The Efficacy and Frequency of Self-monitoring of Blood Glucose in Non-insulin-Treated T2D Patients: a Systematic Review and Meta-analysis

Abstract

Background

Self-monitoring of blood glucose (SMBG) is a useful tool in diabetes management, but its efficacy and optimal application in type 2 diabetes (T2D) patients treated without insulin have been controversial. We aimed to evaluate the efficacy of SMBG in controlling blood glucose levels in non-insulin-treated T2D patients and to determine the optimal frequency and the most appropriate population to benefit from SMBG.

Methods

Eligible publications from January 2000 to April 2022 were retrieved from PubMed, Embase, Cochrane Library, and ClinicalTrials.gov databases. Randomized controlled trials comparing SMBG with no SMBG or structured SMBG (S-SMBG, SMBG with defined timing and frequency of glucose measurements) were included. Meta-analyses and sub-analyses were performed to assess the efficacy, optimal frequency, and most appropriate population for SMBG. Risk of bias was assessed regarding randomization, allocation sequence concealment, blinding, incomplete outcome data, selective outcome reporting, and other biases.

Results

Twenty-two studies involving 6204 participants were identified, including 17 comparing SMBG with no SMBG and 4 comparing SMBG with S-SMBG. SMBG reduced HbA1c (MD −0.30%, 95% CI −0.42 to −0.17) compared with no SMBG, and S-SMBG performed better than SMBG (MD −0.23%, 95% CI −0.38 to −0.07). Subgroup analyses showed that HbA1c control was better with SMBG at 8–11 times weekly (MD −0.35%, 95% CI −0.51 to −0.20) compared with other frequencies and with lifestyle adjustments (MD −0.37%, 95% CI −0.50 to −0.23) than with no adjustments. No significant differences in HbA1c were observed between baseline HbA1c subgroups (≤ 8% and > 8%, P = 0.63) and between diabetes duration subgroups (≤ 6 years and > 6 years, P = 0.72), respectively.

Discussion

SMBG was effective for controlling HbA1c in non-insulin-treated T2D patients, although lacking detailed monitoring design. Better outcomes were seen with SMBG at 8–11 times weekly and lifestyle adjustment based on SMBG results.

Trial Registration

PROSPERO (CRD42021285604)

Supplementary Information

The online version contains supplementary material available at 10.1007/s11606-022-07864-z.

INTRODUCTION

According to the 9th edition of the International Diabetes Federation (IDF) Diabetes Atlas report, 1 in 10 adults worldwide have diabetes¹. The total diabetes-related health expenditure is USD 966,000, with a diabetes-related health expenditure per person of USD 1838.4, indicating a heavy burden on society¹.

Blood glucose monitoring to control glycemia is a fundamental part of diabetes management². The invention of portable glucometers has made self-monitoring of blood glucose (SMBG) possible³. SMBG is an adjunct management tool that aids in blood glucose control by helping patients make lifestyle changes and assisting physicians in selecting treatment options^3,4. Acknowledged by the American Diabetes Association (ADA)⁵, the IDF⁶, and the National Institute for Health and Care Excellence⁷, SMBG is a useful tool for the management of patients with type 1 diabetes (T1D) or insulin-treated type 2 diabetes (T2D). Nevertheless, the efficacy of SMBG in non-insulin-treated T2D patients remains controversial^6,8. Several systematic reviews and meta-analyses found that SMBG could help reduce glycated hemoglobin A1c (HbA1c) and body mass index (BMI)^9–11 in non-insulin-treated T2D patients, while other studies identified that SMBG has limited benefits in these patients for diabetes management^12–15. Meanwhile, as an auxiliary tool for lifestyle adjustment, SMBG potentially offers benefits in risk factors for diabetic complications, including blood pressure and cholesterol levels¹⁶. Studies have reported that structured SMBG (S-SMBG, SMBG with defined timing and frequency of glucose measurements) is more effective than normal SMBG and is recommended by some guidelines^5,6. However, there is no consensus on the optimal application of SMBG^{6–8,12,17–19}. For example, existing SMBG frequencies are generally recommended by experience, lacking objective analysis^13,20,21. Most guidelines for diabetes management have defined no optimal regimen except the Clinical application guideline for blood glucose monitoring in China. The latter recommendations were also based on expert consensus rather than clinical evidence¹⁶.

The purpose of this work was to conduct a systematic review to assess the effects of SMBG and S-SMBG in controlling blood glucose in non-insulin-treated T2D patients, to identify the optimal monitoring frequency, and to determine the most appropriate population for SMBG through subgroup analysis.

METHODS

This systematic review and meta-analysis followed the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-analyses statement²². This study was registered in the International Prospective Register of Systematic Reviews (Registration Number: CRD42021285604).

Data Sources and Searches

Using SMBG as the main search algorithm, PubMed, Cochrane Library, Embase, and ClinicalTrials.gov were searched. The detailed search strategy is available in Table S1. Eligible publications from January 2000 to April 2022 were reviewed. A manual search was done by checking the bibliographies of eligible meta-analyses^11,23 and some included RCTs^19,24 to identify original trials that met the inclusion criteria.

Study Selection

In this paper, SMBG was referred to as self-monitoring of blood glucose at least once a week, S-SMBG was referred to as monitoring at a specific time and frequency, and no SMBG was referred to as standardized usual care (management with diet or oral hypoglycemic agents only with HbA1c levels measured by health professionals every 3 months) without regular monitoring. Studies including adjustments to patients’ lifestyles and/or doctors’ treatment plans were further examined in sub-analyses.

Randomized controlled trials (RCTs) that enrolled patients with non-insulin-treated T2D, compared SMBG with no SMBG or S-SMBG, and assessed glycemic control using HbA1c were included. Studies evaluating SMBG in patients on insulin treatment were excluded unless separate data on non-insulin-treated T2D patients were available.

Data Extraction

Two researchers (YZ and SZ) independently reviewed the titles and abstracts of eligible studies and screened full-text articles to extract the data. Any disagreement was addressed through discussion, and all disputed evaluations were adjudicated by a third researcher (GL).

The extracted information contained general information (first author’s name, publication year, and study country), study design (sample size, study duration, frequency of SMBG, and adjustments after monitoring), and demographic characteristics (age, sex, duration of diabetes, and baseline biochemical parameters). The primary outcome was HbA1c, and secondary outcomes included BMI, weight, waist circumference, total cholesterol, fasting plasma glucose, systolic blood pressure, and diastolic blood pressure. Extracted data, if possible, included outcomes at the end of the study and at 3, 6, and 12 months during the intervention. Adverse events were also extracted. The trial including two eligible intervention arms was separated into two studies or combined using formulas shown in the supplementary file.

Quality Assessment

The Cochrane collaboration tool for risk of bias²⁵ was used to assess the flaws in the following areas: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other areas. Publications with established protocols or registries were considered to have a low risk of publication bias. The quality of evidence was high, moderate, low, or very low, according to the Grading of Recommendations Assessment, Development, and Evaluation guidelines²⁶.

Statistical Analysis

Meta-analyses were performed with Review Manager version 5.4. The primary outcome was the difference in HbA1c between the intervention and control groups at the end of the study. The primary and secondary outcomes were presented as mean difference (MD) with 95% confidence interval (CI). Estimates were obtained by the random-effect meta-analysis model. Odds ratio (OR) and 95%CI were determined for adverse events, primarily hypoglycemia, by the random effects Mantel-Haenszel method. Heterogeneity was assessed by the Cochrane Q and I-square test (I²), considering values greater than 50% as indicative of substantial heterogeneity²⁷. P<0.05 indicated significant heterogeneity. Publication bias was assessed by funnel plot analysis. The outcomes were graded for quality using GRADEprofiler.

Subgroup analyses were conducted according to intervention duration (3, 6, and 12 months) and the frequency of monitoring (≤ 7 and > 7 times per week), including detailed subgroups (8–11 and ≥ 12 times per week). Other subgroup analyses were performed for baseline HbA1c (≤ 8% and >8%), duration of diabetes (≤ 6 and >6 years), adjustments based on SMBG (by physicians, by patients, and none), and diabetes education (yes and no). Data at 3, 6, and 12 months for the above factors were also used in subgroup analyses to examine study heterogeneity. For S-SMBG, a sub-analysis was also performed based on intervention time with data from 3, 6, and 12 months.

Sensitivity analyses were conducted for (a) studies with a sample size above 100, (b) studies using intention-to-treat (ITT) or per-protocol (PP) population, (c) studies after the exclusion of telemonitoring, and (d) studies after the exclusion of trials not specifying the monitoring time and frequency.

RESULTS

Study Selection and Characteristics

A total of 4483 studies were identified through electronic databases and 5 additional studies^28–32 were identified by manual search (Fig. 1). After removing duplicate studies and reading abstracts and full texts, 22 studies, including 17^{12,13,17–19,21,24,31–40} that compared SMBG with no SMBG and 4^20,41–43 that compared S-SMBG with SMBG, were selected. Four studies^19,24,35,40 used remote devices that could transmit the data from devices to clinicians based on SMBG. Of these, a trial by Parsons et al.¹⁹ had two intervention groups (SMBG and SMBG with telephone consultation) that met the criterion of being combined into a single group for quantitative analysis. Similarly, the intervention groups were combined in studies by Lee et al.³⁵ (infrequent SMBG and frequent SMBG) and Young et al.²⁴ (SMBG and SMBG with messaging). Moreover, the two intervention groups with different SMBG frequencies reported by Farmer et al.¹² were considered to be separate.

Figure 1.

Flow chart of study selection. SMBG, self-monitoring of blood glucose; S-SMBG, SMBG defined timing and frequency of glucose measurements.

Overall, 4375 patients (45.43% female, mean age of 60.15 years, and mean diabetes duration of 4.89 years) were included in SMBG analyses; 1829 patients (40.92% female, mean age of 50.07 years, and mean diabetes duration of 8.41 years) were included in S-SMBG analyses. No significant difference was found in baseline HbA1c between the SMBG and no SMBG groups and between the S-SMBG and SMBG groups (Table 1).

Table 1.

Characteristics of the Design of Included Studies

Studies	Country	Population analyses	Frequency (per week)	Adjust		Education†	Duration of intervention (months)	Number of patients	Women (%)	Mean age (years)	Mean diabetes duration (years)	Baseline HbA1c (%)		Baseline BMI (kg/m2)
				Treatment plan	Lifestyle							SMBG	No SMBG	SMBG	No SMBG
SMBG vs No SMBG
Schwedes 2002	Germany	PP	12	No	Yes	Yes	6	223	48	59.6	5.34	8.47±0.86	8.35±0.75	31.0±4.60	31.9±5.50
Guerci 2003	France	ITT	>6	No	Yes	Yes	6	689	44.85	61.55	8	9.00±1.30	8.90±1.30	30.4± 6.10	29.7±4.80
Davidson 2005	America	ITT	36	No	No	Yes	6	88	43.95	50.35	5.65	8.50±2.20	8.40± 2.10	33.4±7.00	31.7±6.70
Farmer 2007	Britain	ITT	6	Yes	No	No	12	302	42.7	65.35	3	7.41±1.02	7.49±1.09	31.9±6.20	30.9±6.10
		ITT	>6	No	Yes	Yes	12	303	43.25	65.9	3	7.53±1.12	7.49±1.09	31.0±5.30	30.9±6.10
Barnett 2008	Britain	ITT	>10	Yes	No	Yes	6	610	50	56	2.8	8.12±0.89	8.12±0.84	--	--
O’Kane 2008	Britain	ITT	8	No	Yes	Yes	12	184	41.1	59.3	0	8.80±2.10	8.60±2.30	34.0±6.98	32.0±6.23
Durán 2010	Spain	ITT	>12	Yes	Yes	Yes	12	161	53.85	64.5	0	6.60±0.52	6.60±0.89	--	--
Franciosi2011	Italy	ITT	3	Yes	Yes	Yes	6	62	21.45	48.8	3.3	7.90± 0.60	7.90± 0.60	30.2±3.90	31.4±4.60
Lu 2011	China	ITT	16	No	Yes	Yes	6	70	38.5	55.15	3.5	8.50±1.10	8.50±1.70	24.7±3.34	25.5±4.67
Harashima 2013	Japan	PP	>9	No	Yes	Yes	6	109	42.1	64.1	8.38	7.44±0.88	7.46±0.76	--	--
Kempf, K 2013	Germany	PP	>7	Yes	Yes	Yes	18	124	--	57.45	3.4	7.40±1.60	7.50±1.00	--	--
Malanda 2016	Netherlands	PP	12	No	Yes	No	12	108	30	61	6.5	7.50±0.50	7.40±0.60	--	--
Lee 2017	Korea	--	<3	No	Yes	No	6	144	38.45	55.53	--	9.30±1.40	9.20±1.50	34.8±6.46	35.5±6.00
Sodipo 2017	Nigeria	ITT	6	Yes	No	No	3	120	50	58.93	--	8.70±2.60	8.70± 2.30	--	--
Young 2017	America	ITT	7	Yes	No	No	12	450	53.3	62	6	7.58±1.04	7.52±1.12	--	--
Parsons 2019	Britain	ITT	>12	Yes	Yes	Yes	12	298	43	61.8	6.9	8.55±1.11	8.70±1.07	33.5±6.68	33.4±6.03
Lee 2020	Malaysia	ITT	6	Yes	Yes	No	12	240	55	56.2	6.65	9.00±0.17	9.00±0.17	--	--
Mean	--	--	9.83	--	--	--	9.17	249.28	45.43	60.15	4.89	8.20±0.72	8.22±0.69	31.59±2.73	31.07±2.45
S-SMBG vs SMBG
Scherbaum 2008	Germany	ITT	4 vs 1	No	Yes	Yes	12	202	38	61.41	8.06	7.20±1.40	7.20±1.00	--	--
Polonsky 2011	America	ITT	>1* vs random	Yes	Yes	No	12	483	46.8	55.9	7.6	8.90±1.20	8.90±1.20	35.0±7.80	35.1±6.70
Bosi 2013	Italy	ITT	12 vs random	Yes	Yes	Yes	12	1024	39.75	60.3	6.2	7.40±0.67	7.30±0.67	--	--
Kan 2017	China	ITT	14 vs random	Yes	Yes	Yes	6	120	39.15	58.67	11.76	9.59±1.52	9.52±1.52	--	--
Mean	--	--	--	--	--	--	9.51	268.27	40.92	50.07	8.41	7.94±1.00	7.83±1.00	35.0±7.80	35.1±6.70

SMBG, self-monitoring of blood glucose; S-SMBG, SMBG defined timing and frequency of glucose measurements; HbA1c, glycated hemoglobin; BMI, body mass index; ITT, intention-to-treat; PP, per-protocol

^*7-point SMBG profile (fasting, pre-prandial, and 2-h post-prandial at each meal, and at bedtime) on 3 consecutive days prior to each scheduled study visit (months 1, 3, 6, 9, and12). ^†Yes: the intervention or control group was educated about how to manage their diet and exercise according to blood glucose levels. -- Not reported

Primary Outcome

SMBG was associated with a significant decrease in HbA1c (MD −0.30%, 95% CI −0.42 to −0.17) in comparison with no SMBG at the end of the intervention, with high heterogeneity (I² = 81%; Fig. 2 and Table 2). S-SMBG reduced HbA1c more substantially than SMBG (MD −0.23%, 95% CI −0.38 to −0.07), with no significant heterogeneity (I² = 0%; Fig. 3 and Table 2).

Figure 2.

Forest plot of meta-analysis.

Table 2.

Outcomes of the Meta-analysis

Outcomes	No. of trials	No. of patients (SMBG)	No. of patients (no SMBG)	Mean difference (95% CI)	P value	I2 (%)	Quality of evidence assessed by GRADE
Primary outcome
HbA1c level (mmol/mol)	18	2416	2071	−0.30 [−0.42, −0.17]	<0.001	81	++-- Low
HbA1c level (mmol/mol) (S-SMBG vs SMBG)	4	919	910	−0.23 [−0.38, −0.07]	0.004	0	++-- Low
Secondary outcomes
BMI (kg/m2)	10	1049	852	−0.18 [−0.31, −0.04]	0.01	0	++-- Low
Weight (kg)	9	1466	1268	−0.33 [−0.62, −0.05]	0.02	0	++-- Low
Fasting plasma glucose (mmol/L)	5	871	807	−0.27 [−0.43, −0.12]	<0.001	0	++-- Low
Waist circumference (cm)	4	439	263	−1.12 [−2.17, −0.06]	0.04	0	++-- Low
Total cholesterol (mmol/L)	8	922	786	−0.08 [−0.18, 0.01]	0.10	26	+--- Very Low
Systolic blood pressure (mm Hg)	7	937	944	0.04 [−1.40, 1.48]	0.95	58	+--- Very Low
Diastolic blood pressure (mm Hg)	7	937	944	−0.08 [−0.66, 0.51]	0.80	0	+--- Very Low

SMBG, self-monitoring of blood glucose; S-SMBG, SMBG defined timing and frequency of glucose measurements; BMI, body mass index; GRADE, Grading of Recommendations Assessment, Development, and Evaluation

Figure 3.

Forest plot of meta-analysis for structured SMBG versus SMBG.

Secondary Outcomes

Compared with no SMBG, the reductions of BMI, weight, fasting plasma glucose, and waist circumference in the SMBG group were −0.18 kg/m² (95% CI −0.31 to −0.04), −0.33 kg (95% CI −0.62 to −0.05), −0.27 mmol/L (95% CI −0.43 to −0.12), and −1.12 cm (95% CI −2.17 to −0.06), respectively. However, SMBG showed no significant benefit in total cholesterol, systolic blood pressure, or diastolic blood pressure (Table 2). The forest plots of the secondary outcomes were provided as supplementary data (Fig. S7).

Subgroup Analyses

There was a significant difference (P = 0.03) in the effect of SMBG between the two groups categorized by monitoring frequency. HbA1c control was better among patients with SMBG >7 times per week (MD −0.39%, 95% CI −0.54 to −0.23) compared with SMBG ≤ 7 times per week (MD −0.17%, 95% CI −0.29 to −0.05). However, no significant difference (P = 0.75) in the control of HbA1c was found between SMBG at 8–11 times per week (MD −0.35%, 95% CI −0.51 to −0.20) and SMBG at ≥ 12 times per week (MD −0.40%, 95% CI −0.65 to −0.15) (Table 3).

Table 3.

Subgroups of the Meta-analysis

Subgroups		No. of trials	No. of patients		Mean difference (95% CI)	I2 (%)	Test for subgroup differences P value
			SMBG	No SMBG
Frequency (per week)	≤ 7	8	1129	1087	−0.17 [−0.29, −0.05]	81	0. 03
	>7	10	1139	933	−0.39 [−0.54, −0.23]
	8~11	4	538	438	−0.35 [−0.51, −0.20]	59	0.75
	≥ 12	6	453	495	−0.40 [−0.65, −0.15]
Duration of intervention (months)	3	7	1129	976	−0.22 [−0.32, −0.12]	80	0.46
	6	11	1418	1193	−0.35 [−0.52, −0.17]
	12	8	1225	969	−0.23 [−0.41, −0.05]
Adjusted therapy	Yes	10	1110	1059	−0.31 [−0.48, −0.14]	81	0.85
	No	8	903	870	−0.29 [−0.45, −0.12]
Adjusted therapy for 12 months	Yes	4	658	554	−0.36 [−0.62, −0.10]	81	0.01
	No	4	419	415	−0.03 [−0.08, 0.01]
Adjusted lifestyle	Yes	12	1434	1243	−0.37 [−0.50, −0.23]	81	<0.01
	No	6	982	777	−0.09 [−0.18, −0.01]
Diabetes education	No and both two groups	14	1946	1596	−0.32 [−0.48, −0.17]	81	0.50
	Invention group	5	470	424	−0.23 [−0.45, −0.02]
HbA1c (%)	≤8	8	891	728	−0.27 [−0.42, −0.12]	81	0.63
	>8	10	1525	1292	−0.33 [−0.53, −0.13]
Duration of diabetes (years)	≤6	10	1255	1068	−0.28 [−0.38, −0.18]	82	0.80
	>6	6	903	729	−0.24 [−0.50, 0.02]

Adjusting the lifestyle based on SMBG reduced HbA1c more substantially (MD −0.37%, 95% CI −0.50 to −0.23) compared with the non-adjusting group (MD −0.09%, 95% CI −0.18 to −0.01; P <0.01). No significant difference in HbA1c decline was observed when analyzing subgroups with treatment adjustment by physicians (yes and no; P = 0.85), baseline HbA1c (≤ 8% and >8%; P = 0.63), diabetes duration (≤ 6 and >6 years; P = 0.72), intervention duration (3, 6, and 12 months; P = 0.46), or diabetes education (yes and no; P = 0.50). Compared to the primary meta-analyses, other sub-analyses with data at 3, 6, or 12 months revealed no differences (Table S2). Only at 12 months did studies adjusting therapy based on SMBG show a decrease in HbA1c (MD −0.36%, 95% CI −0.62 to −0.10) as opposed to studies that did not (MD −0.03%, 95% CI −0.08 to 0.01). The advantage of S-SMBG over SMBG was not affected by intervention time (Table S3).

Adverse Events

Nine studies^{12,13,19,21,24,37,39,41,43} included in the review reported adverse events. Among them, 4 studies^12,13,21,39 reported hypoglycemic events in the SMBG and no SMBG groups, and only two studies^12,21 defined the grade of hypoglycemic events. The incidence of hypoglycemia in the SMBG group was higher compared with that in the no SMBG group (OR 2.13, 95% CI 1.42 to 3.22) (Fig. S11). Bosi et al.⁴¹ reported that non-severe hypoglycemia was more commonly detected in the S-SMBG group than in the ordinary SMBG group (1.32 compared with 0.42 per patient-year). No differences between the SMBG and no SMBG groups were observed among these 9 studies with respect to other adverse events, as no separate data were reported.

Quality of the RCTs and Meta-analyses

All trials had a high risk of performance bias, as patients themselves carried out the intervention tests. Most of the studies provided detailed information about random sequence generation and allocation concealment, while some used PP, which may cause attrition bias^18,31,36,38. A funnel plot of HbA1c showed moderate asymmetries across studies, indicating a potential publication bias. The quality of the eligible studies is shown in the supplementary figure (Fig. S1), and the strengths of evidence are shown in Table 2.

Sensitivity of the Analyses

Visual observation showed that Lee et al.’s⁴⁰ study was an outlier. However, after removing this research, no significant change in HbA1c reduction between SMBG and no SMBG was observed (MD −0.32%, 95% CI −0.43 to −0.21). Similarly, removing another 3 studies^19,24,35,40 that used remote devices (MD −0.31%, 95% CI −0.41 to −0.21) showed no difference. No significant change in HbA1c was observed after excluding studies with a sample size below 100 (MD −0.23%, 95% CI −0.36 to −0.09) or eliminating those not specifying the monitoring time and frequency (MD −0.27%, 95% CI −0.40 to −0.15). There was no difference between subgroups using ITT (MD −0.31%, 95% CI −0.45 to −0.16) or PP (MD −0.30%, 95% CI −0.58 to −0.02) alone and pooled analysis (MD −0.30%, 95% CI −0.43 to −0.18) (Table S4).

DISCUSSION

This systematic review confirmed the efficacy of SMBG in non-insulin-treated T2D patients, contributing to a 0.30% reduction in HbA1c. Moreover, the structured SMBG was found superior over SMBG, and SMBG at 8–11 times per week exhibited advantages over other frequencies of monitoring.

A significant decrease in HbA1c was observed in patients with SMBG compared with no SMBG (−0.30%, similar to the previous meta-analyses^{11,15,23,44–46}), which may have clinical significance for long-term glycemic control and reduction of complications. This was consistent with recommendations by the ADA and reports in certain meta-analyses^11,47. A UK prospective diabetes study (UKPDS) showed that every 1% reduction in HbA1C corresponds to a 37% reduction in microvascular complication risk². According to Holman et al.⁴⁸, earlier, improved metabolic management can reduce blood pressure and cholesterol levels. As part of diabetes treatment, SMBG helps maintain a healthy lifestyle and may accordingly improve blood pressure and cholesterol levels. Although there was no difference between SMBG and non-SMBG patients in changes in total cholesterol, systolic blood pressure, and diastolic blood pressure in the present work, this could not negate the influence of a healthy lifestyle on cardiovascular disease.

In this study, it was found that compared with SMBG, S-SMBG had advantages in reducing HbA1c. S-SMBG was considered to be more efficient, technological, and resource-saving than normal SMBG, considering the frequency and time of self-monitoring of blood glucose^20,41. Furthermore, although the concept of S-SMBG was proposed after 2010, most of the studies conducted before 2010 adopted a structured approach. Only two studies in this meta-analysis did not specify the monitoring time and frequency in detail^38,39. However, after the elimination of trials without specified timing and frequency, sensitive analyses showed no substantial impact on HbA1c data.

In light of the benefits of S-SMBG, we further explored the optimal frequency of monitoring blood glucose. Considering the potential harm and distress brought to patients, the optimal frequency 8–11 times per week of monitoring blood glucose was further recommended. The frequency 8–14 times weekly was reported by Xu et al.’s study⁴⁴, but more studies with SMBG frequency higher than 11 times weekly were included in our study and no additional benefit was observed. Meanwhile, some studies analyzed the monitoring time that may add to the evidence of SMBG in practical application. According to Mohan et al.⁴⁹ in contrast to the fasting and pre-prandial groups, post-prandial SMBG was associated with a significant reduction in HbA1c.

Subgroup analyses were conducted to explore the source of heterogeneity. First, we compared the differences in glycemic control with the adjustments made by physicians or patients after SMBG. SMBG is essentially about blood glucose management through the patient’s response to blood glucose readings. Patients are recommended to adjust their lifestyle, such as exercise or diet, and seek medical treatment adjustments from physicians. This study also confirmed that SMBG had a limited effect in reducing HbA1c without lifestyle adjustments. Meanwhile, we found that changes in the physician’s treatment schedule were not significant until 12 months. This may indicate that adjusting the therapy by the doctor may take longer than lifestyle adjustment by the patient to exert an effect. This result contradicted reports in some previous meta-analyses^23,45. Patients may pay less attention to the adjustment of their therapy by the medical staff, which may cause poor participation^17,39. Adjustment for SMBG was specified at the beginning of the intervention in most original trials. However, in research without specific guidelines, it is impossible to completely rule out the possibility of self-adjustment. Among the RCTs included, only a study by Davidson et al. did not specify the response to SMBG¹⁷, and their results suggested that SMBG has no advantage in reducing HbA1c.

Patient education, particularly how to handle high readings and hypoglycemia, and how to adjust diet and exercise, is essential and important in responding to SMBG. Some studies suggested that the effects shown by SMBG may be essentially the result of diabetes education¹⁴. This study found that diabetes education itself may be useful but shows no apparent influence on the outcomes of SMBG, suggesting HbA1c reduction could not be attributed to the level of diabetes education.

Patients with higher baseline HbA1c levels did not benefit more, although it is generally recognized that patients decompensated at baseline have more room for improvement. This result contrasted findings by Poolsup et al. and Machry et al.^11,46. Possible reasons may be that Machry et al.’s study included insulin-treated patients and Poolsup et al.’s meta-analysis lacked sufficient data with baseline HbA1c of ≤ 8%. However, the present study included 8 RCTs with baseline HbA1c of ≤ 8% and 10 RCTs with baseline HbA1c of > 8%. Meanwhile, we found that diabetes duration had no influence on the effect of SMBG. The differences in HbA1c reduction observed between individual trials did not suggest that differences in the characteristics (baseline HbA1c and diabetes duration) of trial participants contributed to the observed effect.

As for adverse events, this meta-analysis found that SMBG was associated with increased hypoglycemic events. However, most of these events were asymptomatic hypoglycemia. The majority of RCTs required patients with SMBG to record hypoglycemic events in their diaries, but a detailed description of the no SMBG group was absent. The increased recordings of hypoglycemia in the SMBG arm may have resulted from increased awareness of low blood glucose levels from using the meter rather than a true biochemical difference between groups¹². Besides, patients on insulin secretagogues were at high risk for hypoglycemia⁵⁰, while most studies did not analyze the baseline drug treatment.

Telemonitoring is a new tool that has been applied in some studies combined with SMBG. However, sensitivity analysis showed that the remote device had no effect on HbA1c results. This was consistent with the findings of Young et al.²⁴. In addition, a variety of new glucose monitoring technologies have emerged, including real-time continuous glucose monitoring (rt-CGM) and flash glucose monitoring (FGM), which is also termed intermittently scanned CGM^51–55. They make up for some of the shortcomings of SMBG, including pain, behavioral and technical skills, motivation, and intrusiveness^56,57. Patients utilizing CGM must have access to SMBG testing, nevertheless, as advised by the ADA and the US Food and Drug Administration (FDA) because of concerns about accuracy and other factors⁵. Several studies have confirmed the higher effectiveness of CGM compared with SMBG 4 times daily in non-insulin-treated T2D patients^58,59. However, the most cost-effective or appropriate use of CGM as a supplement to SMBG is likely when directed at T1D cases who continue to have poor control despite escalated insulin administration, insulin-appropriate T2D patients, and particular groups requiring additional monitoring of blood glucose⁶⁰. Investigators are still working on developing non-invasive blood glucose monitoring devices, and further studies are required to identify an optimal device. Nevertheless, SMBG remains the main tool for glucose monitoring, especially in developing countries. In this regard, the economic effect of the suggested SMBG frequency should be further investigated.

Strengths and Limitations of the Study

This study analyzed the advantages of SMBG and S-SMBG in non-insulin-treated T2D patients based on the latest evidence and evaluated the frequency of monitoring blood glucose in detail. We performed subgroup analyses based on intervention duration, baseline HbA1c, diabetes duration, and treatment adjustments based on SMBG. Additionally, we included the impacts of patients’ lifestyle adjustments and diabetes education in subgroup analyses for the first time. While ensuring the rigor of the analysis, we strived to explore the problem more comprehensively.

This study had limitations. Some factors not retrieved from the original trials could affect the analysis of hypoglycemic events, including methods of reporting hypoglycemic events by patients, and types of antidiabetic drugs.

CONCLUSION

SMBG is effective in controlling blood glucose levels in non-insulin-treated T2D patients. SMBG at 8–11 times weekly may be the most suitable frequency. Personal lifestyle adjustments according to SMBG were associated with HbA1c reduction, and the adjustment of treatment plans guided by SMBG may have long-term benefits.

Author Contribution

CZ contributed to the study conception and design, and revised the article for important intellectual content. YZ contributed to the search strategy, data analysis, and interpretation, and drafted the manuscript. SZ contributed to the data analysis and manuscript drafting. GL contributed to the search strategy and manuscript revision. All authors have reviewed and agreed to the submitted version of the manuscript.

Declarations

Conflict of Interest

The authors declare that they do not have a conflict of interest.