Optimal Number of Steps per Day to Prevent All-Cause Mortality in People With Prediabetes and Diabetes

Step-based recommendations may be an ideal solution to increase the physical activity levels of people with prediabetes and diabetes (1), which is key to lowering the risks of mortality of this prevalent subgroup of the population (2). Previous evidence has shown an inverse association between steps count and risk of mortality (3). While existing studies may have included participants with prediabetes and diabetes, research restricted to this specific subpopulation is lacking. The current study aimed, for the first time, to examine the detailed dose-response associations between the number of steps per day and all-cause mortality in a sample of U.S. adults with prediabetes and diabetes.

This prospective analysis used data from the 2005–2006 wave of the National Health and Nutrition Examination Survey (NHANES) (4). We included adults (≥18 years old) with prediabetes and diabetes with at least 1 valid day (>10 h/day of wear time) of accelerometry data (3). Diabetes was defined as a self-reported physician diagnosis of diabetes or as elevated levels of fasting glucose (≥126 mg/dL) or HbA_1c (≥6.5%) measured in the same blood sample in a person without a previous diagnosis of diabetes (5,6). Prediabetes was ascertained through self-reported physician diagnosis or levels of fasting glucose from 100 to 125 mg/dL or HbA_1c between 5.7 and 6.4% (5). Data were linked to death records from the National Death Index through 25 February 2019. Participants wore accelerometers (AM-7164, ActiGraph, LLC, Fort Walton Beach, FL) on their hip for 7 consecutive days, from which the number of steps per day was extracted and aggregated for each participant by using previously validated algorithms (3).

We examined the dose-response association between the number of steps per day and all-cause mortality using restricted cubic spline Cox proportional hazards regression models and sampling weights to account for the complex survey design of NHANES. Knots were placed at the 10th (reference), 50th, and 90th percentiles of the exposure distribution (7). Nonlinearity was assessed by a Wald test (7). Additionally, we estimated the nadir of the curve (i.e., the exposure value at which the maximum significant risk reduction was observed) and associated E-values (8), which quantify the minimum strength of association on the relative risk of mortality that an unmeasured confounder must have with both the exposure and outcome to negate the observed exposure–outcome association. We adjusted our models for age, sex, ethnicity, education, smoking, alcohol, diet, diabetes medication, and valid daily wear time. To minimize the possibility of reverse causation, participants with <1 year of follow-up were removed from the analysis. We conducted two sensitivity analyses. First, we combined the samples of participants with prediabetes and diabetes. Second, another analysis including participants with at least 3 days of accelerometry data was conducted. We ran an additional exploratory analysis to examine the associations between steps per day and cardiovascular disease (CVD) mortality. We used R 4.2.1 software (R Core Team, 2017) in our computations.

The data used in this study are publicly available.

We included 1,194 participants with prediabetes (11,137 person-years; median follow-up 9.83 years; 200 deaths) and 493 participants with diabetes (4,296 person-years; median follow-up 8.75 years; 138 deaths). Participants with prediabetes were on average 55 years old, had a BMI of 30 kg/m², were mostly male (56%), had a median of 8,500 steps per day, had an average HbA_1c of 5.71%, and provided 5.5 days of valid days of accelerometry data (Supplementary Table 1). Participants with diabetes were older (61 years old), had a higher BMI (32 kg/m²), were mostly male (51%), had a lower median steps per day (6,300), had a mean HbA_1c of 7.36%, and provided on average 5.5 days of valid accelerometry data (Supplementary Table 1).

In participants with prediabetes, compared with taking 3,779 steps per day (i.e., reference, 10th percentile), taking 10,678 steps per day (i.e., nadir) was associated with significantly lower all-cause mortality (hazard ratio [HR] 0.25 [95% CI 0.16–0.36]; E-value, 7.58 [upper CI 4.85]). In participants with diabetes, compared with taking 2,532 steps per day (i.e., reference, 10th percentile), taking 10,177 steps per day (i.e., nadir) was associated with a lower risk of all-cause mortality (HR, 0.25 [95% CI, 0.14–0.47]; E-value, 7.32 [upper CI 3.61]) (Fig. 1).

Sensitivity analyses using the combined sample (Supplementary Fig. 1) and separately restricted to participants who provided at least 3 days of valid accelerometry data (Supplementary Fig. 2), mirrored the results from the main analysis. For CVD mortality, we found a nonlinear association for participants with prediabetes (49 CVD deaths) and a linear association for those with diabetes (36 CVD deaths) (Supplementary Fig. 3). It is important to note the low number of CVD events when interpreting these results.

Previous studies have examined the association between step counts and mortality among the general population (3). This study shows first-time evidence of nonlinear associations between the number of steps per day and all-cause mortality specifically among those with prediabetes and diabetes.

There are several key clinical implications of our results. First, even low levels of stepping (i.e., just above the 10th percentile of the exposure distribution ∼3,700 steps per day for participants with prediabetes and ∼2,500 steps per day for those with diabetes) may reduce the risk of all-cause mortality among the studied population. Second, our findings suggest that ∼10,000 steps per day may be the optimal dose to lower the risk of all-cause mortality among people with prediabetes and diabetes. These results contrast with recent findings suggesting that 6,000–8,000 steps per day may be the optimal dose to lower the risk of all-cause mortality among the general population (9,10). Ultimately, our results highlight the need for specific recommendations tailored to individuals with prediabetes or diabetes.

Limitations of this study include its observational design and the lack of repeated measures of steps, which precludes from making any causal claims. The low number of participants with diabetes may have resulted in uncertain estimates. Residual confounding may still be present, although the large E-values showed this possibility is minimal. Owing to the lack of data at higher levels of the exposure, the association between high numbers of steps per day and mortality should be interpreted with caution. ∼20% of eligible participants had invalid accelerometery data and were removed from the analyses, which may have introduced selection bias (i.e., healthier individuals may have provided valid accelerometry data). The device used in this study may underestimate step counts, particularly among older individuals with slower gait speeds (11). However, step counts from the ActiGraph 7164 have been shown to be accurate at the population level (12).

Notwithstanding these limitations, our study has generated novel insights that may help shape step-based physical activity recommendations for people with prediabetes and diabetes. Such recommendations may be particularly relevant for people who accumulate their physical activity in an unstructured manner and for whom it may be challenging to determine whether they are sufficiently active in relation to the current minute- and intensity-based physical activity guidelines (i.e., 150–300 min/week of moderate to vigorous physical activity) (13).

Accumulating more steps per day was associated with a lower risk of all-cause mortality in people with prediabetes and diabetes. Our findings suggest that ∼10,000 steps per day may be optimal to lower the risk of all-cause mortality of this population. If provided with effective behavioral strategies, our results could inform evidence-based practice to lower the increased mortality risks associated with prediabetes and diabetes.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. J.d.P.-C. and F.A.-B. contributed to data preparation. J.d.P.-C., F.A.-B., D.G.-G., and B.d.P.C. interpreted the data. J.d.P.-C., F.A.-B., D.G.-G., and B.d.P.C. critically revised the manuscript for important intellectual content. J.d.P.-C. and B.d.P.C. contributed to concept and design. J.d.P.-C. and B.d.P.C. drafted the manuscript. D.G.-G. and B.d.P.C. contributed to statistical analysis. J.d.P.-C. and B.d.P.C. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.