that weight loss was associated with increased mortality, independent of weight status (underweight, normal weight, overweight, or obesity, classification by BMI),[16,17] while some studies found no association.[18,19] Furthermore, the association between weight or BMI change and disease risk may be influenced by lifestyle and genetic factors.[20] It is worth noting that most of the studies in these two meta-analyses were conducted in western countries or regions such as the US and Europe, and few studies were conducted in Asia, especially in China. Compared with white people of the same age, sex, and BMI, Asians typically have a higher percentage of body fat.[21] Asians with the same BMI as Americans and Europeans have a higher prevalence of metabolic diseases, particularly type 2 diabetes.[22] It means that there are genetic backgrounds, dietary habits, and lifestyle differences between Western countries and Asia, which have been proved before.[23] The above differences may affect the results of extrapolation from western population to Chinese population.

In the 1980s, a Nutritional Intervention Trial (NIT) was initiated in Linxian, Henan province of China. More than 20,000 participants were followed up to collect data on all-cause mortality. Herein, using the data from the NIT cohort 1986–2016, we aimed to examine the relationship between five-year weight change among healthy adults (aged 40–69 years) and long-term risk of all-cause and cause-specific mortality.

Methods Study population

This prospective observational study was based on the NIT cohort. Detailed design of the NIT cohort has been described elsewhere.[24] Briefly, the Linxian General Population Nutrition Intervention Trial enrolled healthy adults (aged 40–69 years) from four towns (Rencun, Hengshui, Donggang, and Yaocun) in northern Linxian in 1985. Subjects with chronic wasting diseases at baseline or who regularly took vitamin and mineral supplements were excluded from this study. Finally, a total of 29,584 subjects were included in the NIT study in 1986 to receive daily combined vitamin/mineral supplement or placebo. The four factors tested in the study were as follows: factor A (retinol/zinc), factor B (riboflavin/niacin), factor C (vitamin C/molybdenum), and factor D (selenium/vitamin E/beta-carotene). Doses for daily multivitamins and minerals supplementation are shown in Supplementary Table 1, https://links.lww.com/CM9/B848. Subjects were randomly assigned to one of eight intervention groups according to a 24 fractional factorial design, which received factors ABCD, AB, AC, AD, BC, BD, CD, or placebo, respectively. Before enrollment, each participant signed informed consent forms. The intervention lasted from March 1986 to May 1991. This study was approved by the Institutional Review Boards of the Cancer Hospital Chinese Academy of Medical Sciences (CHCAMS) and the US National Cancer Institute (NCI) (No. 22/445-367). The trial was registered with ClinicalTrials.gov (NCT00342654). All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Declaration of Helsinki 1975, as revised in 2000. All research was conducted in accordance with relevant guidelines and regulations. Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.

Baseline information

A self-designed questionnaire was used to collect demographic characteristics and lifestyles at baseline in 1985, including age, gender, smoking, alcohol drinking, education level, family history of cancer, and dietary habits. Smoking was defined as regular cigarette or pipe use for at least six months, and the use of alcohol was defined as any alcohol consumption in the past 12 months. Family history of cancer was considered positive if a subject had one or more first-degree relatives diagnosed with any cancer. History of major illness was defined as illness that prevents subjects from working. Physical examination was conducted to measure body height, weight, and blood pressure (BP). Body height and weight were measured while subjects were not wearing shoes according to a standard protocol. BMI was calculated as weight in kilograms divided by height in meters squared (kg/m2). All subjects received BP measurements using a mercury sphygmomanometer in a seated position after resting for 5 min. Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were measured as the first and fifth Korotkoff sounds, respectively. BP was repeatedly measured three times within 5 min, and the mean value of these three measures was used for the final analysis.

Weight change

Body weight was measured in 1986 and once again in 1991, and the same procedure for measurement was used for both time points. In our study, we used the weight trajectory patterns during 1986–1991 to evaluate the association between weight change and risk of mortality.

Mortality follow-up

The initial date for follow-up was defined from the most recent body weight examination (1991). The end date of participant follow-up was determined as earliest occurrence among the last known follow-up date, date of death, or the date of the study closure (March 2016). Primary causes of death were identified by the International Classification of Disease, Tenth Revision (ICD-10): cancer (C00–C97), stroke (I60–I69), and heart diseases (I00–I25, I28–I59). During the follow-up, local medical workers continued to visit participants every three months, and new cancer cases and all-cause deaths were verified by a panel of American and Chinese experts (1991–1996) or senior Chinese doctors (1996–2016) according to diagnostic materials such as case records, biochemical results, ultrasound, X rays, endoscopy, surgery reports, and pathology or cytology slides.

Statistical analysis

We used two alternative categorizations for weight change. First, absolute weight change was divided into four categories according to the difference between two weight measurements (weight1991 minus weight1986), including weight loss ≥2 kg, weight maintenance (weight change <2 kg), weight gain of ≥2 kg and <5 kg, and weight gain ≥5 kg. In addition, using weight status at two time points, we defined seven weight status change patterns for each interval [Supplementary Table 2, https://links.lww.com/CM9/B848]: stable normal weight, underweight to normal weight, overweight to normal weight, becoming underweight, becoming overweight, stable underweight, and stable overweight. Differences in baseline demographic characteristics across different weight change categories were compared using chi-squared test for categorical variables and nonparametric Kruskal–Wallis test for continuous variables. Univariable and multivariable Cox proportional hazards regression models were used to calculate hazard ratios (HRs) and 95% confidence intervals (95% CIs) for the association between weight change and risk of all-cause and cause-specific mortality, taking the weight maintenance group or stable normal weight group as the reference group. Potential covariates in models included age at baseline, gender, baseline BMI categories (underweight: BMI <18.5 kg/m2, normal weight: 18.5 ≤BMI ≤23.9 kg/m2, and overweight or obesity: BMI ≥24.0 kg/m2; referred to BMI categories in China), smoking, alcohol drinking, family history of cancer, history of major illness, education level, towns, nutrition intervention arms, pulse rate, SBP, and DBP. Cumulative mortality rates were estimated by the Kaplan–Meier method, and Log-rank tests were used to examine the difference between cumulative mortality curves. Restricted cubic splines of change in body weight were used to graphically assess the association between weight change and our predefined outcomes after adjusting for confounders mentioned above, while weight or BMI change of 0 was defined as the reference. In the subgroup analysis, we examined interactions with age at baseline, sex, and BMI at baseline. Sensitivity analysis was performed by excluding 10% of participants randomly, participants with extreme BMI, or participants with chronic wasting diseases at baseline. Statistical analyses were performed using IBM-SPSS Software (version 23.0, IBM Corp., Armonk, NY, USA) and R Statistical Software, (version 4.1.3, R Foundation for Statistical Computing, Vienna, Austria). All tests were two-sided, and P <0.05 was considered statistically significant.

Results

A total of 21,028 participants were included in the final analysis. The flow diagram of the current study is shown in Supplementary Figure 1, https://links.lww.com/CM9/B848. Weight in 7752 participants was well-maintained (weight change <2 kg) during the two measurements, while the number of individuals who had weight gain ≥2 kg and <5 kg, weight gain ≥5 kg, and weight loss ≥2 kg were 4978, 2875, and 5423, respectively. There are significant differences in age, BMI, pulse rate, SBP, DBP, gender, smoking status, towns, and education levels at baseline. Subjects who had weight loss ≥2 kg tended to be males, elderly, smokers, have higher BMI and SBP, live in Yaocun, and receive less education (all P-values <0.05) [Table 1]. The correlation between weight change and selected variables at baseline was also estimated [Supplementary Tables 3 and 4, https://links.lww.com/CM9/B848].

Table 1 - Baseline information by absolute weight change categories. Baseline characteristics Weight change <2 kg (n = 7752) Weight gain ≥2 kg and <5 kg (n = 4978) Weight gain ≥5 kg (n = 2875) Weight loss ≥2 kg (n = 5423) U/ χ 2 P values Age (years) 51 (45, 58) 50 (43, 57) 48 (41, 55) 53 (47, 60) 462.188† <0.001 BMI (kg/m2) 21.6 (20.3, 23.1) 21.5 (20.1, 23.1) 21.6 (20.3, 23.2) 22.3 (20.9, 23.9) 411.373† <0.001 Pulse rate (times/min) 72 (68, 76) 72 (68, 76) 72 (68, 76) 72 (68, 78) 31.016† <0.001 SBP (mmHg) 125 (110, 140) 120 (110, 140) 120 (110, 140) 130 (115, 140) 127.485† <0.001 DBP (mmHg) 80 (70, 90) 80 (70, 90) 80 (70, 90) 80 (75, 90) 13.522† 0.004 Gender 229.316‡ <0.001 Female 4485 (57.9) 3146 (63.2) 1946 (67.7) 2839 (52.4) Male 3267 (42.1) 1832 (36.8) 929 (32.3) 2584 (47.6) Smoking* 203.864‡ <0.001 No 5514 (71.1) 3772 (75.8) 2297 (79.9) 3615 (66.7) Yes 2234 (28.8) 1204 (24.2) 577 (20.1) 1805 (33.3) Unknown 4 (0.1) 2 (0.0) 1 (0.0) 3 (0.0) Alcohol drinking* 8.129‡ 0.043 No 6011 (77.5) 3889 (78.1) 2168 (75.4) 4186 (77.2) Yes 1737 (22.4) 1087 (21.8) 706 (24.6) 1234 (22.8) Unknown 4 (0.1) 2 (0.0) 1 (0.0) 3 (0.0) Family history of cancer 5.472‡ 0.140 No 5164 (66.6) 3237 (65.0) 1857 (64.6) 3558 (65.6) Yes 2588 (33.4) 1741 (35.0) 1018 (35.4) 1865 (34.4) History of major illness 3.426‡ 0.330 No 7123 (91.9) 4574 (91.9) 2628 (91.4) 4939 (91.1) Yes 629 (8.1) 404 (8.1) 247 (8.6) 484 (8.9) Towns* 1065.520‡ <0.001 Yaocun 2426 (31.3) 1088 (21.9) 552 (19.2) 2356 (43.4) Rencun 1748 (22.5) 1091 (21.9) 491 (17.1) 1129 (20.8) Donggang 1638 (21.1) 1107 (22.2) 820 (28.5) 1066 (19.7) Hengshui 1936 (25.0) 1690 (33.9) 1011 (35.2) 869 (16.0) Unknown 4 (0.1) 2 (0.0) 1 (0.0) 3 (0.0) Education levels 83.228‡ <0.001 Never 3192 (41.2) 2000 (40.2) 1087 (37.8) 2305 (42.5) <5 years education 2387 (30.8) 1542 (31.0) 885 (30.8) 1750 (32.3) Primary school or higher education 1446 (18.7) 988 (19.8) 685 (23.8) 872 (16.1) Not sure 727 (9.3) 448 (9.0) 218 (7.5) 496 (9.1) Nutrition intervention arms 29.390‡ 0.105 ABCD 1012 (13.1) 628 (12.6) 334 (11.6) 682 (12.6) AB 922 (11.9) 652 (13.1) 352 (12.2) 722 (13.3) AC 980 (12.6) 649 (13.0) 360 (12.5) 614 (11.3) AD 980 (12.6) 587 (11.8) 350 (12.2) 712 (13.1) BC 971 (12.5) 600 (12.1) 374 (13.0) 700 (12.9) BD 963 (12.4) 615 (12.4) 383 (13.3) 650 (12.0) CD 921 (11.9) 618 (12.4) 342 (11.9) 680 (12.5) Placebo 1003 (12.9) 629 (12.6) 380 (13.2) 663 (12.2)

Data are presented as median (Q1, Q3) or n (%). *Data of unknown classification were not included in the analysis; †U values; ‡χ2 values. BMI: Body mass index; DBP: Diastolic blood pressure; SBP: Systolic blood pressure.

Table 2 presents the associations between weight change and risk of all-cause and cause-specific mortality. The median survival time was 18.4 years. From 1991 to 2016, a total of 13,565 deaths occurred, including 3713 deaths from cancer, 4540 deaths from stroke, and 3504 deaths from heart disease. After adjusting for age and gender, risks of all-cause, cancer, and heart diseases mortality in participants who had weight loss ≥2 kg were all significantly increased by 14%, respectively (HRAll-cause = 1.14, 95% CI: 1.09–1.18, P <0.001; HRCancer = 1.14, 95% CI: 1.05–1.23, P = 0.001; HRHeart diseases = 1.14, 95% CI: 1.05–1.24, P = 0.001). The difference was still significant after controlling for potential confounders including age, gender, baseline BMI, smoking, alcohol drinking, family history of cancer, history of major illness, education level, towns, nutrition intervention arms, pulse rate, SBP, and DBP (HRAll-cause = 1.14, 95% CI: 1.09–1.19, P <0.001; HRCancer = 1.12, 95% CI: 1.03–1.21, P = 0.009; HRHeart diseases = 1.21, 95% CI: 1.11–1.31, P <0.001). Weight gain ≥5 kg was associated with an 11% decreased risk of cancer mortality (HRCancer = 0.89, 95% CI: 0.79–0.99, P = 0.033) and a 23% increased risk of stroke mortality (HRStroke = 1.23, 95% CI: 1.12–1.34, P <0.001). The association between weight status change and mortality was also evaluated after excluding 41 participants with missing height information in 1991. Weight change from overweight to normal weight (HRAll-cause = 1.18, 95% CI: 1.09–1.27, P <0.001), becoming underweight (HRAll-cause = 1.35, 95% CI: 1.25–1.46, P <0.001), stable underweight (HRAll-cause = 1.16, 95% CI: 1.05–1.28, P = 0.003), and stable overweight (HRAll-cause = 1.11, 95% CI: 1.05–1.17, P <0.001) could all increase the risk of all-cause mortality by 18%, 35%, 16%, and 11%, respectively [Supplementary Table 5, https://links.lww.com/CM9/B848]. There were significant differences in all-cause and cause-specific mortality risk among the groups (all P-values from log-rank test <0.001, Figure 1 and Supplementary Figure 2, https://links.lww.com/CM9/B848).

Table 2 - HRs and 95% CIs for the associations between weight change and risk of all-cause and cause-specific mortality. Cause of death Weight change <2 kg (n = 7752) Weight gain ≥2 kg and <5 kg (n = 4978) P values Weight gain ≥5 kg (n = 2875) P values Weight loss ≥2 kg (n = 5423) P values Total No. of deaths 5003 3060 1643 3859 Crude HR (95% CI) 1 0.91 (0.87–0.95) <0.001 0.81 (0.77–0.86) <0.001 1.24 (1.19–1.30) <0.001 Age and gender-adjusted HR (95% CI) 1 0.99 (0.95–1.04) 0.757 1.04 (0.98–1.10) 0.169 1.14 (1.09–1.18) <0.001 Multivariable adjusted HR (95% CI)* 1 0.99 (0.95–1.04) 0.711 1.03 (0.97–1.09) 0.319 1.14 (1.09–1.19) <0.001 Cancer No. of deaths 1399 826 411 1077 Crude HR (95% CI) 1 0.88 (0.81–0.96) 0.003 0.73 (0.65–0.81) <0.001 1.23 (1.13–1.33) <0.001 Age and gender-adjusted HR (95% CI) 1 0.94 (0.87–1.03) 0.192 0.87 (0.77–0.97) 0.010 1.14 (1.05–1.23) 0.001 Multivariable adjusted HR (95% CI)* 1 0.96 (0.88–1.05) 0.409 0.89 (0.79–0.99) 0.033 1.12 (1.03–1.21) 0.009 Stroke No. of deaths 1632 1101 639 1168 Crude HR (95% CI) 1 1.00 (0.93–1.08) 0.919 0.97 (0.89–1.07) 0.539 1.16 (1.08–1.25) <0.001 Age and gender-adjusted HR (95% CI) 1 1.09 (1.01–1.18) 0.024 1.25 (1.14–1.37) <0.001 1.07 (0.99–1.15) 0.092 Multivariable adjusted HR (95% CI)* 1 1.08 (1.00–1.16) 0.060 1.23 (1.12–1.34) <0.001 1.07 (0.99–1.16) 0.079 Heart disease No. of deaths 1329 744 387