Abstract
Glycemic outcomes of adults with type 1 diabetes may be affected by depression. Our aim was to compare outcomes of “depressed” (Patient Health Questionnaire-9 ⩾ 10, N = 83) to “not-depressed” matched control (Patient Health Questionnaire-2 < 3, N = 166) adults with type 1 diabetes with objective measures. The depressed group had poorer blood glucose control and, for those with glucose meter downloads, fewer glucose tests/day. The groups did not differ on glucose variability or episodes of hypoglycemia. Depression in adults with type 1 diabetes is associated with poorer glycemic control and less blood glucose monitoring. Future research should examine whether treatment of depression results in better self-care and glycemic outcomes.
Depression in adults with type 1 diabetes (T1D) is a significant concern because depression has been shown to relate to worse glycemic control, poorer quality of life, greater morbidity, and higher health care costs (Bachle et al., 2015; Barnard et al., 2006; Gilsanz et al., 2018; Sacco and Bykowski, 2010; Strandberg et al., 2014; Trief et al., 2014). Depression has also been linked to poorer self-care (e.g. poorer diet, less activity, and poorer medication adherence) in persons with type 2 diabetes (Gonzalez et al., 2007; Lin et al., 2004) and missing insulin doses and exercising less often in those with T1D (Trief et al., 2014).
Most studies that have examined the relationship between glycemia and depression in adults with T1D used glycated hemoglobin A1c (HbA1c) as the measure of glycemic control (Nathan et al., 1984). HbA1c is a widely used, reliable, and valid measure of blood glucose control over the preceding 2–3 months; a higher value indicates poorer blood glucose (glycemic) control. Prior reports found that, in a large sample of adults with T1D (N = 6172), depression (defined as a Patient Health Questionnaire (PHQ)-8 score ⩾10 and as a continuous variable) was associated with worse clinical outcomes (Trief et al., 2014). HbA1c was higher in participants who were depressed, and those who were depressed were more likely to report on surveys that they had missed insulin doses, engaged in less physical activity, and checked their blood glucose less frequently. This poorer regimen adherence leads to poorer glycemic control and may increase the risk of hypoglycemia, that is, seriously low blood glucose that can lead to transient symptoms like shakiness, sweating, and confusion, and serious consequences that include loss of consciousness, seizures or death. Therefore, the American Diabetes Association (ADA) Standards of Care recommends annual screening for depression in all persons with diabetes (ADA, 2018; Young-Hyman et al., 2016).
The aim of the current study was to compare adults with T1D screened for depression seen in a real-world setting and found to have probable depression to those without depression on glycemia-related outcomes. To our knowledge, this is the first study to examine the association of depression in adults with T1D with a number of glycemic measures; we include HbA1c and other objective data from blood glucose meter downloads. These outcomes include documented hypoglycemia, measures of glycemic variability, and frequency of self-monitoring of blood glucose (SMBG). SMBG is required for adults with T1D, as it provides real-time information about their blood glucose level. This helps them appropriately schedule food, activity, and correct insulin doses to maximize the likelihood of achieving good glucose control and avoiding hypoglycemia (Kirk and Stegner, 2010). We hypothesized that the depressed group would demonstrate poorer glycemic control and poorer self-care (defined as less frequent SMBG).
Methods
Participants
Adults with T1D at a diabetes center are routinely screened for depression by medical personnel during the rooming process, with versions of the PHQ. Using a case-control design, we examined data from individuals who had visits at this center from 1 September 2016 to 31 December 2017 and had results of depression screening available in the electronic medical record (EMR). Inclusion criteria were T1D duration >1 year, age 18–85 years, and absence of dementia or hemoglobinopathy. We excluded adults with evidence of serious renal pathology (glomerular filtration rate (GFR) <30 mL/minute/1.73 m2), since depression is reported to be more common in patients with poor renal function (Muscat et al., 2018). Table 1 displays demographic and health data for the total study sample (N = 249) by depression status.
Participant characteristics by depression status.
PHQ: Patient Health Questionnaire; SD: standard deviation.
Overall, 59.4 percent were female, 92.0 percent defined themselves as non-Hispanic white, with a mean (standard deviation (SD)) age of 41.8 (15.7) years. They had been diagnosed with T1D for a mean (SD) of 21.7 (14.5) years, and had a mean (SD) HbA1c = 8.9 (2.0) percent.
Measures
Depression
Versions of the PHQ are commonly used, reliable, and validated measures for screening and monitoring for depression (Kroenke et al., 2001, 2003). Adults with T1D at our diabetes center are routinely screened for depression using the PHQ-2, a two-item measure that queries about the presence of two core depression symptoms (depressed mood and anhedonia). A PHQ-2 score of 3 or greater has sensitivity for major depression of 83 percent, specificity of 90 percent, and a positive likelihood ratio of 2.9 (Kroenke et al., 2003). If the PHQ-2 score is
The PHQ-9 consists of nine items that measure how many days in the past 2 weeks the individual has experienced any of the nine symptoms of depression defined by the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV). One of the nine symptom criteria (“thoughts that you would be better off dead or of hurting yourself in some way”) is scored positive if present at all, regardless of duration. As a severity measure, the PHQ-9 score ranges from 0 to 27, as each of the nine items can be scored as 0 (“not at all”), 1 (“several of the days”), 2 (“more than half of the days”) or 3 (“nearly every day”). Cut points of 5, 10, 15, and 20 represent the thresholds for probable mild, moderate, moderately severe, and severe depression, respectively. The common single screening cut point chosen for probable major depression is a PHQ-9 score ⩾10, which has a sensitivity for major depression of 88 percent, a specificity of 88 percent, and a positive likelihood ratio of 7.1 (Kroenke and Spitzer, 2002). The ADA recommends the PHQ-9 as an appropriate validated measure for screening and monitoring for depression in people with diabetes. In the current report, participants were defined as “depressed” if the PHQ-9 score was ⩾10 or “not-depressed” if the PHQ-2 score was <3.
The gold standard for diagnosing depression is generally considered to be a structured clinical interview, and some have argued that a self-report survey yields too many false positives (Fisher et al., 2016; Martin-Subero et al., 2017). An alternative to using a PHQ-9 cut-off score is used in the Behavioral Risk Factor Surveillance Survey (BRFSS), based on an algorithm applied to the PHQ for symptoms of depression. The BRFSS is a national system for collecting health-related data through telephone surveys of US residents in all 50 states, completing more than 400,000 adult interviews each year. The BRFSS algorithm scores the PHQ-8, which includes eight of the nine DSM-IV depression criteria (excluding the ninth criterion, thoughts of suicide/self-harm). Based on this algorithm, participants are classified as “depressed” if, for “more than half the days,” they met at least five of the eight criteria, including “little interest or pleasure in doing things” (anhedonia) and/or “feeling down, depressed, or hopeless” (depressed mood) (Centers for Disease Control and Prevention (CDC), 2010). We also note that the PHQ-9, whether scored by cut-off or algorithm, can only define probable major depression, and has been found to have a high rate of false positives in a sample of adults with T1D (Fisher et al., 2016). However, we use the labels “depressed” and “not-depressed” throughout this article for ease of discussion, noting these caveats.
2. Demographic data: age, sex, race, ethnicity, marital status, insurance, and employment status, from EMR
3. Clinical data: HbA1c, duration of diabetes, and body mass index (BMI), from EMR
4. Clinical data from devices (blood glucose meters and continuous glucose monitoring (CGM) devices) (a) Data are downloaded from blood glucose meters (patients are asked to bring their blood glucose meters to office visits). These meters store the following information that is routinely downloaded: frequency of SMBG, number of glucose readings per day, mean and standard deviation (SD) of blood glucose readings, episodes of serious hypoglycemia (defined as blood glucose <54 mg/dL), and the coefficient of variation (CV) (defined as (SD glucose)/(mean glucose) × 100). The CV is recommended to assess glycemic variability, with SD as a key secondary measure (Danne et al., 2017; Monnier et al., 2017). CV ⩽36 percent has been suggested as the threshold to distinguish high from low glucose variability, with high variability defined as CV >36 percent (Monnier et al., 2017). (b) Some individuals use a CGM device, which tracks glucose levels throughout the day and night automatically by taking glucose measurements every 5–15 minutes. For individuals with CGMs, we recorded mean and SD of blood glucose readings and percent of glucose readings below, above, and within target range (70–180 mg/dL).
Procedures
We reviewed EMRs (as described above) of all participants with PHQ-9 scores ⩾10 (“depressed”) and of a selected sample of age and sex-matched participants with PHQ-2 scores <3 (“not-depressed”). This study was reviewed by the Institutional Review Board for the Protection of Human Subjects (IRBPHS) at State University of New York Upstate Medical University, and was determined to be exempt from IRBPHS review.
Statistical analyses
We compared demographics, clinical characteristics, and glycemic measures between the depressed and not-depressed groups for each variable individually using Fisher’s exact tests for categorical variables (e.g. sex, race, and marital status), and Student’s t-tests for continuous variables (e.g. HbA1c, BMI, and CV). All p-values from the two-way tests are reported at the nominal level, with p < 0.05 defining significance. In addition, a multiple regression was run to assess the impact of depression status on A1c levels while controlling for the demographic and clinical variables.
Results
Descriptive analyses
Of 1292 adults with T1D screened for depression, 83 (6.4%) were defined as “depressed” using the PHQ-9 ⩾10 cut-off score system (mean (SD) PHQ-9 score = 16.8 (4.3)). This depressed group was compared to an age and sex-matched group of 166 T1D adults defined as not-depressed using the PHQ-2 <3 cut-off (mean (SD) PHQ-2 score = 0.5 (0.8)). As shown in Table 1, age, sex, duration of diabetes, and BMI did not differ between the groups. Depressed adults were more likely to be non-white, unmarried, and unemployed, and to have Medicaid health insurance (a marker for low socioeconomic status).
The association between HbA1c and depression status is shown in Table 2.
Hemoglobin A1c by depression status.
PHQ: Patient Health Questionnaire; SD: standard deviation; SMBG: self-monitoring of blood glucose noted on meter downloads; CGM: continuous glucose monitoring.
Adjusted p-value.
HbA1c was significantly higher in the depressed group versus the not-depressed group (9.6% ± 2.3% vs 8.6% ± 1.9%, p < 0.001). Within the depressed group, 56.6 percent had HbA1c ⩾9.0 percent (Table 2) as compared to 30.7 percent of the not-depressed group. Depression status had a significant impact on HbA1c, even after controlling for all demographic and clinical variables in a multiple regression analysis. Depression status accounted for an increase in HbA1c of 0.8 percent, on average (p = 0.003).
SMBG and CGM use were associated with lower HbA1c in both the depressed and not-depressed groups.
The relationships of other glycemia and self-care factors to depression in the sub-sample of patients who had blood glucose meter downloads available are shown in Table 3.
Glycemia-related self-care factors by depression status.
PHQ: Patient Health Questionnaire; SMBG: self-monitoring of blood glucose noted on meter downloads; SD: standard deviation; BG: blood glucose; CV: coefficient of variation; CGM: continuous glucose monitor.
CV value was missing for one participant in the depressed group.
Meter downloads were available for 211 of the 249 participants. Based on these meter downloads, the depressed group had significantly higher mean blood glucose levels than the not-depressed group (236.2 ± 75.6 vs 206.4 ± 63.0 mg/dL, p = 0.008). There were no significant differences between groups in mean CV or percentage of participants with CV > 36 percent (high glycemic variability) (Table 3). There was also no difference between the proportions of depressed versus not-depressed participants having at least one occurrence of serious hypoglycemia.
Glycemic control and blood glucose testing and use of meters/CGM devices
Those in the depressed group, on average, were testing their blood glucose 2.8 times daily compared to 3.6 tests per day in the not-depressed group (p = 0.046). Meter use, based upon the patient bringing a meter to the clinic visit, was also significantly less in the depressed group (p = 0.001), although it is possible that some participants used meters but did not bring them to their clinic visit. CGM was used by only 37 of the 249 participants. Despite this small sample, we found that CGM use was significantly less in the depressed group (6%, n = 5) compared to the not-depressed group (19.3%, n = 32, p = 0.005).
Using the BRFSS PHQ-8 scoring algorithm, 62 (4.8%) of those screened with the PHQ-9 were defined as depressed. We have previously reported a similarly low rate of depression using the algorithm as compared to the cut-off (Trief et al., 2014). The results of the other analyses described above, using the algorithm scoring method to define depression, were similar to those obtained using the PHQ-9 cut-off scoring method. All tests that were non-significant using the cut-off, were also non-significant using the algorithm. Most of the comparisons that were statistically significant using the cut-off were also significant using the algorithm, except for two comparisons. Comparing depressed to not-depressed groups on race/ethnicity, the p-value was 0.019 (i.e. statistically significant) using the cut-off, but 0.068 (i.e. non-significant) using the algorithm; for number of blood glucose tests/day, the p-value was 0.046 using the cut-off, but 0.055 using the algorithm (data not shown).
Discussion
This is the first study we are aware of to examine the association of screened depression with a variety of glycemic measures, including objective data from device downloads at the time of depression screening and blood glucose testing, in a sample of adults with T1D seen in a real-world setting. Depression, defined using two well-accepted scoring methods for the PHQ, was associated with worse glycemic control and less SMBG.
Of the 1292 adults with T1D screened for depression using the PHQ-2, 6.4 percent had PHQ-9 scores ⩾10 and thus defined as likely to meet criteria for major depression. Recently published US data from the National Health and Nutrition Examination Survey, which used the PHQ-9 score ⩾10 to estimate depression in adults ⩾20 years of age, reported a prevalence of depression of 7.9 percent in non-Hispanic whites (5.2% in men and 10.5% in women), with the highest prevalence observed in those living below the federal poverty level (Morbidity and Mortality Weekly Report (MMWR), 2018). While our study was not designed to estimate prevalence, our data do support other findings that suggest that depression may not be more common in adults with T1D than in the general population (Fisher et al., 2016; Trief et al., 2014). However, the findings that those who are depressed have worse glycemic control and self-care are quite significant and consistent with previous reports (Trief et al., 2014). The key strength of the study is that we had objective data from blood glucose meter and CGM downloads, and thus analyses were not reliant on self-report. These objective data confirmed that depression was related to poorer glycemic control (both higher HbA1c and higher mean blood glucose) and less SMBG.
Fisher et al. (2016) recently raised concerns about the interpretation of PHQ data to define depression. They compared results using several PHQ cut-off scores to results using the Mood Disorders Module of the Structured Clinical Interview (SCID), considered the diagnostic “gold standard,” and identified a high false positive rate. However, they note that reliability and validity concerns have also been raised concerning the SCID (Wakefield, 2016). One relevant concern is that it results in a categorical diagnosis when severity of symptoms should be considered (Shankman et al., 2018). In contrast to our data, they found no relationships between depression (using any of these varied assessment methods) and HbA1c. However, they did not measure HbA1c on the same day as the depression screening, but within 3 months of PHQ-8 completion. Since depression is diagnosed based on symptoms that occur only 2 weeks prior to the assessment, it is possible that the time discrepancy between the HbA1c and depression measurements affected their results. Other data they present suggest that the PHQ-8 may tap into diabetes distress and not reflect an underlying psychiatric disorder. We also note that there were some small discrepancies noted when we defined depression using the more conservative algorithm scoring method. Since the PHQ measures are routinely used for efficient depression screening, this issue requires further study and clarity.
How might depression affect glycemic control in adults with T1D? There may be shared biological vulnerabilities between depression and glycemia. Several plausible biological mechanisms have been suggested, such as the effects of insulin deficiency on metabolism of neurotransmitters that may underlie depression and potential effects of chronic high blood glucose on the hypothalamic-pituitary-adrenal axis (Korczak et al., 2011). Or, the feelings of hopelessness and helplessness that are hallmarks of depression (Liu et al., 2015) may affect the individual’s motivation to engage in active self-care, and T1D requires a very high level of vigilance and attention that may be difficult to muster if one is depressed. And, knowing that poor glycemic control contributes to the development of future complications, yet having difficulty achieving good glycemic control, feelings of hopelessness and helplessness may arise. Studies are needed to understand the possible biological and/or psychosocial underpinnings of the demonstrated relationship between depression and glycemic outcomes in adults with T1D.
There was less use of CGM in the depressed group. If a patient is depressed, he or she may be less likely to request a CGM, since it involves a significant increase in patient effort to be of value. Or, providers may perceive depressed patients as less capable of using CGM and therefore do not recommend it. However, in those using CGM, the mean HbA1c of the depressed group did not significantly differ from the mean HbA1c of the not-depressed group (7.4% vs 7.8%, p = 0.521), suggesting that CGM use should not be withheld in the presence of depression. As CGM becomes more widely available and recommended, the use of CGM in depressed individuals should be examined more thoroughly.
Strengths and limitations
The use of objective meter and CGM downloads is a significant strength. Also, these data were derived from individuals living with T1D in a real-world setting, where depression screening is routine and performed in >90 percent of all T1D patients. Limitations include the small number of participants using CGM, and that this is a one-center study, with mostly non-Hispanic white adults with T1D. As such, these results may not generalize to a more diverse sample. And, we did not have data about complications and co-morbidities that may have affected results.
In conclusion, adults with T1D who were defined as depressed using the PHQ-9 were found to have worse glycemic control and were less likely to do regular blood glucose self-monitoring, putting them at higher risk for diabetes-related complications. A better understanding of the role of depression in T1D management should inform future research aimed at improving the health outcomes and quality of life for adults with T1D. Whether successful treatment of depression can result in improvement in these outcomes will require further investigation.
Footnotes
Declaration of Conflicting Interests
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Drs Trief and Weinstock receive National Institutes of Health research grant support. In addition, Dr Weinstock receives research grant support from Medtronic, Minimed, Oramed Ltd, Kowa Research Institute, Diasome Pharmaceuticals Inc, and the Jaeb Center for Health Research.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported in part by the Type 1 Diabetes Exchange Quality Improvement Collaborative; and Unitio. The funding sources had no role in the study design, collection, analysis or interpretation of data.
