Higher number of steps and breaks during sedentary behaviour are associated with better lipid profiles

In this prospective cohort study, the participants were recruited from Tampere City Hospital Breast Clinic in Finland. Recruitment efforts were conducted one week per month for a total of 11 weeks. The target group for our study included women who were 50 to 60 years of age and participated in the breast cancer screening programme. Written information about this study was provided to all attendees. Those willing to participate returned the signed informed consent forms via mail.

A total of 3366 women attended the screening programme during the 11 weeks. Of these women, 880 (26.1%) were willing to participate in the study (Fig. 1). The participants (n = 650,73.9%) wore an accelerometer on their waist for a 14-day (two weeks) data collection period. From the original group of women, 667 (75.6%) underwent the laboratory test. Finally, this study comprised of 571 (64.9%) women who met the criteria for sufficient accelerometer data (at least four days per week with 10 or more hours per day based on guidelines used in previous reports [1, 21]) and underwent the laboratory tests.

Flow chart of the study population is presented in Fig. 1. An accelerometer, questionnaire (including items on comorbidities, medications, weight, height, smoking status and quality of life measured with the SF-36 and 15D, Appendix) and instructions for the laboratory tests were sent to the participants. Within two weeks the participants were instructed to visit laboratory for fasting venous blood samples. Plasma and serum was separated and stored at − 70° until analysed. The participants were randomized (1:1) to the intervention and control groups for long-term follow-ups. After the baseline measurements the participants received feedback of the analysis of the PA and SB collected by the accelerometer. In addition, the intervention group was given an exercise prescription. Further, the participants that were randomized at baseline, where invited to 1-year and 3-year follow-up measurements. Here, we report the baseline results before randomization.

PA and SB were measured at a 100 Hz sampling rate with tri-axial accelerometers (Hookie AM20 (Traxmeet Ltd., Espoo, Finland) and UKK AM30 (UKK Terveyspalvelut Oy, Tampere, Finland)) attached to a flexible belt that was worn on the right hip. The participants were given written instructions on how to use the accelerometer for 14 consecutive days during their waking hours. The accelerometer was not allowed to be worn during water activities, such as shower or swimming. The raw accelerometer data were stored on a mass storage device and sent back via mail to the research group for further analysis. All data were relocated to an analysing program written in Visual Basic for Applications to Microsoft Excel 2016 (Microsoft Corp. Redmond, WA, USA) and analysed according to the instructions described below.

We considered the mean amplitude deviation (MAD) values determined from the resultant acceleration of the three orthogonal acceleration components in six-second epochs as a valid indicator of incident oxygen consumption during movement [22, 23]. The MAD values were converted to the metabolic equivalent of task values (METs) (3.5 ml/kg/min of oxygen consumption) for each epoch. Intensity was calculated as the one-minute moving exponential average of the estimated MET values.

PA was classified into three intensity categories in terms of the METs: light PA (1.5–2.9 METs), moderate PA (3.0–5.9 METs and vigorous PA (more than 6.0 METs) [1]. Moderate and vigorous PA are combined with MVPA in many studies.

SB was defined as the time spent in a seated and/or resting position without movement (≤1.5 METs). Standing without movement was analysed separately. The classification of body posture was determined based on two facts: Earth’s gravity vector is constant, and the body’s posture while walking is upright. The accelerometer orientation with respect to the gravity vector was taken as the reference orientation, and the posture was determined from the incident accelerometer orientation with respect to the reference vector [9]. A transition from sitting to standing was considered a break in SB. The daily number of sit-to-stand transitions was based on the amount of lying/sitting periods during which the previous one minute estimated MET value indicated no movement and which ended up with a clear vertical acceleration followed by a standing position or movement [23]. The step detection algorithm split the measured acceleration data into vertical and horizontal components. The vertical component was bandpass filtered (1–4 Hz), and the positive values were integrated. When the integral value exceeded the specified limit, a step was detected [23].

Peripheral blood were collected at baseline for analysis of the serum TC (mmol/l), HDL-c (mmol/l), LDL-c (mmol/l), and triglyceride levels (mmol/l). The study participants were instructed to fast for 10–12 h before the samples were collected. The samples were taken in the laboratory at the Pirkanmaa Cancer Society in Tampere, Finland [24] and transported to be analysed at the United Medix Laboratories Ltd. in Helsinki, Finland [25]. In this report, the analysis of the TC, HDL-c, LDL-c and triglyceride levels was performed by the enzymatic method. The normal ranges were as follows: < 5.0 mmol/l for TC, > 1.2 mmol/l for HDL-c, < 3.0 mmol/l for LDL-c and < 1.7 mmol/l for triglycerides.

The study population was divided into quartiles based on the following:

The two middle subgroups (Q2 and Q3) were combined for the statistical analysis so that it was possible to compare the highest quartile (Q4) and the middle group (Q2–3) to the lowest quartile (Q1). The cut-off points for the lowest and highest quartiles were 54 and 66%, respectively, which corresponded to 6.8 and 10.8% of the total time accumulated for SB and MVPA, respectively. The corresponding cut-off points were 28 and 41 for the daily breaks and 5600 and 9100 for the daily steps.

Descriptive statistics (means and standard deviations (SDs)) were reported to summarize the participants’ characteristics. Linear regression analysis was used to examine the relationship between SB, MVPA, number of daily breaks, and number of steps per day and the TC, HDL-c, LDL-c, and triglyceride levels, as well as the difference between the groups (Q1, Q2–3, and Q4) in the TC, HDL-c, LDL-c, and triglyceride levels. Log-transformation for triglycerides was used in the linear regression models to obtain normally distributed residuals. Both unadjusted models and models adjusted for the BMI, hypothyroidism medication use, and hormone replacement therapy were generated. For analysis of the serum TC and LDL-c, the participants using statin medication as a lipid lowering therapy (n = 29) and the participants who did not respond questions concerning lipid lowering medication (n = 33) were excluded (total n = 62).

The baseline characteristics of 571 participants are summarized in Table 1. Out of the 15 diabetic patients, 14 (2.5%) reported taking medications for diabetes. Only 86 (16%) of the participants reported using hormone replacement therapy. In addition, 66 (12%) used over-the-counter products for menopausal symptoms. Smoking was uncommon; 58 (11%) reported being occasional or daily smokers, and 477 (89%) were never smokers or had stopped smoking.

In the baseline physical activity measurements, the mean amount of SB was 60% (SD 8.9%) of the total measurement time. From the total measured time, the participants spent 18% (SD 5.9%) standing, 14% (SD 3.9%) performing light PA, and 8.9% (SD 2.9%) performing MVPA in addition to the 60% of SB. On average, the participants took 7483 steps daily (SD 2682), and they had 35 (SD 9.6) breaks during sedentary behaviour (Table 1).

According to the linear regression analysis, light physical activity was associated with the concentration of HDL-c (coefficient 0.019, p < 0.001) (Table 2). After the models were adjusted for BMI, hypothyroidism medication use and hormone replacement therapy (HRT), this association remained still significant (coefficient 0.011, p = 0.005). The light PA was inversely related to the concentration of triglycerides in the unadjusted model (coefficient − 0.011, p = 0.012), but this association disappeared after the adjustments were made (coefficient − 0.004, p = 0.40) (Table 2). MVPA was associated with HDL-c both in the unadjusted model (coefficient 0.033, p < 0.001) and adjusted models (coefficient 0.018, p = 0.001) (Table 2). In addition, MVPA was inversely associated with the triglyceride concentration (coefficient − 0.017, p = 0.004), but this association disappeared after the model was adjusted for BMI, hypothyroidism medication use and HRT (coefficient − 0.004, p = 0.49) (Table 2). The number of breaks in SB was also associated with the concentration of HDL-c both in the unadjusted (coefficient 0.010, p < 0.001) and adjusted models (coefficient 0.005, p = 0.007). Breaks were associated with the triglyceride concentration in the unadjusted model (coefficient − 0.009, p < 0.001) and in the adjusted model (coefficient − 0.004, p = 0.028) (Table 2). The number of daily steps was associated with the concentration of HDL-c both in the unadjusted (coefficient 0.037, p < 0.001) and in the adjusted models (coefficient 0.019, p = 0.002). Additionally, the number of daily steps was inversely associated with the triglyceride concentration in the unadjusted model (coefficient − 0.024, p < 0.001), but this association disappeared after the model was adjusted for the BMI, hypothyroidism medication use and HRT (coefficient − 0.008, p = 0.23) (Table 2).

The relationship between standing and LDL-c was significant in the unadjusted model (coefficient − 0.014, p = 0.016) but not in the adjusted model (− 0.011, p = 0.072) (Table 2). MVPA and light PA were associated with TC both in the unadjusted model (coefficient 0.033, p = 0.009 and coefficient 0.019, p = 0.045, respectively) and adjusted models (coefficient 0.035, p = 0.008 and coefficient 0.019, p = 0.050, respectively).

The mean HDL-c level was the highest (1.92 (SD 0.43) mmol/l) in the participants who had the least amount of SB (SB less than 54% of waking hours) (Table 3).

The triglyceride concentration was on average 0.25 mmol/l (p < 0.001) lower in the group with SB ≥ 66% than in the group of women with SB less than 54% (Fig. 2). The differences between the groups disappeared after the adjustments were made to the model (Fig. 2). Compared to the SB < 54% group, the SB 54–65.9% group and SB ≥ 66% group had lower HDL-c concentrations by 0.12 mmol/l (p = 0.001) and 0.32 mmol/l (p < 0.001), respectively (Fig. 2). The differences between the groups remained significant after the model was adjusted: the concentration of HDL-c was 0.08 mmol/l lower in the SB 54–65.9% group (p = 0.035) and 0.14 mmol/l lower in the SB ≥ 66% group (p = 0.003) than in the SB <  54% group (Fig. 2).

When the number of daily breaks during SB were distributed to quartiles the mean concentration of HDL-c was the highest (1.89 mmol/l, SD 0.38), and the mean concentration of triglycerides was lowest (1.04 mmol/, SD 0.48) in the quartile with the most daily breaks (≥ 41) compared to the quartile with the lowest number of breaks (< 28).

In the unadjusted models, the HDL-c level was 0.16 mmol/l higher (p < 0.001) in the intermediate group (28–40.9 breaks) and 0.25 mmol/l higher (p < 0.001) in the group with the most breaks (≥ 41) than in the group with the fewest breaks during SB (< 28). After the model was adjusted for BMI, hypothyroidism medication use and HRT, the HDL-c level was 0.11 mmol/l higher (p = 0.017) in the group with ≥41 daily breaks during SB than in the group with < 28 breaks (Table 4).

The concentration of triglycerides was significantly lower both in the group with 28–40.9 breaks (− 0.15 mmol/l, p = 0.001) and in the group with ≥41 breaks (− 0.22 mmol/l, p < 0.001) than in the group with the fewest breaks (< 28). The differences remained significant even after the adjustments were made; the concentration of triglycerides was − 0.09 mmol/l lower (p = 0.031) in the group with 28–40.9 breaks and − 0.10 mmol/l lower (p = 0.032) in the group with ≥41 breaks when compared to the group with the fewest breaks (< 28) (Table 4).

The mean number of steps per day for the whole study population was 7483 (SD 2682). In the group with the fewest steps (< 5600 steps/day), the mean concentration of HDL-c was 1.63 (SD 0.36) mmol/l. In the intermediate group (5600–9099 steps/day), the mean HDL-c was 1.80 (SD 0.41) mmol/l. The group with the most daily steps (≥ 9100) had the highest mean HDL-c level, which was 1.90 (SD 0.41) mmol/l.

In the unadjusted models, the HDL-c level was 0.16 mmol/l higher (p < 0.001) in the 5600–9099 steps/day group and 0.26 mmol/l higher (p < 0.001) in the ≥9100 steps/day group than in the group with the fewest steps (< 5600). After the model was adjusted for BMI, hypothyroidism medication use, and HRT, the HDL-c level was 0.13 mmol/l higher (p = 0.006) in the ≥9100 steps/day group than in the < 5600 steps/day group (Table 5).

The concentration of triglycerides was significantly lower in the groups with 5600–9099 (− 0.16 mmol/l, p < 0.001) and at least 9100 (− 0.19 mmol/l, p < 0.001) daily steps than in the group with fewer than 5600 daily steps. These significant findings disappeared after the adjustments were made (Table 5).

As expected, the mean HDL-c level was the highest in the quartile with the highest amount of MVPA (at least 10.8% of waking hours). In this group, the mean HDL-c level was 1.90 (SD 0.39) mmol/l (Table 3). In the unadjusted linear regression models, the HDL-c level was 0.14 mmol/l (p = 0.001) higher in the MVPA 6.8–10.79% group and 0.25 mmol/l (p < 0.001) higher in the MVPA ≥10.8% group than in the MVPA less than 6.8% group. After the adjustments were made, the concentration of HDL-c was 0.14 mmol/l (p = 0.002) higher in the MVPA ≥10.8% group than in the MVPA < 6.8% group. In the unadjusted models, the concentration of triglycerides was 0.12 mmol/l (p = 0.005) lower in the MVPA 6.8–10.79% group and 0.15 mmol/l (p = 0.002) lower in the MVPA ≥10.8% group than in the MVPA < 6.8% group (Table 6). All these significant differences disappeared after the adjustments were made (Table 6).

The limitation of this type of study is that the participants are likely to comprise healthy volunteers who are already interested in their health. The volunteers are also likely to be more physically active. Altogether, 4161 women were invited for screenings with mammograms during the weeks we recruited study participants. Of these women, 3366 (81%) attended a screening, and 880 (26%) were willing to participate in this study. The mean BMI of the study population was slightly lower (26 kg/m² (SD 4.6)) than in the FinHealth2017 Survey population study with the mean BMI of 28 kg/m² among Finnish women in 2017 [41].

The recruitment method for the study was not very intensive. There were posters and flyers about the study in the screening mammography waiting room. Nurses at the screening facility mentioned about the recruiting process if they had the time to do so. This explains the loss of many participants and why the ones who wanted to participate are already quite physically active. However, there were 880 participants in this study, which is a reasonable amount.

The accelerometers were instructed to be used during waking hours and not during water-based activities. Therefore, some of the PA data were lost from the measurement period. However, the data are more objective and reliable than data in previous studies because these data are based on objective measurements and not on questionnaires only [42].

We lost a few participants and their data because a mail workers’ strike occurred during our study period. Because of the postal delay in the delivery, the battery for the accelerometers did not last for the whole measurement period.

Dietary behaviour could be a confounding factor in studies measuring serum lipids. In the present study we tried to control this important issue by collecting the participants from the breast cancer screening, which have a high participation rate in order to get normal population using different type of diet involved. Acute dietary effects were controlled by using overnight fasting before blood samples were taken.

The study data from the questionnaires are not objective. It might be tempting for the participants to embellish their responses, e.g., their weight, height or smoking status. Menopausal status was not addressed in our questionnaire, but due to the study’s target age group (50–60 years), it is likely that most of the participants were postmenopausal.