A cross-sectional study of health-related quality of life deficits in individuals with comorbid diabetes and cancer

Diabetes is a chronic medical condition that affects approximately 5% of Canadians aged 20 years and older, with type 2 diabetes accounting for 90% of all diagnosed cases of diabetes [1]. Type 2 diabetes is associated with several microvascular complications, such as retinopathy, nephropathy, and neuropathy, and macrovascular complications, such as heart disease, which results in significant morbidity and mortality [2–4]. A reduced self-reported health status or health-related quality of life (HRQL) reflects the significant health burden in this patient population [5–8].

In addition to the commonly recognized micro- and macrovascular complications, diabetes is associated with other comorbidities. A number of epidemiologic studies have identified an increased risk of developing cancer in people with type-2 diabetes [9–12]. The association appears to be mediated through the metabolic syndrome (also known as the insulin resistance syndrome). The metabolic syndrome is present in almost one-half of all older individuals and is a condition associated with hyperinsulinemia, insulin resistance and a predilection to type 2 diabetes [13].

It has been suggested that hyperinsulinemia combined with insulin resistance might promote carcinogenesis [14, 15]. Several types of cancers have been found to be associated with type 2 diabetes, such as breast cancer [10, 11], endometrial cancer [16, 17], pancreatic cancer [11], and colorectal cancer [11, 12].

The potential HRQL deficits associated with patients who have both type 2 diabetes and cancer may be quite large. There is extensive evidence in the literature of a reduced HRQL in patients with cancer [18, 19]. However, despite the recognition of the link between type 2 diabetes and cancer, very little is known about HRQL in individuals with these comorbid chronic conditions. The objective of this study was to assess the HRQL burden in the following groups: 1) Individuals with comorbid diabetes and cancer compared to individuals with either condition alone, and 2) Individuals with comorbid diabetes and cancer compared to individuals without either condition. In both cases, we hypothesized that individuals with comorbid diabetes and cancer would have a significantly worse HRQL.

The unadjusted mean (standard deviation, SD) overall HUI3 for the comorbid diabetes and cancer group was 0.67 (0.30). This value was significantly lower than that for the other groups: 0.78 (0.27) for diabetes alone, 0.78 (0.25) for cancer alone, and 0.89 (0.18) for the reference group (Table 1).

Respondents who were younger, married/common law, physically active, not obese, not depressed, and had no chronic medical conditions had significantly higher overall HUI3 scores (Table 2). On the other hand, males, smokers, or respondents that had not completed high school had significantly lower overall HUI3 scores (Table 2). Of note, there were clinically important differences in the following variables for overall HUI3 score: age, physical activity level, education level, depression status, and number of chronic medical conditions (Table 2).

We used population-based data from the Canadian Community Health Survey Cycle 1.1 to determine HRQL, using the overall HUI3 score, in respondents with self-reported diabetes and cancer compared to a reference group without diabetes or cancer. As hypothesized, we found that patients with diabetes and cancer had a significantly lower and clinically important difference in overall HUI3 score compared to respondents who had no diabetes or cancer, even after controlling for potential confounding variables.

All covariates included in the model were associated with significant differences in overall HRQL; however, not all of the variables revealed clinically important differences. Number of chronic medical conditions and specific comorbidities, such as depression, had the largest impact on HRQL. Respondents who were not depressed had a 0.14 higher HUI3 overall score compared to respondents who were depressed. Also, the fewer chronic medical conditions apart from diabetes and cancer that respondents had, the higher their overall HUI3 score (0.15 for no chronic medical conditions compared to respondents who had 3 or more chronic medical conditions). These findings are in agreement with Maddigan et al, who found that comorbidities such as cardiovascular disease and depression had the largest impact on HRQL in patients with type 2 diabetes [7, 8].

Previous research has reported lower overall HUI3 scores in individuals that have either type 2 diabetes [7, 25] or cancer [18, 19]. These deficits are likely a result of the complications, comorbidity, and treatment regimens associated with these chronic conditions. There is strong evidence of an association between type 2 diabetes and cancer; this association appears to be mediated through the metabolic syndrome [9–11]. Furthermore, there is an increased mortality for patients with comorbid diabetes and cancer [9, 39]. Despite these associations, however, very little is known about the HRQL deficits in this patient population.

The overall HUI3 scores we observed in our sample are similar to that of other studies. A study by Maddigan et al, observed the same mean overall HUI3 score of 0.78 in patients with type 2 diabetes as our study did in patients that have diabetes; this finding is not surprising considering the authors also used CCHS data [32]. They also observed mean overall HUI3 scores of 0.77 for arthritis patients (a figure similar to the 0.78 we found for our cancer group), 0.54 for patients who have had a stroke, and 0.90 for the general population (a figure similar to the 0.89 we found for our reference group that did not have diabetes or cancer) [32].

There are some limitations inherent in this study. Firstly, the cross-sectional study design of the CCHS is not ideal for the purpose of answering this question. A longitudinal study design would more appropriately address this question, as changes in HRQL over time could also be assessed. Since both are chronic conditions, however, the duration of follow-up would be prohibitive. Furthermore, as a cross-sectional study, in conditions which have substantial mortality such as cancer, there is likely a survival bias and respondent bias, with only those more healthy individuals in the community being respondents. This would, of course, lead to an over estimate of the respondents in the cancer groups.

We also lacked information on potentially useful clinical variables. Recent research has revealed that waist circumference or waist to hip ratio may be more effective than BMI in predicting risk for type 2 diabetes or cancer [40, 41]. Furthermore, there is evidence that obesity, as measured by waist to hip ratio and/or waist circumference, are associated with various measures of HRQL [42, 43]. It would also be useful to have clinical information on the different types of cancer and on disease severity for both diabetes and cancer, all of which would differentially impact the HRQL of individuals [26]. Although type of cancer and a variable that allows calculation of time since cancer diagnosis are collected as part of the CCHS Cycle 1.1 microdata, these variables are not available in the CCHS PUF.

Another limitation of this dataset was the inability to separate the diabetes cohort into type 1 diabetes and type 2 diabetes. The group with type 2 diabetes may have been of particular interest, as there is well-documented evidence of an increased risk of various types of cancer in patients with type 2 diabetes [10–12, 17]. However, there is no literature to support the evidence of a link between type 1 diabetes and cancer, and we could expect that approximately 10% of the diabetes cohort has type 1 diabetes [1, 8]. Insulin use is available in the CCHS PUF, there is a substantial amount of missing data on this variable (95.2% or 124,609/130,880). Also, because the variables in the CCHS are self-reported, there is a potential for recall bias or social desirability bias. For example, there is evidence that respondents may respond in a socially desirable manner when self-reporting their BMI (height and weight) [44], smoking status [45], and physical activity status [46].

We recognize that the PUF version of the CCHS data is not as precise as the micro data available through Statistics Canada. In the PUF, some data elements are aggregated, and variables that have been collapsed, and more importantly there is no information on the single attribute utility scores from the HUI3. There were also missing data on a number of the covariates (particularly for physical activity level, 6.5%; depression status, 3.0%; and BMI 2.3%). We noted that respondents with missing data on any of these three variables were younger and had lower overall HUI3 scores. There was no relationship between missing BMI and reporting diabetes or cancer. Respondents with cancer were more likely to be missing data on physical activity, although the differences were small. Respondents with both diabetes and cancer were more likely to be missing data on depression. As such, excluding these individuals likely overestimated the HUI3 scores across all groups. Of note, the diabetes and cancer group was most likely to have had missing data on depression; had those individuals been included, the overall HUI3 score would have been even lower. One variable in particular that had a substantial amount of missing data (>10% for all respondents) was income; therefore, this variable was not included in our analyses. It has been shown that income, as a determinant of health, is a distinct aspect of socioeconomic status that is useful in predicting HRQL, independent of education level [47].

Despite the above mentioned limitations in this study, there are some key strengths that must be recognized. Most importantly, the CCHS has a very large sample that is considered to be representative of 98% of the Canadian population. Also, we used the HUI3 as our measure of HRQL. There is evidence of the validity of this measure in people who have type 2 diabetes [27] and in people who have cancer [48]. It has also been used in other large national population health surveys [7, 21, 22]. Finally, to our knowledge, this is the first study that has assessed HRQL, using the HUI3 as an outcome measure, in patients who have comorbid diabetes and cancer.

We found that patients with comorbid diabetes and cancer had a clinically meaningfully worse overall HUI3 score compared to respondents who had no diabetes or cancer, differences which were considerable, even after controlling for potential confounding variables. While our results are intriguing, they should be considered hypothesis-generating given the limitations inherent in this study. Diabetes (especially type 2 diabetes) and cancer are largely preventable through lifestyle. Separately, the public health burden of these two chronic diseases is large; together their burden is even greater. A better understanding of the relationship between diabetes and cancer, and their associated comorbidities and complications may have important implications for prevention and management of these chronic conditions. These prevention and management strategies will in turn have positive effects on the HRQL of individuals affected by these conditions.