Diabetes-Related Distress, Insulin Dose, and Age Contribute to Insulin-Associated Weight Gain in Patients With Type 2 Diabetes: Results of a Prospective Study

Insulin therapy is frequently needed to achieve adequate glycemic control in patients with type 2 diabetes mellitus (T2DM) but often at the expense of weight gain. This is obviously a disadvantage of insulin therapy in an already obese population. The reported weight gain during the first year of insulin therapy ranges from ∼2 to 6 kg (1). This weight gain shows large interindividual differences, with some patients experiencing substantial insulin-associated weight gain, while others do not show any weight gain at all or even lose weight.

Identification of clinical variables related to insulin-associated weight gain in T2DM would be valuable to predict or potentially prevent this side effect. While a number of putative mechanisms are postulated to explain this weight gain (2,3), formal prospective studies including these different clinical variables are lacking.

Improvement in glycemic control (change in HbA_1c) by itself seems to be related to insulin-induced weight gain, but the contribution appears to be rather small and even nonsignificant in the long term (4). Even so, a decrease in HbA_1c is inevitably related to insulin therapy. Insulin dose should therefore be regarded as a potential contributor to weight gain rather than changes in HbA_1c.

Insulin-associated weight gain may be related to a decrease in physical activity, but this possibility has received limited attention. In a cross-sectional study, we showed that patients with pronounced weight gain (“gainers”) during insulin therapy performed less daily physical activity overall compared with those who did not gain weight (“nongainers”) (5).

Diabetes mellitus is associated with a significantly increased risk for depressive symptoms (6). Depressive complaints and anxiety are related with larger weight gain and an increase in cumulative incidence of obesity in men and women (7). As a result, it can be reasoned that depressive complaints and anxiety in patients with T2DM incur weight gain.

In the current study, we first prospectively investigated whether baseline (i.e., at the start of insulin therapy) glycemic control, physical activity, insulin dose, and diabetes-related distress and also changes of these variables during the course of the study (changes over time) correlated with insulin-associated weight gain in patients with T2DM. Second, we modeled the relationship between baseline variables and changes over time in order to predict the amount of weight gain 1 year after the start of insulin treatment.

In this prospective, multicenter, observational study, patients with T2DM who were starting insulin therapy were included. Patients were recruited from one university hospital (Radboud University Nijmegen Medical Center, Nijmegen, the Netherlands), three nonuniversity teaching hospitals (Jeroen Bosch Hospital, ‘s-Hertogenbosch; Slingeland Hospital, Doetinchem; and Bernhoven Hospital, Oss, the Netherlands), and four large primary care practices (de Teselaar, Bemmel; de Vier Kwartieren, Boxtel; Hoevense Veld, Uden; and Berghem, Berghem, the Netherlands). The diagnosis of T2DM was made according to the diagnostic criteria of the World Health Organization. The decision to start insulin treatment was at the discretion of the responsible physician and was always based on a failure of glycemic control while on oral glucose-lowering agents and diet. All patients were treated according to national guidelines (8). The choice of insulin preparation, insulin regimen, dose, and titration was left to the responsibility of the treating physician. Patients were excluded if they already used insulin or had steroid-induced diabetes, latent autoimmune diabetes in adults (which was excluded by testing anti-GAD antibodies), or maturity-onset diabetes of the young. Furthermore, clinical evidence of psychiatric, renal, cardiovascular, liver, or other diseases and medication (prednisone) that may influence study results regarding glucose and weight, hormonal disorders that may influence weight (i.e., thyroid diseases, unless properly treated with stable hormonal levels), bariatric treatment, excessive alcohol consumption (>20 g/day), drug abuse, use of thiazolidine derivatives, and pregnancy or intention to become pregnant during the study were all considered exclusion criteria.

Patients were followed for 12 months after the initiation of insulin therapy. Patient characteristics (age, sex, race, smoking habits) were noted, and clinical data were retrieved at baseline (i.e., prior to start insulin) and at 3-month intervals, which included body weight, blood pressure, HbA_1c, diabetes duration, comorbidities, number of hypoglycemic events, lipids, type of oral glucose-lowering medication, and insulin type and dose.

Before and at 6 and 12 months after the initiation of insulin therapy, physical activity levels, diabetes-related distress and depressive complaints both measured by questionnaires, and caloric intake were assessed.

The inclusion and exclusion criteria were reviewed at a screening visit, where patients underwent a history taking and a complete physical examination. The study protocol was approved by the local ethics committee (Commissie Mensgebonden Onderzoek region Arnhem-Nijmegen, the Netherlands). All patients provided written informed consent.

Patients were studied in fasted-overnight conditions. Weight was measured with subjects wearing light underwear only. Scales were calibrated annually. Waist circumference was measured midway between the lower rib margin and the iliac crest at expiration and hip circumference over the greater trochanter; waist-to-hip ratio was calculated. Skin fold measurements at five specific sites (abdomen, next to the belly site; suprailiac, just above the iliac crest of the hip bone; quadriceps, middle of the upper thigh; triceps, the back of the upper arm; and subscapula, beneath the edge of the shoulder blade) on the right site of the body were taken by using the Slim Guide Skinfold Caliper (Creative Health Products, Ann Arbor, MI). The scale of 0 to 85 mm is especially useful in assessing skin folds at the upper ranges, which are not accommodated by most other calipers. Accuracy may be interpolated to 0.5 mm. Measurements were taken twice and averaged. Body composition was assessed by using formulas described by Jackson et al. (9,10).

Blood pressure was measured twice by a manual sphygmomanometer in supine position after a minimum of 5-min rest at the right arm. The average blood pressure (mean systolic and diastolic blood pressure) was calculated. Furthermore, fasting blood samples were drawn to assess glucose, HbA_1c, thyroid hormone, and lipids.

Physical activity was objectively measured using a SenseWear Pro Armband (Body Media, Pittsburgh, PA) (11). The device was placed on the right upper arm over the triceps muscle for 4 consecutive days at baseline and at 6- and 12-month follow-ups. Measurements between 7:00 a.m. and 11:00 p.m. were used for calculations if >90% of these data were available. Outcome variables from the activity monitor included average MET (1 MET = consuming 1 kcal/kg of body weight per hour), time (minutes/day) spent at different activity intensity categories averaged per day over the measurement period, and the number of steps per day. SenseWear InnerView Professional software 6.1 was used to analyze the data.

Dietary intake was measured by food records. Participants returned a handwritten food record (modified from Thompson and Subar [12]) in which self-reported detailed description of the types and amounts of food, beverage, and/or supplements was documented over a 3-day period. Some of the patients provided weighed diet records (i.e., participants weighed used food and beverages). Food records were obtained prior to the start of insulin therapy and at 6- and 12-month follow-ups. Food records were checked by a dietitian, and the intake of nutrients was calculated with a computer program (2008 Vodisys Medical Software b.v., the Netherlands). This computer program used a national database on the composition of foods (13).

Diabetes-related distress and depression was measured with the Problem Areas in Diabetes (PAID) scale questionnaire (14,15). PAID scores were calculated using a five-point Likert-scale with options ranging from 0, not a problem, to 4, serious problem. Summing all item scores and multiplying by 1.25 resulted in an overall PAID score. A minimum score of 0 indicated no diabetes-related distress. A maximum score of 100 indicated significant diabetes-related distress. The questionnaire was taken at baseline and at 6- and 12-month follow-ups.

The sample size was based on data from a pilot study where physical activity was measured in nine patients with T2DM. This group showed a physical activity level of 1.31 MET and an SD of 0.30 MET. This study was powered to detect a difference in physical activity of 15% after 12 months of insulin therapy, with a power of 90% at a significance level of 0.05%. This translated to the inclusion of 65 subjects. It was estimated that 10–15% of the included patients would drop out, resulting in the need for 72–75 patients to be included.

The results of the study are displayed as mean ± SD, unless otherwise indicated.

To study changes in clinical variables (insulin dose, body weight, HbA_1c, physical activity, caloric intake, and PAID score) during the course of the study, a general linear model for repeated measures was used.

To determine whether selected baseline variables were related to the change in body weight after the start of insulin treatment, a univariate approach by simple correlations was performed using the Pearson correlation coefficient for normally distributed variables and the Spearman correlation coefficient for non-Gaussian distributed data.

Relationships between changes in clinical variables during the course of the study (changes over time) were investigated by a linear mixed model for repeated measures (16).

Finally, a multiple linear regression model was designed in order to estimate the relationship between the dependent variable (insulin-associated weight gain) and different independent variables or predictors. Both baseline parameters (from the correlation analyzes) and the change in parameters over time (from the linear mixed model for repeated measures) were included in the model.

For both the multiple linear regression model and the linear mixed model, estimated regression parameters are presented, with the appropriate SE and 95% CI. Details of both models are presented in the Supplementary Data.

All calculations were made using SPSS 20.0 for Windows (SPSS Inc., Chicago, IL). A P value <0.05 was considered significant in all statistical comparisons.

A total of 79 patients were eligible to enter the study. Seventy-seven patients consented and underwent screening. Of these, 74 patients with T2DM enrolled in the study. We excluded nine patients (eight patients with missing follow-up data or withdrawal and one patient diagnosed with maturity-onset diabetes of the young). Finally, 65 subjects were included in the analysis. Table 1 displays the descriptive characteristics of the patients at baseline.

Patients started with a mean of 18 ± 2 units (0.21 ± 0.02 units/kg) of insulin per day. At 6 and 12 months, patients used a mean of 39 ± 3 units (0.42 ± 0.03 units/kg) and 46 ± 4 units (0.49 ± 0.04 units/kg) of insulin per day, respectively (P < 0.05 compared with baseline). In addition, at 6 months, a shift in the use of different types of insulin was observed: 9% of the patients who used long-acting insulin were switched to biphasic or prandial insulin. At 12 months, 46% of the patients used long-acting insulin, 14% used biphasic insulin, and 40% used basal/bolus insulin. Within the study period, no severe hypoglycemic episodes requiring a third party occurred.

Mean body weight increased from 89.4 ± 2.4 to 92.4 ± 2.5 kg at 12 months of insulin therapy (P < 0.001) (see Fig. 1A), and waist circumference increased from 106 ± 2 to 109 ± 2 cm at 12 months (P < 0.01). Seventy-one percent of the patients showed body weight gain, whereas 29% had stable body weight or even lost weight after 12 months of insulin treatment (Fig. 1B). Calculated lean body mass decreased from 70 ± 2 to 69 ± 2 kg at 12 months (P = NS), while calculated body fat increased from 21 ± 1 to 24 ± 1% at 12 months (P < 0.001).

Mean HbA_1c decreased from 8.8 ± 0.2 (73) to 7.3 ± 0.1% (56 mmol/mol) after 12 months (P < 0.001).

After 12 months of insulin therapy, quantitatively measured physical activity was significantly reduced (Fig. 2A–E). Average MET decreased from 1.39 ± 0.04 to 1.32 ± 0.04 at the 12-month follow-up (P < 0.05), and time spent in sedentary physical activity level numerically increased, but this difference was not statistically significant. The amount of moderate physical activity (MET 3–6) at 12 months was significantly lower compared with 6 months of insulin therapy (P < 0.05). The average number of steps per day diminished significantly after the start of insulin therapy (from 6,441 ± 426 steps/day at baseline to 5,746 ± 343 steps/day after 12 months of treatment; P < 0.05).

Patients suffered from minor depressive complaints at baseline, and this did not change significantly over time (PAID scores: 10.0 ± 1.4 vs. 9.7 ± 1.3 vs. 9.4 ± 1.4 at baseline and after 6 months and 12 months, respectively; P = NS).

After insulin initiation, reported caloric intake decreased (from baseline 1,599 ± 72 to 1,392 ± 41 kcal/day at 12 months; P < 0.05), which cannot explain the insulin-induced weight gain.

Simple correlation coefficients were calculated between baseline variables and insulin-associated weight gain after 12 months of insulin therapy. These results are shown in Table 2. The baseline variables HbA_1c, insulin dose, and PAID score were positively correlated with weight gain. The baseline variables age and body fat were negatively correlated with weight gain.

Using a linear mixed model for repeated measures, we assessed the relationship between changes in clinical variables over a 12-month period of insulin treatment and weight gain. Details from the outcome of the linear mixed model are described in Supplementary Tables 1–3. Both the change in HbA_1c and insulin dose over time were significantly related with insulin-associated weight gain. A decrease in HbA_1c and an increase in insulin dose contributed to weight gain. Changes in physical activity and diabetes-related distress over time were not related to weight gain.

With a multiple linear regression model, we retrieved a cluster of clinical parameters consisting of baseline variables as well as variables that change over time that are related to weight gain. The change in HbA_1c was omitted from the prediction model (see CONCLUSIONS). Table 3 shows that baseline diabetes-related distress as measured by PAID questionnaire/score, initial insulin dose, age, and change in insulin dose over time were the main contributors to the prediction model with respect to weight gain after the initiation of insulin therapy. In this model, a significant interaction between age and PAID score was observed. Hence this indicates that the aforementioned variables are significantly related to insulin-associated weight gain but should be corrected for the interaction between age and diabetes-related distress.

From our prospective study that was performed in a fairly large number of patients, we were able to predict insulin-associated body weight gain with the use of a cluster of clinical variables. Diabetes-related distress (PAID score), insulin dose at the beginning of insulin therapy, and age, together with the change in insulin dose during the study period (i.e., 12 months), are the main independent variables contributing to this prediction model. Physical exercise decreased after the initiation of insulin treatment but was not related to insulin-induced weight gain.

In this real practice-based study, most patients (58%) started with a basal insulin regimen. The uptitration of insulin and the corresponding decrease in HbA_1c was well in line with the LANMET trial and other previous reports (1,17). The mean insulin-associated weight gain in our study patients was ∼3 kg after 12 months of insulin therapy and is in line with previous reports (4,18).

Our study is the first to explore the relationship between the change in insulin-associated body weight and diabetes-related distress (PAID score). We showed that diabetes-related distress significantly contributed to the model that predicts the amount of weight gain after the initiation of insulin therapy. An increase in diabetes-related distress reflected by a higher PAID score before the start of insulin therapy is related to a more pronounced weight gain. The PAID scale is a reliable and validated questionnaire, which is able to assess emotional distress and significant motivational problems in patients related to their diabetes management (14,19). Although a relationship between emotional factors and BMI has been described (20), no link has been identified between emotional distress per se and subsequent changes in weight after the initiation of insulin therapy. This study was not powered to investigate the influence of specific components of the PAID survey having an impact on body weight gain. It should be pointed out that patients with evident depressive disorders or on psychiatric medication, both known to have a significant impact on body weight, were excluded from the study. However, it can be hypothesized that diabetes-related distress contributes to an unhealthy diet or to physical inactivity (21). From the multiple linear regression analyzes, it was shown that diabetes-related distress (PAID scale) and age interact with each other. This means that age negatively affected the strength of the relationship between PAID score and weight gain; in other words, the predictive value of diabetes distress with respect to insulin-associated weight gain will be less with increasing age.

Intriguingly, we found that a higher initial insulin dose at the start of the study significantly contributed to a more pronounced weight gain after 12 months of insulin therapy. Also, an increase in the insulin dose during the course of the study significantly contributed to the increase of body weight. Both an increase in total body lipid stores due to an increase in lipid synthesis and an inhibition of fatty acid release (lipolysis) (22) as well as fluid retention (23) may play a role in insulin-associated weight gain. In a recent article, Balkau et al. (24) found in a large cohort of patients with T2DM starting on any type of insulin that starting insulin dose was associated with a greater weight gain, which is in agreement with our data. Their study did not investigate whether the dose escalation was also associated with weight gain. The reason why the initial insulin dose independently contributes to body weight gain after 12 months is unclear. The starting amount of insulin was established by the treating physician and may be partly based on the level of hyperglycemia and/or on the estimated insulin resistance. A separate analysis showed that HbA_1c and body weight at baseline were similar between patients starting with a low insulin dose compared with a high insulin dose, indicating that initial insulin dose was not determined by initial body weight or HbA_1c. Furthermore, it was shown that HbA_1c at the end of the study in patients who started with a relatively low insulin dose or high dose was well matched. This may suggest that a gradual decrease of HbA_1c compared with a steeper decrease has a favorable effect on ultimate weight gain. A possible explanation for this observation may be that patients suffering from a steep decrease of HbA_1c will experience more hypoglycemic episodes (which indeed was observed), leading to an increase in caloric intake. The change in reported caloric intake between these two groups, however, did not differ.

Age was shown to have a significant effect on insulin-associated weight gain. Older patients starting insulin therapy were prone to gaining more body weight. This could be related to a decrease in daily physical activity and fitness (25,26). Indeed, a subanalysis of changes in physical activity revealed that reduction in physical activity (MET and number of steps) was more pronounced in older patients.

As stated, physical activity did not seem to be related to insulin-associated weight gain in our study, i.e., it could not be identified as an independent predictor of insulin-associated weight gain. Age and physical activity are covariates. From the analysis, it was shown that only age was an independent variable. Nevertheless, it can still be viewed as an unfavorable finding from the perspective of general health. The decrease in energy expenditure after insulin initiation may be explained by a number of mechanisms. Patients who start insulin may experience (or at least fear) hypoglycemic episodes and, as a consequence, perform less physical activity in order to prevent hypoglycemia. In our patient cohort, the number of hypoglycemic episodes was low, and no severe hypoglycemia occurred. However, fear of hypoglycemia remains a potential explanation. Alternatively, a low physical exercise level may be a consequence of rather than a cause of body weight gain.

An increase in caloric intake is frequently mentioned as a determinant of weight gain. Especially after initiating insulin therapy, it can be reasoned that caloric intake may increase in relation to (the avoidance of) hypoglycemic events resulting in weight gain (27). In this current study, however, reported food intake did not increase, and even diminished significantly over time, arguing against increased food consumption being an important determinant of body weight gain. It should be noted that reported calorie intake was lower than expected based on body weight and level of physical activity. Probably, substantial underreporting in food intake has occurred, as is often encountered with food records and is quite typical in obese people. Several factors such as the intervals between the time points of taking the questionnaire, recall bias, true changes, and intrinsic feelings could have had an impact on reported caloric intake outcomes.

We and several others reported that changes in HbA_1c are related to insulin-associated weight gain, although we found that this contribution was rather small (4). In the current study, we prospectively confirmed that an improvement in glycemic control is significantly related to insulin-associated weight gain. Nevertheless, the primary intention of insulin therapy is to improve glycemic control and decrease HbA_1c. As such, the change in HbA_1c is the consequence of insulin therapy. Therefore, although the change in HbA1_c was related to weight gain in the linear regression model, we did not include this variable in the multiple linear regression model because it does not represent a true independent variable.

Importantly, ∼30% of the patients did not gain or even lost weight after the initiation of insulin therapy. From the results of the multiple linear regression model, it can be concluded that a young patient with a low PAID score, starting on a low insulin dose, and in need of a more lenient uptitration of insulin to achieve adequate glycemic control will probably suffer from less body weight gain after 12 months of insulin treatment.

Our study has limitations. First, this study cannot assess whether the decrease in physical activity is a consequence or a cause of body weight gain. Second, between-center differences and treatment effects could have influenced outcomes but could not be tested because of the small numbers of participating centers and because the study involves a relatively small sample size. Third, there was a discrepancy between reported caloric intake and body weight, which is probably explained by underreporting of food intake, a limitation of the use of food records. When filling in food records, subjects become more aware of their consumption and tend to report more socially preferred information. Given the increase in body weight together with a decrease in physical activity observed in our study, the underreporting seems to be more pronounced with the second measurement. Hence, the results of caloric intake should be interpreted with caution and do not allow us to conclude that caloric intake truly decreased. Finally, while our study provides predictors of insulin-associated weight gain, it cannot determine whether intervening on these factors, for example, a lower starting dose and a more lenient insulin uptitration, or interventions aimed at mood improvement will affect weight gain.

Our study results may have implications for clinical practice. The decrease in physical activity, although not related to insulin-associated weight gain in this study, is an unfavorable consequence of insulin therapy. Health education should be aimed at improving physical fitness in patients starting with insulin therapy, as this may be beneficial to reduce cardiovascular risk. The clinical importance of identifying emotional factors that might impact body weight gain in patients with T2DM adds value to understanding this novel relationship. Furthermore, it can be suggested that low initial insulin dose and a more gradual titration of insulin over time may limit or even prevent body weight gain.

In conclusion, the initiation of insulin therapy in patients with T2DM is associated with weight gain and a decrease in physical activity. Although the reduction in physical activity was not associated with weight gain, it is an important issue for patients starting insulin therapy. A cluster of clinical variables predicts insulin-associated weight gain. This includes diabetes-related distress (PAID score), initial insulin dose, change in insulin dose, and age.

Further prospective studies are needed to investigate the exact role of emotional factors, types of insulin regimens, and strategies of insulin titration on weight gain.

Acknowledgments. The authors thank P. Bouter and L. Kemink (Jeroen Bosch Hospital, ‘s-Hertogenbosch), A. Mulder (Slingeland Hospital, Doetinchem), M. Leclercq (Bernhoven Hospital, Oss), J. Rossen (Primary Care Practice, de Teselaar, Bemmel), J. Janssen (Primary Care Practice, de Vier Kwartieren, Boxtel), and A. Hoevenaars (Primary Care Practice, Hoevense Veld, Uden) for their contribution to this article. The authors also thank A. Ek and R. Bekkers (Jeroen Bosch Hospital, ‘s-Hertogenbosch); S. Hendriks (Radboud University Nijmegen Medical Center, Nijmegen); C. Wielink (Primary Care Practice, de Teselaar, Bemmel); J. van Boxtel (Primary Care Practice, Berghem, Berghem); P. Meeuwissen (Bernhoven Hospital, Oss); M. Meekes, S. Kemperman, and E. Grandia (Slingeland Hospital, Doetinchem); and all diabetes specialist nurses for including the study patients and performing measurements. The authors thank Dr. J. de Vries (Department of Human Nutrition, Wageningen University) for valuable advice on the interpretation of food intake data.

Duality of Interest. The authors thank Novo Nordisk A/S, Alphen aan den Rijn, the Netherlands, for providing an unrestricted grant to perform this study. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. H.J.J. and G.M.M.V. researched and discussed data and wrote and reviewed the manuscript. A.F.J.d.H. contributed to the statistical analyzes and reviewed the manuscript. P.M.N. and W.J.d.G. reviewed the manuscript. C.J.T. discussed data and reviewed the manuscript. C.J.T. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.