Self‐monitoring of blood glucose in patients with type 2 diabetes mellitus who are not using insulin

Previous reviews and meta‐analyses

The present review is an update of the Cochrane review performed by Welschen et al in 2005 (Welschen 2005a). In this review, substantial clinical heterogeneity between included trials was noted. Consequently, qualitative analyses were performed and it was concluded that self‐monitoring of blood glucose might be effective in improving glycaemic control in patients with type 2 diabetes who are not using insulin. The same systematic review has been published in Diabetes Care with the addition of a meta‐analysis, on request of the editor (Davidson 2005a; Kleefstra 2005; Welschen 2005b). In the meta‐analysis, the overall effect of SMBG was a statistically significant and clinically relevant decrease of 0.39% in HbA_1c in favour of SMBG compared with control groups. Since then, 11 other reviews on the effect of SMBG in patients with type 2 diabetes not using insulin have been published. Nine reviews included RCT’s only (Allemann 2009; Clar 2010; Jansen 2006; Kleefstra 2009; McIntosh 2010; Poolsup 2008; Poolsup 2009; Sarol 2005; Towfigh 2008),  one included cross‐sectional, longitudinal and (non)randomised trials (McAndrew 2007) and two combined observational studies and RCT’s (McGeoch 2007; St John 2010). In all reviews, change in HbA_1c was the primary outcome measure. Seven reviews performed a meta‐analysis (Allemann 2009; Clar 2010; Jansen 2006; McIntosh 2010; Poolsup 2008; Poolsup 2009; Sarol 2005; St John 2010; Towfigh 2008) with  HbA_1c varying from ‐0.42% to ‐0.16% in favour of SMBG versus no‐SMBG. Three reviews performed qualitative analyses only and concluded similar to Welschen et al (Welschen 2005a) that SMBG might be effective in glycaemic control but that the heterogeneity in design and quality between trials complicated drawing an overall conclusion on the effectiveness of SMBG (Kleefstra 2009; McAndrew 2007; McGeoch 2007).

Since the publication of our first review new studies have been published, possibly with good or improved methodological quality and design. We performed this update in order to explore if these new trials provide new evidence of the effect of SMBG on glycaemic control in patients not requiring insulin. In addition, if possible, an estimation of the effect of SMBG on glycaemic control in patients with type 2 diabetes not requiring insulin will be obtained. Furthermore, assessment of risk of bias of included trials illustrate limitations or enhance strengths of the studies.

Types of studies

Published and unpublished randomised controlled clinical trials (RCTs).

Types of participants

Patients diagnosed with type 2 diabetes and who are not using insulin therapy. To be consistent with changes in classification and diagnostic criteria of diabetes mellitus through the years, the diagnosis should be established using the standard criteria valid at the time of the beginning of the trial (for example ADA 1999; ADA 2008; WHO 1998). Ideally, diagnostic criteria should have been described. If necessary, we used authors' definition of (type 2) diabetes.

Types of interventions

Studies describing self‐monitoring of blood glucose (SMBG) as primary intervention compared to control are investigated. Studies concerning the comparison between SMBG and self‐monitoring of urine glucose (SMUG) were included as well (SMUG as control group).

Types of outcome measures

Electronic searches

Electronic search strategies were used to identify relevant RCT’s and reviews or meta‐analyses (for identification of additional eligible trials).

We used the following sources for the identification of trials:

• The Cochrane Library (issue 3, 2011);

• MEDLINE (until July 2011);

• EMBASE (until July 2011);

• PsycINFO (until July 2011).

We also searched databases of ongoing trials: 'Current Controlled Trials' (www.controlled‐trials.com ‐ with links to other databases of ongoing trials).

For detailed search strategies please see under Appendix 1.    

Additional key words of relevance could have been detected during any of the electronic or other searches. If this was the case, we would have modified the electronic search strategies to incorporate these terms. Studies published in any language were included.

Searching other resources

We tried to identify additional studies by searching the reference lists of included trials and (systematic) reviews, meta‐analyses and health technology assessment reports noticed.

Selection of studies

The search for publications were performed by one of the review authors (IR) supported by the Cochrane Metabolic and Endocrine Disorders' trials search coordinator. The MeSH terms and search strategy used were agreed upon and tested by two review authors (IR, UM).

Studies were selected for full text reading in three steps:

Data extraction and management

We used data concerning details of study design, intervention and outcomes employing a standardized extraction form and included the following items.

• General information (authors, title, details of journal, year of publication).

• Trial characteristics (study duration, design, methods, geographical region, temporal setting, sequence generation, allocation concealment, randomisation).

• Participants (total number per group, baseline characteristics).

• Interventions (specific details of intervention/control group).

• Outcomes (primary and secondary outcomes, timing of the outcome assessment).

• Results (number of participants analysed per group, dropouts/missing participants, summary data for each group, all available results on outcomes).

• Notes (any information reported that can be important; e.g. conflicts of interests).

A pilot test, using two trials excluded from the review preceded the data extraction of the selected RCT's. This test was likely to identify data that were not needed or missing to optimise the data extraction sheet.

Data extraction and data entry was performed independently by two review authors (UM, LW). Any discrepancies between authors were resolved by discussion. If necessary a third review author (SB) was consulted for the final decision. We sought any relevant missing information on the trial from the original author(s) of the article, if required.

Assessment of risk of bias in included studies

Two review authors (UM, LW) assessed each trial independently. Possible disagreement were resolved by consensus, or with consultation of a third party in case of disagreement. Interrater agreement for key bias indicators (e.g. allocation concealment, incomplete outcome data) was calculated using the kappa statistic (Cohen 1960). In cases of disagreement, the rest of the group was consulted and a judgement was made based on consensus. A pilot, using two trials excluded from the review, preceded the assessment of risk of bias of the RCTs.

The results and the rationale for the decision are presented in a methodological quality graph, summary and table. No trials were excluded based on the assessment of risk of bias.

The risk of bias in included studies and the internal validity of included studies was assessed with the Cochrane Collaboration's recommended 'Risk of bias' tool (Higgins 2009). With this tool, a study’s risk on selection‐, performance‐, attrition‐, detection‐, and reporting bias can be critically evaluated and judged.

We used the following criteria:

• sequence generation (was the allocation sequence adequately generated?);

• allocation concealment (was the allocation adequately concealed?);

• blinding of participants, personnel and outcome assessors (was knowledge of the allocated intervention adequately prevented during the study?);

• incomplete outcome data (were incomplete outcome data adequately addressed?);

• selective outcome reporting (were reports of the study free of suggestion of selective outcome reporting?);

• other sources of bias (was the study apparently free of other problems that could put it at a high risk of bias?

Measures of treatment effect

Dichotomous data were expressed as odds ratio (OR) or risk ratio (RR) with 95% confidence intervals (CI). Continuous were expressed as differences in means (MD) with 95% CI.

Unit of analysis issues

We took into account the level at which randomisation occurred, such as cross‐over trials, cluster‐randomised trials and multiple observations for the same outcome.

Dealing with missing data

We obtained relevant missing data from authors, if feasible and carefully performed evaluation of important numerical data such as screened, eligible and randomised patients as well as intention‐to‐treat (ITT) and per‐protocol (PP) population. We investigated attrition rates such as drop‐outs, losses to follow‐up and withdrawn study participants.

Assessment of heterogeneity

In the event of substantial clinical‐, methodological‐, or statistical heterogeneity, study results were not reported as meta‐analytically pooled effect estimates. We identified heterogeneity by visual inspection of the forest plots, by using a standard Chi² test and a significance level of α = 0.1, in view of the low power of this test. We specifically examined heterogeneity with the I² statistic quantifying inconsistency across studies to assess the impact of heterogeneity on the meta‐analysis (Higgins 2002; Higgins 2003), where an I² statistic of 75% and more indicates a considerable level of inconsistency (Higgins 2009).

When heterogeneity was found, we attempted to determine potential reasons for it by examining individual study and subgroup characteristics.

Assessment of reporting biases

We planned to use funnel plots to assess for the potential existence of small study bias. There are a number of explanations for the asymmetry of a funnel plot (Sterne 2001). Therefore, we carefully interpreted results (Lau 2006). Due to small number of included studies we did not employ funnel plots.

Data synthesis

Quantitative analyses

We summarised data statistically if they were available, sufficiently similar and of sufficient quality. We performed statistical analyses according to the statistical guidelines referenced in the newest version of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2009).

Meta‐analyses were conducted using a random‐effects model . The results of each RCT were plotted as point estimates with corresponding 95% confidence intervals. Statistical heterogeneity was tested using the Z score and the Chi² statistic with significance being set at P < 0.10. Quantification of the effect of heterogeneity was assessed by means of I², ranging from 0% to 100% including its 95% confidence interval (Higgins 2009). The I² statistic demonstrates the percentage of total variation across studies due to heterogeneity and was used to judge the consistency of evidence. If the evidence of statistical heterogeneity would be substantial (I² greater than 50%), the potential sources of variation between the RCTs would be investigated using subgroup analyses.

Qualitative analyses (best‐evidence synthesis)

When severe clinical or statistical heterogeneity was found, a qualitative analysis (best‐evidence synthesis) was performed to summarize the results of the included studies in terms of strength of the scientific evidence. Findings were considered consistent if more than one of the studies reported the same direction of the effect on the outcome measure.

Subgroup analysis and investigation of heterogeneity

If the data permitted, we planned to perform subgroup analyses to determine whether there were any systematic differences between groups of patients. We mainly carried out subgroup analyses if one of the primary outcome parameters demonstrated statistically significant differences between intervention groups. In any other case subgroup analyses were planned to be clearly marked as a hypothesis generating exercise.

A priori defined subgroup analyses were:

• HbA_1c level at baseline (subdividing into three groups of low (less than 7.0%), medium (between 7.0% and 11.0%) and high level (11.0% or higher) ‐ based on data);

• diabetes duration (up to one year past diagnosis vs duration over one year);

• duration of intervention (short‐term (up to six months follow‐up), medium‐term (between 6 and 12 months follow‐up), long term (12 months follow‐up or more)).

• age groups (below 60 years, over 60 years);

• gender;

• presence of complications (e.g. diabetic complications);

• different comparison interventions;

• type of treatment: oral hypoglycaemic agents, diet, exercise, no treatment;

• weight (normal (body mass index ‐ BMI: women less than 25, men less than 27), overweight (BMI: women 25 to 30, men 27 to 30) obese (BMI more than 30)).

Sensitivity analysis

We planned to perform sensitivity analyses in order to explore the influence of the following factors on effect size:

• repeating the analysis excluding unpublished studies;

• repeating the analysis taking account of risk of bias:

• repeating the analysis excluding very long or large studies to establish how much they dominate the results;

• repeating the analysis excluding studies using the following filters: diagnostic criteria, language of publication, source of funding (industry versus other), country.

We also planned to test the robustness of the results by repeating the analysis using different measures of effect size (relative risk, odds ratio etc.) and different statistical models (fixed‐effect model and random‐effects model).

Results of the search

The electronic database search identified a total of 1138 citations (Figure 1). After excluding titles and abstracts clearly not related to the objective of our review, 36 full text publications were retrieved for further examination. Screening of references resulted in another 15 citations. One trial (DiGEM trial 2007) had multiple publications which displayed the design of the trial (Farmer 2005), set out different outcome measures (Farmer 2007; French 2008; Simon 2008) and one overall publication (Farmer 2009). One retrieved publication of Siebolds et al (Siebolds 2006) described additional results of the trial of Schwedes et al (Schwedes 2002), a trial that was included in the original review. Results from that publication were used to supplement the trial of Schwedes et al, now referred to as SMBG study group 2002. The search identified one letter to the editor and four abstracts submitted to international conferences all describing a randomised controlled trial. Three of the abstracts had already been published and identified as an eligible trial (Davidson 2004; Drouin 2002; O'Kane 2006). Detailed trial information on the letter to the editor (Shiraiwa 2010) and the fourth abstract (Atsumi 1997) could not be retrieved and were therefore excluded (Atsumi 1997). One trial stopped following the control group after three months (Chidum 2011). We unsuccessfully requested additional data on the control group from the corresponding author. Therefore, we excluded this study as well (Chidum 2011).

Searching the database of ongoing trials identified five registered trials related to our objective. Two trials were completed and had already been published and identified as an eligible trial (DiGEM trial 2007; O'Kane 2008) and three trials were still ongoing (Bergenstal 2005; Malanda 2009; Kleefstra 2010). The authors of these trials were contacted and asked to provide (published or unpublished) data for the review. Only one author provided the requested data (Kleefstra 2010).

Finally, six eligible RCTs met all inclusion criteria and were added to the six trials included in our previous review. In total, 12 trials were included in this review.

We asked the corresponding authors of the DiGEM trial 2007, Durán 2010, Kleefstra 2010 and O'Kane 2008 to provide short‐term follow‐up data if available. They were asked if they could calculate changes in HbA_1c for short‐term follow‐up and to provide this in terms of means (SD). All authors responded and provided the additional data as requested.

Included studies

See: Characteristics of included studies and Table 2 for an overview of study populations and Appendix 2 for baseline characteristics of included studies.

Nine trials compared self monitoring of blood glucose (SMBG) with usual care without monitoring (Barnett 2008;Davidson 2005;Durán 2010;Franciosi 2011;Guerci 2003;Kleefstra 2010;Muchmore 1994;O'Kane 2008;SMBG study group 2002), one study compared SMBG with self‐monitoring of urine glucose (SMUG) (Allen 1990), one study was a three‐armed trial comparing SMBG and SMUG with usual care (Fontbonne 1989) and one study was a three‐armed trial comparing less intensive SMBG, and more intensive SMBG with a control group (DiGEM trial 2007). Three of these 12 trials had a multi‐centred design with centres in two (SMBG study group 2002), three (Franciosi 2011), and seven (Barnett 2008) countries, respectively. Trial duration ranged from 26 weeks to 12 months. The majority of the trials (seven trials) included over 100 patients in the studies (range 195 to 689).  All trials investigated effects of SMBG in patients with a diabetes duration of at least one year, except for O'Kane 2008 and Durán 2010. These two trials studied SMBG effects in newly diagnosed patients exclusively. Specifications on type, doses or combinations of prescribed oral treatments were provided by Barnett 2008,DiGEM trial 2007,Durán 2010, Franciosi 2011,Davidson 2005,Kleefstra 2010 and O'Kane 2008, however details differed per trial. Furthermore, investigated SMBG interventions differed in accompanying education programmes (Appendix 3).

Excluded studies

Studies were excluded from the review because they had a control group with access to SMBG (i.e. Polonsky 2011), they had included patients using insulin (i.e. Lim 2011), they did not explored one of our primary outcome measures (i.e. Scherbaum 2008), were secondary reports of studies all ready included (i.e. Pignone 2009) or because patients were not randomised (i.e. Bajkowska‐Fiedziukiewicz 2008).

Allocation

Five trials had adequate concealment of allocation ( Barnett 2008; DiGEM trial 2007; Franciosi 2011; Kleefstra 2010; O'Kane 2008) and five trials were unclear about allocation concealment (Allen 1990; Davidson 2005; Durán 2010; Fontbonne 1989; ;SMBG study group 2002;). In the trial of Guerci et al (Guerci 2003) patients were randomised by their general practitioners who were also responsible for intervention and follow‐up, and Muchmore et al (Muchmore 1994) recruited patients were divided over four groups, of which one group was not randomised. Therefore, Muchmore 1994 and Guerci 2003 had a high risk of bias for concealment of allocation.

Adequate sequence generation showed a low risk of bias in six trials (Allen 1990; Barnett 2008; DiGEM trial 2007; Franciosi 2011; Kleefstra 2010; O'Kane 2008), an unclear risk of bias in five trials (Davidson 2005; Durán 2010; Fontbonne 1989; Guerci 2003; SMBG study group 2002) and was not described in one (Muchmore 1994). The DiGEM trial 2007 used a computerised partial minimisation procedure for sequence generation; in the Allen 1990 et al trial sequence was computer generated. Barnett 2008 used a non‐defined random sequence, O'Kane 2008 used randomly generated codes for consecutively numbered sealed envelopes, Kleefstra 2010 had consecutively numbered non‐transparent envelopes (range 1 to 60) and Franciosi 2011 used computer‐generated randomisation tables (random permuted blocks).

Blinding

In none of the studies patients were blinded to the intervention. The care provider was only blinded in the trial of Davidson et al (Davidson 2005). In that trial the care provider was a study nurse who was kept blinded for the allocated intervention and followed detailed algorithms to make therapeutic decisions, regardless of randomisation group.

The primary outcome (HbA_1c) was assessed independently of staff responsible for performing analyses in three trials (DiGEM trial 2007; Kleefstra 2010; O'Kane 2008). Additionaly, treatment allocation was concealed for laboratory staff. Blinding of the outcome assessor was not done in Franciosi 2011 and unclear or not described in the other trials.

Incomplete outcome data

Eight trials described drop‐out rates and provided reasons for it (Allen 1990; Barnett 2008; Davidson 2005; DiGEM trial 2007; Durán 2010; Franciosi 2011; Kleefstra 2010; O'Kane 2008). Therefore, in these eight trials we assessed risk of bias concerning drop‐out rate as low. Drop‐out was not clearly described in the SMBG study group 2002 trial.

Intention‐to‐treat analysis was performed in five trials (Davidson 2005; DiGEM trial 2007; Durán 2010; Kleefstra 2010; O'Kane 2008), which we defined as having a low risk of bias. Five trials performed per‐protocol analyses only and were considered having a high risk of bias (Allen 1990; Barnett 2008; Fontbonne 1989; Guerci 2003; SMBG study group 2002) and one trial did not describe performing either intention‐to‐treat or per‐protocol analyses (Muchmore 1994).

Selective reporting

Three trials (DiGEM trial 2007; Kleefstra 2010; O'Kane 2008) had a low risk of selective reporting bias by either publishing their design (DiGEM trial 2007), by registering their protocol in a trial register (DiGEM trial 2007; Kleefstra 2010; O'Kane 2008) or both. We rated Durán 2010 and Franciosi 2011 as unclear because they both registered their protocol to a trial register three years after the start of the trial. In addition, Durán 2010 used a hard to interpret primary outcome measure. The rest of the included studies were rated as unclear as no information on pre‐designated endpoints or a‐priori defined subgroup analysis was identified.

Other potential sources of bias

All trials had similar groups at baseline for the most important prognostic indicators, except for Kleefstra 2010. In that study, diabetes duration differed between the intervention and control group. With the exception of SMBG study group 2002, co‐interventions were similar or avoided in all studies. We evaluated that the SMBG group in SMBG study group 2002 received a co‐intervention by means of a structured counselling program every four weeks during the intervention period while the control group only received non‐standardised counselling.

See: Appendix 3 for an overview of education programmes for included studies.

Heterogeneity

Included studies differed in baseline characteristics and in delivered SMBG education. Due to clinical heterogeneity we decided not to conduct a pooled analysis of all trials. We performed random‐effects subgroup meta‐analyses on the basis of diabetes duration and follow‐up.

Primary outcomes

Secondary outcomes

Clinical relevance of findings

The main results suggest that long‐term SMBG in new‐onset patients is beneficial in lowering HbA_1c. However, when diabetes duration is over one year, the overall glycaemic effect of SMBG is small and more likely to be present at short‐term. Because subgroup meta‐analyses could not fully take the presence of clinical heterogeneity into account, clinical interpretation and translation into practice of these results is difficult and should be done with caution. Different levels of probability and estimates of outcome variables of included studies might account for differences in presented subgroups. In addition, differences in requested monitoring frequency, HbA_1c level at baseline and SMBG and diabetes education may have contributed to the differences as well. Besides HbA_1c we paid attention to important outcome measures such as health‐related quality of life, well‐being, patient‐satisfaction and hypoglycaemic episodes as well. Because few included trials reported outcomes on health‐related quality of life and well‐being and used self‐report measures varied, presenting a general effect estimate is not possible and interpretation of best‐evidence synthesis is difficult. However, similar effects on similar sub‐scales or dimensions in similar directions suggested a clinical non‐relevant effect of SMBG on general (health‐related) quality of life and well‐being. In addition, since all trials that measured patient‐satisfaction did not report an SMBG related effect, this is considered clinical relevant as well. Experiencing hypoglycaemic events can be confirmed with SMBG. Therefore, it is in the line of expectation that more frequent hypoglycaemic events are reported when using SMBG.

Risk of bias

Most included trials and specifically earlier trials were exposed to selection and attrition bias . With the advent of new studies (Barnett 2008; DiGEM trial 2007; Durán 2010; Franciosi 2011; Kleefstra 2010; O'Kane 2008) performed after our first review (Welschen 2005a) the risk of bias was reduced.

Primary outcome

SMBG was embedded differently in usual care across included trials and instructions for self‐monitoring frequency, 'lifestyle' adjustment and SMBG integration in diabetes management varied between trials. However, we considered clinical heterogeneity induced by trial design not significant and decided to combine the data in subgroup meta‐analyses.

We made an a‐priori decision to separate pooled estimates of the effect of SMBG on HbA_1c for newly diagnosed patients and patients with a diabetes diagnosis of at least one year. Initiating diabetes management in newly diagnosed and never treated patients results in larger and differential effects in glycaemic control compared to patients with a longer diabetes duration (Schwedes 2002). In addition, being confronted with having a major chronic disease as type 2 diabetes can be accompanied with newly gained worries, which directly reflects in glycaemic control (Schwedes 2002).

Secondary outcomes

In none of the trials, psychological measures were the primary outcome measures. Some trials mentioned contradictory or no psychological effects as secondary outcome measures. Finally, hypoglycaemic episodes were more reported in SMBG groups than in the control groups (Barnett 2008; DiGEM trial 2007; Guerci 2003; O'Kane 2008). Because patients in the SMBG groups can use their SMBG device to confirm both periods of asymptomatic and symptomatic hypoglycaemic, this is according to expectations.

Summary

Welschen reports some conclusions and implications for the practice that do not seem to be closely and accurately based in the results of the review.

The conclusions of Welschen are: …self‐monitoring of blood glucose (SMBG) might be effective in improving glycaemic control in patients with type 2 diabetes who are not using insulin..….by using SMBG, patients can achieve a more individual management of their disease and thereby a better quality of life and this might result in a decrease in consultations with the general practitioner.

In my opinion, it is unclear on what results is Welschen based to write these conclusions. In fact, after reading the results I believe the conclusions should have been neutral or even the opposite ones. In Results Welschen wrote:

‐ The studies of Allen 1990 and Davidson 2005 [no efficacy of SMBG] were considered of high quality… The studies of Fontbonne 1989, Muchmore 1994 ……and Schwedes 2002 and Guerci 2003 [the only two trials reporting efficacy of the SMBG] …were considered to be of low quality.

‐ Heterogeneity... Because of differences in baseline data of the patients and type of interventions between the studies, it was not possible to perform either a meta‐analysis and/or subgroup or sensitivity analyses.

‐ Glycaemic control measured by HbA1c… Fontbonne 1989, Muchmore 1994 and Davidson 2005 [the only one blinded and with a high quality score] found no statistically significant differences in the decrease of HbA1c between the SMBG and the control groups… Schwedes 2002 [low quality] found a statistically significant difference of 0.5% in HbA1c in favour of SMBG…Guerci 2003 [low quality] also found a statistically significant difference of 0.4% in HbA1c at the end of the study between the SMBG and control group.

‐ We considered that the SMBG group in Schwedes 2002 did receive a co‐intervention by means of a structured counseling program every four weeks during the intervention period. The control group only received a non‐standardised counseling [see results]……. However, because this was considered as a co‐intervention, the effect of SMBG only is not clear [see Discussion].

‐ Guerci 2003 reported a dropout rate of more than 40% [48% in the intervention group], which was considered non‐acceptable.

‐ In addition, there is no evidence that SMBG has a beneficial effect on fasting plasma glucose, quality of life, well‐being, patient satisfaction and number of hypoglycaemic episodes.

In medical statistics the presumption of efficacy does not exist, but it is the opposite. It starts with the null hypothesis (there are not significant differences between an intervention [SMBG] versus other [blood glycated haemoglobin every 2‐4 months]) and clinical trials are performed to try to pull the null hypothesis down.

Welschen has failed to achieve it in her review. Her conclusions on the efficiency of the self‐monitoring are based on two studies that herself has almost rejected before for deficient quality: one because a co‐intervention in favour of the group of self‐monitoring (risk of attrition bias), and the second because a clinical trial where the drop‐outs and withdrawals rise up to 48 % of the participants loses the benefit of the randomisation, of the sample size, set doubts about the internal validity and more obvious doubts about the external validity (the applicability to the general population) of the results.

None of these inconveniences have been reported in the conclusions.

1. Coster S. Monitoring blood glucose control in diabetes mellitus: a systematic review. NHS R&D HTA Programme. Health Technology Assessment. 2000;4(12).

Reply

In our review, we tried to assess the effects of self‐monitoring of blood glucose (SMBG) in patients with type 2 diabetes who are not using insulin in order to eliminate the debate on the effectiveness of SMBG as a tool in the self‐management of these patients. 
 Sáenz comment was that our conclusions were not based on accurate results. We respectfully disagree with Sáenz and we will argue below that we still stick to the conclusion from our review that SMBG might improve glycaemic control.

The review of Coster 2000 (1) cannot be used as an adequate summary of the evidence for the effect of SMBG by health care professionals, as was suggested by Sáenz, because it is out‐dated and because of the inclusion of a trial with patients using insulin (Wing 1986). 
 Our review included all randomised controlled trials until September 2004 (2).

The direction of the effect of the combined studies is clearly in favour of SMBG and definitely not the opposite as Sáenz stated. The absence of a statistical significant effect should not be interpret as evidence of no effect (3). Moreover, since the largest studies included in our review found statistically significant differences suggests that the other smaller studies possibly did not include enough patients.

In our review we reported clearly that the evidence for an effect of SMBG on HbA1c was moderate, based on the overall methodological quality of the trials. In addition, in our methods section it was described that we did not intend to exclude trials on the basis of methodological quality criteria as suggested by Sáenz. In our discussion section, we explicitly described all methodological issues that should be taken into account before reading definitive conclusions.

We concluded that SMBG might be effective which in our opinion implies that the conclusions should be interpreted with caution. Because of the remaining uncertainty we also recommend a high quality randomised controlled trial to provide more solid evidence of the effect of SMBG on a large range of outcomes. We believe that, apart from glycaemic control, quality of life, well‐being and patient satisfaction are all very important outcome measures. Unfortunately, these outcome measures where not measured in most of the studies and therefore we could not draw conclusions on these important outcomes.

With respect to our review, we conclude that we paid sufficient attention to all the limitations in the available evidence in our review. In our view it is too early to reject a potentially helpful tool for patients with diabetes who need to incorporate their chronic disease into their daily lives.

Laura MC Welschen, MSc 
 Evelien Bloemendal, MSc 
 Giel Nijpels, MD, PhD 
 Jacqueline M. Dekker, PhD 
 Robert J. Heine, MD, PhD 
 Wim A.B. Stalman, MD, PhD 
 Lex M. Bouter, PhD

Contributors

Antonio Sáenz, Instituto Madrileno de la Salud, Spain

Summary

In the review, the authors conclude that “the overall effect of self‐monitoring of blood glucose on glycaemic control in patients with type 2 diabetes who are not using insulin is small up to six months after initiation and subsides after 12 months” [1]. For the meta‐analysis on the effects of SMBG vs. control in patients with a diabetes duration greater than one year, nine studies with a six months follow‐up (2324 patients) were included. For the corresponding 12 months follow‐up, two of the 9 studies (493 patients) were included. The HbA1c reduction of ‐0.3% (‐0.4 to ‐0.1) at six months [1] was statistically significant and in view of UKPDS, in which a 1% reduction in HbA1c was associated with a 37% decrease in risk for microvascular complications and a 21% decrease in the risk of any end point or death related to diabetes [2], a statistically significant reduction of HbA1c 0.3% at six months should be recognized and valued for the treatment of type 2 diabetes. We feel it is a limitation of the analysis that only two studies, DiGEM trial and ZODIAC‐17 [3,4], were included in the 12 month follow‐up meta‐analysis. In the analysis a non‐significant decrease in HbA1c (‐0.13% (‐0.13 to ‐0.04)) was reported. Both studies, furthermore, are characterized by several limitations, which need to be considered: In the DiGEM trial, HbA1c values of the patients in the three different groups ranged from 7.41% to 7.53%. Inclusion of patients with a stable and relatively good metabolic control at entry into the study may have attenuated the need for a modification or intensification of treatment within any of the three groups. In the study, therefore, the usage of oral antidiabetic agents (OADs) was increased only in less than one‐third of the patients. In both the less intensive and the more intensive intervention groups, OADs were not increased more frequently as compared with the control group (29% and 32% vs 30%). No specific algorithm for modification of treatment plans was mentioned. It is also noteworthy, that the enrolled patients were a highly selected population (453 patients out of 2986 total eligible ones) [5]. ZODIAC‐17 is a small Dutch study, in which only 22 patients were included in the SMBG group, of whom 17 performed at least 80% of the requested glucose measurements. The authors of the study mention in the discussion the sample size as an important limitation of the study. In the study, structured testing of blood glucose was not applied and any information on modification of treatment is missing [3]. We, therefore, would like to point out that due to the limitations of the two studies the conclusions given in the Cochrane analysis for the 12 months follow‐up are not warranted. The fact that recent prospective and randomized studies which demonstrated benefits of structured self‐monitoring of blood glucose approaches in non‐insulin treated type 2 diabetes mellitus, e.g. STeP‐study, St. Carlos study and ROSES [6‐8], report significant outcome results opposite to your meta‐analysis, further limits your conclusions.

References

1. Malanda UL, Welschen LM, Riphagen, II, et al. Self‐monitoring of blood glucose in patients with type 2 diabetes mellitus who are not using insulin. Cochrane Database Syst Rev 2012;1:CD005060. 
 2. Stratton IM, Adler AI, Neil HA, et al. Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study. BMJ 2000;321:405‐412. 
 3. Kleefstra N, Hortensius J, Logtenberg SJ, et al. Self‐monitoring of blood glucose in tablet‐treated type 2 diabetic patients (ZODIAC). Neth J Med 2010;68:311‐316. 
 4. Farmer A, Wade A, French DP, et al. The DiGEM trial protocol‐‐a randomised controlled trial to determine the effect on glycaemic control of different strategies of blood glucose self‐monitoring in people with type 2 diabetes [ISRCTN47464659]. BMC Fam Pract 2005;6:25. 
 5. Schnell O, Heinemann L. Self‐monitoring of blood glucose in noninsulin‐treated patients with type 2 diabetes: a never ending story? J Diabetes Sci Technol 2007;1:614‐616 6. Polonsky WH, Fisher L, Schikman CH, et al. Structured self‐monitoring of blood glucose significantly reduces A1C levels in poorly controlled, noninsulin‐treated type 2 diabetes: results from the Structured Testing Program study. Diabetes Care 2011;34:262‐267. 
 7. Duran A, Martin P, Runkle I, et al. Benefits of self‐monitoring blood glucose in the management of new‐onset Type 2 diabetes mellitus: the St Carlos Study, a prospective randomized clinic‐based interventional study with parallel groups. J Diabetes 2010;2:203‐211. 
 8. Franciosi M, Lucisano G, Pellegrini F, et al. ROSES: role of self‐monitoring of blood glucose and intensive education in patients with Type 2 diabetes not receiving insulin. A pilot randomized clinical trial. Diabet Med 2011;28:789‐796.

Reply

In our review¹, we explored whether newer trials with possible good or improved methodological quality and design, published after Welschen et al.² would provide new evidence of the effects of self‐monitoring of blood glucose on glycaemic control in patients with type 2 diabetes mellitus who are not using insulin. 
 We thank Schnell et al. for their comments regarding the clinical relevance of a 0.3% reduction in HbA1c after six months of self blood glucose monitoring and the limited amount of available evidence for longer‐term effects in patients with longer duration of diabetes.

In our discussion we have acknowledged that the reduction in HbA1c of 0.3% observed in the subgroup of patients with diabetes duration of at least one‐year and a six month follow‐up is statistically significant. Indeed, in the UKPDS the observed difference of 7.9% and 7.0 % between the control and the intervention groups was associated with considerably reduced morbidity and mortality. Since then, a reduction of 0.5% had generally been accepted to be of clinical relevance³ and most included and more recent trials have based their power on this^4‐7. These expected effects were not achieved, and compared to the expectations, a reduction of 0.3% is relatively small.

We agree that from a public health perspective, small reductions might have an important role ‐ however, under the condition that these are achieved at a large scale and at low cost. For SMBG so far, this is not the case^3;8.

With respect to the limited number of longer‐term studies, we like to stress that our review was performed under the stringent conditions of the Cochrane Collaboration. Therefore, an extensive assessment of risk of bias was performed following previously determined guidelines. Both the DiGEM trial and ZODIAC‐17^4;9 were assessed with low risk of bias on most domains, indicating proper internal validity.

Both trials assessed and compared SMBG effects with patients not using SMBG in an existing usual diabetes care structure and with oral glycaemic titration schedules in line with the national diabetes guidelines at the time the study was performed^9;10.

The subgroup under critique was a‐priori defined, was designed to limit clinical heterogeneity and met the widely acknowledged GRADE working group criteria¹¹. The results from the 12‐months analysis show an estimate of the effect comparable to the estimate found in the 12 months analysis ([‐0.4 to ‐0.1] vs. [‐0.3 to 0.04]) indicating a similar precision in effect. Further, a best‐evidence synthesis would have shown a similar non‐significant direction in effect.

Summarized, we stress that the 12‐months analysis is correctly performed, that included studies met the highest criteria for inclusion and thus results are valid.

We agree that patients in the specific subgroup were moderately controlled (range 7.4% to 7.7%) and therefore potentially less susceptive to benefit from SMBG. However, this was according to the predefined cut points. Furthermore, in a recent published individual patient data meta‐analysis¹² the effects of SMBG did not differ between groups with different levels of baseline entry HbA1c.

All studies that were mentioned to demonstrate benefits of structured SMBG were indeed included in our review and taken up in the appropriate subgroup analyses, except for the STeP study⁵. This study did not comply with our inclusion criteria having a control‐group with access to self‐monitoring. The ROSES‐study⁷ is included in the subgroup of patients with a diabetes duration of one year and a follow‐up of six months; the St. Carlos study¹³ is taken up in the subgroup of newly diagnosed patients with a follow‐up of six and 12 months. Conclusions resulting from these subgroup analyses are not comparable with the conclusion of the subgroup under critique and therefore do not limit our analysis.

Thus, though we acknowledge the amount of available evidence for longer‐term effect in patients with diabetes duration of over one year is limited, predefined protocols were followed leading to the inclusion of good quality studies.

At the present state of evidence, we feel our conclusions are justified. Of course, in future studies, with more innovative interventions, extended behavioural strategies or in additional subgroups, SMBG may be proven beneficial on the long‐term. Still, these future studies have to prove cost‐effectiveness. As we stated in our discussion: translation into practice of the presented results is difficult, and should be done with caution.

Reference List: 

     (1)   Malanda UL, Welschen LM, Riphagen II, Dekker JM, Nijpels G, Bot SD. Self‐monitoring of blood glucose in patients with type 2 diabetes mellitus who are not using insulin. Cochrane Database Syst Rev 2012; 1:CD005060. 
      (2)   Welschen LM, Bloemendal E, Nijpels G, Dekker JM, Heine RJ, Stalman WA et al. Self‐monitoring of blood glucose in patients with type 2 diabetes who are not using insulin. Cochrane Database Syst Rev 2005;(2):CD005060. 
      (3)   Clar C, Barnard K, Cummins E, Royle P, Waugh N. Self‐monitoring of blood glucose in type 2 diabetes: systematic review. Health Technol Assess 2010; 14(12):1‐140. 
      (4)   Farmer A, Wade A, Goyder E, Yudkin P, French D, Craven A et al. Impact of self monitoring of blood glucose in the management of patients with non‐insulin treated diabetes: open parallel group randomised trial. BMJ (Clinical research ed ) 2007; 335(7611):132. 
      (5)   Polonsky WH, Fisher L, Schikman CH, Hinnen DA, Parkin CG, Jelsovsky Z et al. Structured self‐monitoring of blood glucose significantly reduces A1C levels in poorly controlled, noninsulin‐treated type 2 diabetes: results from the Structured Testing Program study. Diabetes Care 2011; 34(2):262‐267. 
      (6)   O'Kane MJ, Bunting B, Copeland M, Coates VE. Efficacy of self monitoring of blood glucose in patients with newly diagnosed type 2 diabetes (ESMON study): randomised controlled trial. BMJ 2008. 
      (7)   Franciosi M, Lucisano G, Pellegrini F, Cantarello A, Consoli A, Cucco L et al. ROSES: role of self‐monitoring of blood glucose and intensive education in patients with Type‐2 diabetes not receiving insulin. A pilot randomized clinical trial. Diabetic Medicine 2011. 
      (8)   Davidson MB. Evaluation of Self Monitoring of Blood Glucose in Non‐Insulin‐Treated Diabetic Patients by Randomized Controlled Trials: Little Bang for the Buck. Rev Recent Clin Trials 2010. 
      (9)   Kleefstra N, Hortensius J, Logtenberg SJ, Slingerland RJ, Groenier KH, Houweling ST et al. Self‐monitoring of blood glucose in tablet‐treated type 2 diabetic patients (ZODIAC). Neth J Med 2010; 68(1):311‐316. 
    (10)   Farmer A, Wade A, French DP, Goyder E, Kinmonth AL, Neil A. The DiGEM trial protocol‐a randomised controlled trial to determine the effect on glycaemic control of different strategies of blood glucose self‐monitoring in people with type 2 diabetes [ISRCTN47464659]. BMC Fam Pract 2005; 6:25. 
    (11)   Atkins D, Eccles M, Flottorp S, Guyatt GH, Henry D, Hill S et al. Systems for grading the quality of evidence and the strength of recommendations I: critical appraisal of existing approaches The GRADE Working Group. BMC Health Serv Res 2004; 4(1):38. 
    (12)   Farmer AJ, Perera R, Ward A, Heneghan C, Oke J, Barnett AH et al. Meta‐analysis of individual patient data in randomised trials of self monitoring of blood glucose in people with non‐insulin treated type 2 diabetes. BMJ 2012; 344:e486. 
    (13)   Duran A, Martin P, Runkle I, Perez N, Abad R, Fernandez M et al. Benefits of self‐monitoring blood glucose in the management of new‐onset Type 2 diabetes mellitus: The St Carlos study, a prospective randomized clinic‐based interventional study with parallel groups. Journal of Diabetes 2 (3) ()(pp 203‐211), 2010 Date of Publication: September 2010 2010;(3):203‐211.

Contributors

Oliver Schnell 
 Email Address: oliver.schnell@lrz.uni‐muenchen.de 
 Submitter agrees with default conflict of interest statement: I certify that I have no affiliations with or involvement in any organization or entity with a financial interest in the subject matter of my feedback.

Oliver Schnell(1), Hasan Alawi(2), Tadej Battelino(3), Antonio Ceriello(4), Peter Diem(5), Anne‐Marie Felton(6), Wladyslaw Grzeszczak(7), Kari Harno(8), Peter Kempler(9), Ilhan Satman(10), Bruno Vergès(11)

Author Affiliations: 
 (1) Diabetes Research Group, Helmholtz Center, Munich, Germany 
 (2) Diabetes Centrum Saar, Saarlouis, Germany 
 (3) University Children´s Hospital, Ljubljana, Slovenia 
 (4) Insititut d'Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS) and Centro de Investigación Biomédica en Red de Diabetes y Enfermedades Metabólicas Asociadas (CIBERDEM), Barcelona, Spain 
 (5) Bern University Hospital, Bern, Switzerland 
 (6) Foundation of European Nurses in Diabetes, London, UK 
 (7) Department of Internal Medicine and Diabetology, Zabrze, Poland 
 (8) University of Eastern Finland, Finland 
 (9) Semmelweis University, Budapest, Hungary 
 (10) Istanbul university, Istanbul Faculty of Medicine, Istanbul, Turkey 
 (11) Hôpital du Bocage, Dijon, France

Uriëll Malanda on behalf of the authors.