Changes In HbA1c Level Over A 12-Week Follow-up In Patients With ...

Results

A total of 93 eligible patients were recruited to the study from eighteen health centres. Of those recruited, two patients were withdrawn from the study: one patient died and one failed to take their medication. Of the remaining 91, 8 patients changed their type or dose of medication before end of follow-up and an additional 4 patients did not attend for a 12-week HbA1c measurement leaving 79 patients for analysis. A flow chart of study recruitment and loss to follow-up is shown in Figure 1. The 79-patient analysis population had the following baseline characteristics (shown in Table 1): mean (sd) age was 61.3 (10.8) years, 27 (34%) were female, diabetes duration of 6.0 (4.3) years and baseline HbA1c of 8.7±1.5% (72.0±16.8 mmol/mol). The index medication change was a dose increase of an existing drug in 39 of 79 patients, new drug in 39 patients, and new drug plus dose increase of existing drug in 1 patient. Baseline characteristics for all recruited patients and the excluded patients were similar (Table 1).

Download:

• PPTPowerPoint slide
• PNGlarger image
• TIFForiginal image

Figure 1. Study flow chart.

https://doi.org/10.1371/journal.pone.0092458.g001

Download:

• PPTPowerPoint slide
• PNGlarger image
• TIFForiginal image

Table 1. Baseline characteristics of included patients.

https://doi.org/10.1371/journal.pone.0092458.t001

Mean±standard error (se) change in HbA1c, shown in Table 2 and Figure 2, was −0.11±0.04% (−1.2±0.4 mmol/mol) after 2 weeks, −0.25±0.05% (−2.8±0.6 mmol/mol after 4 weeks, −0.51±0.08% (−5.6±0.9 mmol/mol) after 8 weeks and −0.65±0.11% (−7.1±1.3 mmol/mol) after 12 weeks. Results were similar when two patients who had two simultaneous medication changes were excluded from the analysis. When thirteen patients who did not have their 12 week HbA1c measured within 3 days of 12-weeks from their baseline visit were excluded in a sensitivity analysis, results for the remaining 67 patients were similar (Table 2). These 67 patients had mean change (±se) in HbA1c at 2, 4, 8 and 12 weeks of −0.12±0.04% (−1.4±0.4 mmol/mol), −0.27±0.05% (−3.0±0.6 mmol/mol), −0.47±0.09% (−5.1±1.0 mmol/mol) and −0.56±0.11% (−6.1±1.2 mmol/mol) respectively. There was a small non-significant overall increase in mean (sd) weight (0.16±2.73 kg) and decrease in waist circumference (−0.75±5.0 cm) of participants during study participation. Histograms were plotted to show the distribution in the change in HbA1c at each measurement point (Figure 3); these show that there is an increase in the spread of the change in HbA1c at each measurement point and that some patients had an increase in their HbA1c after their baseline measurement and medication change.

Download:

• PPTPowerPoint slide
• PNGlarger image
• TIFForiginal image

Figure 2. Mean change in HbA1c in mmol/mol at 2, 4, 8 and 12 weeks after an increase in diabetes medications and 95% confidence intervals.

https://doi.org/10.1371/journal.pone.0092458.g002

Download:

• PPTPowerPoint slide
• PNGlarger image
• TIFForiginal image

Figure 3. Histograms showing spread of change in HbA1c 2, 4, 8 and 12 weeks after a change in diabetes medication.

https://doi.org/10.1371/journal.pone.0092458.g003

Download:

• PPTPowerPoint slide
• PNGlarger image
• TIFForiginal image

Table 2. Mean (sd) HbA1c and change in HbA1c in mmol/mol for all patients and excluding patients with measurements outside 3-day window.

https://doi.org/10.1371/journal.pone.0092458.t002

By week 12, thirty-one of the 79 patient analysis population (39%) had achieved glycaemic control. Forty-eight patients who did not achieve control at 12 weeks had a much smaller overall mean change in their HbA1c (−0.3±0.7%, −3.6±7.8 mmol/mol) compared with those who did achieve control (−1.1±1.2%, −12.6±13.3 mmol/mol). Forty-nine patients (62%) had not yet achieved glycaemic control at 8 weeks; of these five did achieve control after 12 weeks though none had an HbA1c below 7.1% (54 mmol/mol).

Forty-two patients achieved a greater reduction in HbA1c at 12 weeks than at 8 weeks, but 37 had either the same or higher HbA1c levels at 12 weeks compared with their 8 week levels. Sixteen patients (20%) had higher HbA1c levels 12 weeks after their medication change than they had at baseline; fourteen of these patients had already exceeded their baseline HbA1c after 8 weeks.

Figure 4 shows graphs of changes in HbA1c at weeks 2, 4 and 8 plotted against change in HbA1c at 12 weeks. Each one of these showed a significant correlation between change at the earlier testing times and the 12-week change in HbA1c with corresponding correlation coefficients of 0.46 (95% CI 0.27 to 0.62), 0.58 (95% CI 0.41 to 0.72) and 0.91 (95% CI 0.87 to 0.95) respectively. Linear regression modelling showed a significant correlation between change in HbA1c at 2, 4 and 8 weeks with change in HbA1c at 12 weeks, with coefficients (SE) of 1.18 (0.35), 1.21 (0.20) and 1.16 (0.06) and R2 values of 0.13, 0.34 and 0.83 respectively. The change in HbA1c at each time interval was significantly predictive of the 12-week change in HbA1c (p = 0.001 at 2 weeks, p<0.0001 at 4 and 8 weeks). Among patients who had their 12-week HbA1c tested within 3 days either side of the date 12 weeks from baseline, the correlations of change in HbA1c at 2, 4 and 8 weeks with the 12-week change, with correlation coefficients were 0.52, 0.58 and 0.92 respectively, slightly higher than those for the full analysis population.

Download:

• PPTPowerPoint slide
• PNGlarger image
• TIFForiginal image

Figure 4. Correlation between 2 week, 4 week and 8 week HbA1c versus change in12 week HbA1c in mmol/mol.

https://doi.org/10.1371/journal.pone.0092458.g004

Table S1 shows change in HbA1c stratified by high, moderate and low medication adherence. The 15 most adherent patients had a mean change in HbA1c after 12 weeks of −0.5±1.2% (−5.6±13.1 mmol/mol), 50 patients with moderate medication adherence had a mean change in HbA1c of −0.7±1.1% (−7.3±11.7 mmol/mol) and 14 patients who had the lowest reported medication adherence had a mean 12-week change in HbA1c of −0.7±0.6% (−8.0±6.5 mmol/mol). Correlation of the 8 week change with 12-week change was highest in the most adherent patients (R = 0.96).

Thirty-eight patients who received an increase in the dose of a single medication had a mean decrease in their HbA1c by 0.2±0.6% (2.2±6.6 mmol/mol) and 0.3±0.7% (2.7±7.4 mmol/mol) at 8 and 12 weeks respectively. Thirty-nine patients who initiated a new type of medication had a mean decrease in their HbA1c of 0.8±0.6% (8.4±6.8 mmol/mol) and 1.0±1.1% (10.7±11.6 mmol/mol) at 8 and 12 weeks respectively. A detailed breakdown of medication type and change in HbA1c is shown in Table S2.

The ROC curve analysis (Figure S1) showed that 8-week HbA1c correctly classified the majority of the 12 week HbA1c with an area under the curve of 0.954 (95% CI 0.912 to 0.996). For example eight-week HbA1c measurements of 8.2% (66.1 mmol/mol) or above had 100% sensitivity for predicting the people who were still be poorly controlled at 12 weeks; alternatively, of 42 patients who were above an HbA1c cut-off of 7.8% (62 mmol/mol) at 8 weeks, 39 (93%) remained uncontrolled at 12 weeks.