Disease severity is associated with the use of complementary medicine to treat or manage type-2 diabetes: data from the 2002 and 2007 National Health Interview Survey

Self-management is the cornerstone of overall diabetes management[1]. Self-management of type-2 diabetes requires complex, continual and demanding self-care behaviors, including dietary control, exercise and frequent medication. More active self-management is generally believed to result in better metabolic control and higher quality of life[2] while also being more cost effective than standard pharmaceutical therapies alone[3, 4]. Yet failure to follow treatment recommendations is reported as a serious and widespread problem in patients with type-2 diabetes and believed to lead to diabetic complications such as blindness, poor wound healing, neuropathy and kidney failure.

Hernanderz et al.[5] and Paterson and Thorne[6] have demonstrated that learning to master self-care in diabetes is a process where the person attempts a variety of self-care strategies according to her or his unique circumstances until discovering what is effective for their own lifestyle and contextual situation. It has been proposed that individuals with diabetes view complementary medicine (CM) as another option for their self management along with conventional options[7], and that individuals highly motivated to control their diabetes will try multiple therapies available to them including CM[8]. This is consistent with numerous observations that CM therapies are predominately used as complements to conventional care rather than in lieu of conventional care[9, 10].

The overall prevalence of CM use among adults in the United States with diabetes has been examined both in representative national samples[11‐13] and in more restricted populations[7, 8]. Prevalence rates in these studies varied from about 30% to 70% depending on the definition of CM used. Much lower rates are observed when examining the prevalence of CM use specifically to treat or manage diabetes – from about 20% of all adults with diabetes in 1997[11] to about 6.7% of all adults with diabetes in 2002 (recalculated from[12]). None of these studies attempted to identify predictors of CM use to treat diabetes among the populations sampled, nor looked for a relationship between CM use and diabetes severity.

In the general population, a number of characteristics are associated with CM use: gender (being female), race/ethnicity (being non-Hispanic White), age (being middle-aged), and education (having a college degree) to name a few[10, 14, 15]. Other predictors of CM use include poor perceived health status and the presence of multiple health complaints[14‐17]. Thus, one might hypothesize that individuals with diabetes are more likely to use CM if they have more severe diabetes and/or one or more complications associated with the disease. However, there are also data from the general population showing that CM is associated with a number of positive health behaviors that would be part of a wellness lifestyle[18] including regular levels of exercise[15], nonuse of tobacco[15, 19, 20], nonuse or moderation in use of alcohol[15, 21], healthy diet choices[21] and preventive screening[22]. It may also be that individuals oriented toward a wellness lifestyle will be more likely to use CM for their diabetes. The present study seeks to address these two, not mutually exclusive hypotheses through analysis of data from the National Health Interview Survey (NHIS). In both 2002 and 2007, the NHIS included extensive questions on the use of a wide range of CM therapies. The NHIS also contained a limited number of questions on the reasons why an individual might have chosen to use a given CM therapy. We analyzed the responses to two of these questions to examine reasons for use of CM among individuals who used CM to treat/manage their diabetes.

We found that 7.2% (S.E. 0.14) of surveyed adults reported having physician diagnosed diabetes. This equates to roughly 15.3 million adults in the United States. Of these, 95.5% (S.E. 0.4), or 14.6 million, were coded as having type-2 diabetes. The majority of individuals with type-2 diabetes had the disease longer than 5 years, were taking either insulin or oral hypoglycemic medication, did not have a functional limitation related to diabetes, did not have weak/failing kidneys or severe vision problems, were at least 45 years old, had greater than a high school education, were Non-Hispanic whites, had private health insurance, were overweight or obese, had hypertension but not coronary heart disease, had smoked tobacco and consumed alcohol at some point in their lives, were not depressed or anxious, and were unable to perform vigorous leisure activity but nevertheless considered themselves in at least good health (Table 1). About equal numbers of men and women had type-2 diabetes. In contrast, individuals with type-1 diabetes predominately had their diabetes more than 10 years, were under 45 years of age, did not have a functional limitation related to diabetes, did not have weak/failing kidneys, severe vision problems, hypertension, coronary heart disease or depression, yet considered their health as only fair or poor (Table 1).

Consistent with the literature, we found that 30.9% (S.E. 0.96) of individuals with type-2 diabetes used CM for some reason, equating to roughly 4.4 million adults (data not shown). However, only 3.43% (S.E. 0.37) of individuals with type-2 diabetes used CM specifically to treat or manage their diabetes versus 7.06% of those with type-1 diabetes (Table 1). Among those using CM to treat/manage their type-2 diabetes, 63% have had their diabetes for more than five years, and 77% used both CM and conventional prescription medicine for their diabetes (Table 2). Only a quarter of individuals using CM to treat/manage their type-2 diabetes reported a functional limitation resulting from their diabetes, severe vision problems or anxiety/depression (Table 2). Very few individuals using CM for their diabetes reported weak/failing kidneys or coronary heart disease (Table 2).

Within the group using CM to treat or manage their diabetes, we observed that a number of sociodemographic and health status variables were associated with CM use at p<0.1 (Table 2): age, race/ethnicity, region of residence, number of co-morbid conditions, BMI level, the ability to perform vigorous leisure activity at least once per week and perceived health status. These variables were included in all adjusted logistic regression models as covariates.

To our knowledge, this is the first report to examine the association between the use of CM and severity of diabetes in a large, nationally representative population sample. We found that only a small proportion of individuals coded with type-2 diabetes used CM to treat or manage their diabetes (3.4%). However, in this small cohort, persons with the most severe disease (as assessed by having 3 or more measures of diabetes severity) had almost twice the odds of using a complementary therapy as those with less severe disease. This association was relatively insensitive to the inclusion of anxiety/depression as a diabetes complication, and to changes in the age at diagnosis cutoff for differentiating type-1 and type-2. Furthermore, we found that persons who had diabetes 10 or more years and those who had a functional limitation resulting from their diabetes were more likely to use at least one CM therapy. Our study could not determine whether the use of CM is a direct result of diabetes severity or whether other unmeasured and uncontrolled factors motivate use.

While our observation that 30.9% of individuals with type-2 diabetes used CM for any reason is consistent with early reports[11, 12], our finding that only 3.4% of individuals with type-2 specifically use CM to treat or manage their diabetes is substantially lower than that reported by Yeh et al., 2002 (20%)[11]. Although it is possible some of these discrepancies reflect real changes in use over time, it is more likely the discrepancies reflect methodological differences in the two surveys. In calculating their prevalence of CM use in 1997, Yeh and colleagues[11] included four types of therapies not included in the present analyses: folk remedies (2.9%)(included in the 2002 but not 2007 NHIS), commercial diets (6.4%), high dose vitamins (1.7%) and self-help groups (2.3%). In addition, Yeh et al.[11] found higher use of herbal medicines to treat or manage diabetes then we did for NVNMDS in the present study, 6.7% vs. 1.1%, respectively. This discrepancy is most likely explained by that fact that Yeh et al.[11] employed a global question on use of “herbal medicine”, while in the NHIS participants were asked if they used specific NVNMDS on a list.

While the inverse association between the use of CM and the use of prescription diabetes medications failed to reach significance in either the unadjusted or adjusted logistic regression models, some comment is still warranted given concerns of herb-drug interactions[38] or the substitution of CM therapies for proven conventional therapies for diabetes[39]. There are two competing interpretations of our observation: 1) those using CM are more likely to forego needed conventional treatment; or 2) individuals with less severe disease not requiring medical interventions are more likely to use CM. When examining CM categories, we observed that diabetes medication use predominates in CM users across categories except in participants using multiple CM therapies (only 35.3% of whom used diabetic medications). Except for multiple CM users, the present CM category data are consistent with a regional survey of older rural adults diagnosed with diabetes (N= 679)[40], where no difference in diabetes medication adherence was found between those who used CM and those who didn’t.

Consistent with our observation that most individuals (77%) who use CM for their diabetes also use conventional diabetes prescription medications are the data of Garrow and Egede[41] who found that individuals with diabetes using CM (for any reason) were more likely to receive regular preventive care than were individuals with diabetes who did not use CM. These quantitative findings are supported by qualitative research[8] suggesting that individuals with diabetes who try CM were highly motivated to control their diabetes and to try all methods available to them, and did not use CM as a substitute for conventional care.

We are aware of only one other article that discussed demographic characteristics of individuals in the United States using CM specifically to treat or manage their diabetes[7]. In this survey of rural older adults in North Carolina, it was found that in those using CM for their diabetes, home remedies were the prevalent type of CM therapy used, with their use associated with race/ethnicity and the number of co-morbidities. In the present study, we also found that the number of co-morbidities was associated with the overall use of CM but did not see a relationship between CM use and race/ethnicity. Unfortunately, the NHIS did not specifically look at use of home remedies, which may explain the differences in our findings from Arcury et al.[7].

Our analyses revealed an interesting paradox in that persons with more severe diabetes are more likely to use CM to treat/manage their diabetes, but persons with poorer perceived general health status were less likely to use CM (OR 0.56, 95% CI .33-.86). Using data from the 2007 NHIS, Nguyen and colleagues[42] observed a similar paradox in the general population where individuals “using CAM were more likely to have chronic illness . . . yet also were more likely to report that their health status was excellent and better than the prior year.” Nguyen et al.,[42] suggested two possible reasons for the paradox, both of which also apply to our analyses: 1) different timeframes for the health status questions (current status) and illness self-reports (previous 12-months); 2) positive expectations of CM use may positively impact perceived health status.

We found that relatively few individuals with diabetes who used CM for their diabetes used more than one type of CM therapy (<10%). This is in contrast to the general population in 2007 (unpublished observation) where 41.4% of CM users used more than one type of therapy, or in 2002 or 1997 where, respectively, 33.4%[43] and 54%[44] of CM users used more than one type of therapy (as in the present analysis, these studies counted use of multiple NVNMDS or multiple diet-based therapies only once each).

This study has several limitations. First, the variables being investigated were self-reported. The scientific literature suggests that most people tend to under-report negative health behaviors[45]. Hence, the effects of alcohol consumption[46, 47] or smoking behavior[48] may have been underreported in this study. Second, the cross-sectional nature of the data prevents us from establishing whether the diagnosis of diabetes actually occurred prior to and resulted in the diseases/conditions we considered as complications of diabetes. Thus, it is likely that for some unknown percent of participants, non-complications were misclassified as complications. From our data, we cannot predict how such misclassification would impact on our results. These cross-sectional data also do not let us investigate the possibility of cohort effects and secular trends in the associations between diabetes severity and CM use. Third, the lack of confirmed diagnoses of type-1 and type-2 diabetes makes some degree of misclassification likely. Based on our criteria for differentiating between type-1 and type-2 diabetes and on published NHANES data[49], we can predict that 7-9% of the cases we classified as type-2 were in fact type-1. Conversely, given the increasing prevalence of type-2 diabetes in youth, we can predict that some unknown proportion of individuals we classified as probably having type-1 diabetes actually had type-2. We sought to more completely capture the population of individuals with type-2 diabetes by lowering the cutoff age for differentiating type-1 and type-2. We found that lowering the cutoff age to less than 10-years old, a cutoff likely to capture almost all, if not all individuals with type-2 diabetes[34], did not substantially change our findings. However, clearly our study needs to be replicated in a dataset that incorporates confirmed physician diagnosis of diabetes by type and also contains information on the use of CM. Fourth, it is possible that additional measures of diabetes severity (as well as other unexplored factors) explain more of the observed relationships. However, the six measures employed are predictors of mortality and morbidity in diabetes[50, 51]. Fifth, because our primary focus was to identify factors associated with the use, versus nonuse, of CM, a dichotomous dependent variable was utilized for the primary analysis. By doing so, information on frequency and/or intensity of use was lost. It may be that substantial differences exist between heavy and light users of one or more CM modalities[44]. Sixth, it has been found that the use of specific types of CM therapies is associated with specific personality styles[52]. These associations might confound our results if specific personality styles (e.g., “openness” or “control”) are also related to better diabetes control. Finally, despite combining data from two very large national surveys in the United States, the number of respondents using CM specifically to treat their diabetes was small (141 respondents, equivalent to 3.4% of individuals with type-2 diabetes). While this may have increased the chance of type-2 error (false negatives) in our analyses and certainly prevented assessment of individual CM therapies, it does not change our confidence in the positive associations we did observe.

Our results demonstrate that individuals with more severe diabetes are more likely to use CM independent of sociodemographic factors. Further studies are essential to determine if CM therapies, either alone or in combination with conventional approaches, actually improve clinical outcomes when used to treat/manage diabetes. Also warranted are studies investigating the cause and effect relationships underlying the associations between diabetes severity and CM use. Knowledge of these specific patterns of use may contribute to tailoring health education programs for diabetes. For health care providers, understanding the motivations behind a patient’s use of CM may assist in the design of an optimal treatment plan[53]. The fact that users of CM appear willing to take active control of their health[54‐56], suggests that this group of individuals with diabetes may well be open to additional recommendations toward managing their disease.