Is therapeutic inertia present in hyperglycaemia, hypertension and hypercholesterolaemia management among adults with type 2 diabetes in three health clinics in Malaysia? a retrospective cohort study

At baseline, participants were sampled consecutively as they came to the clinics over a period of six months. Inclusion criteria at baseline: age 30 years or older, a diagnosis of T2D more than one year ago, and with at least three clinic visits in the previous year. The baseline exclusion criteria were pregnancy or lactating, psychiatric/ psychological disorders that could impair judgment and memory, and participants who could not read or understand English, Malay or Mandarin [29]. The participating health clinics were chosen because they serve different sections of the local population. One health clinic (SK) is urban and is visited mainly by patients of Chinese descent, the second is a rural clinic (DK) visited by proportionally more patients of Indian descent than found in a usual public health clinic, and the third clinic (SL) is in a rural and predominantly visited by the Malays. Before answering the questionnaires in their preferred language, all participants gave written consent while waiting for a medical consultation with the clinic’s doctor. Trained research assistants interviewed participants at baseline who were not able to self-administer the questionnaires. The study protocol was approved by Medical Research Ethics Committee (MREC), Ministry of Health Malaysia.

Baseline demographic data included age, gender, ethnicity, religion, educational level, employment status, monthly income and life event within the past six months [30]. Structured case record forms were used for data collection from the medical records. These included duration of diabetes, HbA1c, diabetes-related complications, blood pressure, lipids, number and type of medication use [29]. At follow-ups, participants were identified by an orange label on their follow-up cards and on their medical records. Follow-up data were retrieved from the medical records on glycaemic control (HbA1c), systolic (SBP) and diastolic blood pressure (DBP), LDL-cholesterol and prescribed medications [29]. These medications included oral anti-diabetic agents (ADA), insulin, oral anti-hypertensive agents (AHA) and lipid-lowering agents (LLA). Non-participants were those who lost to follow-up at the participating clinics and non-attendance for at least more than one year between 2014 and 2016. There was no other injectable ADA besides insulin at the three participating clinics during the period of the study. Every result of HbA1c and LDL-cholesterol, and every reading of SBP and DBP were retrieved. The recommended treatment targets are HbA1c < 7.0% (53 mmol/mol), blood pressure < 140/90 mmHg and LDL-cholesterol < 2.6 mmol/L, respectively [31]. Medications use was retrieved in terms of their names, dose and frequency. The medications prescribed when the annual blood tests were reviewed during the follow-up visit were considered as the index prescription. In the event of absence of an annual blood test or results, the medications prescribed in the last follow-up visit of the year were considered as the index prescription.

Data analyses were carried out using SPSS software version 25.0 (IBM Corp., Armonk, NY). Comparisons of mean levels were performed using the Student’s t-test or Mann-Whitney U test according to the data distribution, and the Chi-square test was used for proportionate samples between groups. Characteristics of the patients who have and did not have the required data for this study are presented as mean (SD) or median (IQR) for continuous variables, and counts and percentages for nominal variables using descriptive statistics according to the three health clinics.

Assessment of therapeutic inertia was done for year 2015 because the required medication use data were captured from year 2014, and not possible for year 2016 as the study ended in the third quarter. Index of ADA, insulin and LLA medications prescription and use of the year were compared to the previous year index prescription. Changes of medication use were classified as 1) no change, 2) stepping up: dose increment or/and replacement with a stronger medication, and 3) stepping down: dose reduction or/and replacement with a weaker medication. All the classification was verified by the author himself and the resident family physicians in SK and SL health clinics. Therapeutic inertia is defined as no change in the medication use in the present of not reaching the recommended treatment targets. Therapeutic inertia was assessed for HbA1c (no change in ADA and insulin uses separately) and LDL-cholesterol (no change in LLA use) by looking at the treatment target in the same year because medication use and change was captured after the tests were reviewed. For those who did not have results for HbA1c and LDL-cholesterol in the same year, the results in the previous year were used. This reflects the actual clinical practice at these health clinics. The therapeutic inertia assessment for blood pressure (no change in AHA use) was assessed by looking for occurrence of persistent SBP/DBP above treatment targets over two occasions with no change of AHA use in the same year. Sensitive analysis was conducted to examine the states of therapeutic inertia for AHA use in the first half and the second half of the year when at least four SBP/DBP measurements were available, and at least two consecutive SBP/DBP measurements that were above the recommended treatment targets in the respective 6-month periods. The results were essentially similar.

Possible clinic and patient’s factors on the therapeutic inertia were further evaluated using the multivariable logistic regression analysis. Besides the clinic, covariates with a P value < 0.20 from the univariable analyses were included in the final multivariable analyses. Category ‘no change’ represents therapeutic inertia and category ‘stepping up and stepping down’ was used as the reference group. The final models were re-run with ‘stepping up’ as the reference group and the results did not change the interpretations. We reported the results with the treatment targets entered as one of the covariate, with the reference category ‘stepping up and stepping down’ because this could account for all possible reasons of therapeutic changes and render the modelling clinically more meaningful [32]. Statistical significance was set at P < 0.05.

In total, 552 participants have the required follow-up data for the assessment of therapeutic inertia, representing 78.9% of the baseline sample (n=700). There were no differences in the prescribed medications between the non-participants and participants except LLA use was more among the participants (72% vs. 83%) (Table 1). Compared to the participants, the non-participants were more often treated at the rural Salak Health Clinic, had higher systolic blood pressure, and had dyslipidaemia (Table 1).

Except for ADA and LLA, there were significant differences in insulin and AHA therapeutic patterns when treatment targets were not achieved at the three health clinics (Fig. 1). DK was less likely to have insulin therapy inertia and SK less likely to have AHA inertia compared to the other health clinics. All clinics had high therapeutic inertia in ADA (61–72%) and LLA therapies (56–77%).

Both ADA and insulin therapeutic inertia were observed, and ADA therapeutic inertia was worse than the insulin (64% versus 47%) (see Additional File Table S1) presumably because of limited dosing range and choices of the oral agents. Patients who were prescribed less medication, had shorter diabetes duration and not having dyslipidaemia were likely to experience ADA therapy inertia (see Additional File Table S1). There was no relation between ADA changes and glycaemic control (OR 1.18, 95% CI 0.77, 1.80; P = 0.454). Insulin therapeutic inertia was more likely among those with shorter diabetes duration (adjusted OR 0.9, 95% CI 0.87, 0.98) (Table 2).

AHA inertia was observed in about half (51%) of the participants with BP > 140/90 mmHg, and only 34% had their AHA therapy stepped up. This inertia happened more among the younger age group, Indians, in active employment, shorter hypertension duration and not diagnosed of hypertension (see Additional File Table S2). Those who had at least two consecutive BP > 140/90 mmHg were more likely to experience AHA therapy intensification (Table 3).

Therapeutic inertia in LLA was observed in 61% of those who had uncontrolled LDL-cholesterol, and this was more common among those having hypertension on top of T2D (see Additional File Table S3). However, no factor was an independent risk factor to LLA therapy inertia after adjusting for LDL-cholesterol treatment target (Table 4). Those who had not achieved LDL-cholesterol treatment target < 2.6 mmol/L were more likely to experience lipid-lowering therapy intensification.

Good sample size, representative samples of the participants [29] and reasonable statistical analyses rendering the study ability to produce credible answers. However, this study was subjected to several potential limitations. This study did not classify diabetes pharmacological regimen as a whole but ADA to insulin changes separately. Thus, escalation of treatment from ADA to insulin therapy in the event of HbA1c ≥ 7.0% was not recorded. Therefore, insulin therapeutic inertia provided an estimate that was closer to the actual therapeutic practices compared to the other studied therapies in the clinical management of adults with T2D. Insulin treatment was recorded from none to initiation, and to dose escalation without the constraints of medication choices when the maximum doses have been reached. Treatment targets used did not reflect risk profiles and customized target levels for the patients. This may lead to underestimation of the proportions of therapeutic inertia because majority of the adults T2D at the primary care setting were of the lower risk and in early stages of diseases, and treatment intensification should occur at lower target levels [41, 42]. Similarly, underestimation of the proportions of therapeutic inertia is likely in the participants especially in ADA and LLA since the non-participants were prescribed less of these medications, and the included cohorts were those tend to follow-up attendance at baseline. The three participating health clinics in this study are of the medium scale with resident family physicians and doctors, and situated in a developed state of Selangor across urban, sub-rural and rural regions. This limits generalizability of the results to other clinics of different scales, without resident doctors or in the more remote areas in the country that may have different healthcare systems and delivery at the meso- and micro-levels.