Essay: A systematic review and meta-analysis: how SGLT2s affect eGFR in a tertiary care diabetes centre

Introduction
Diabetes mellitus is a chronic condition that affects around 415 million people around the world. Diabetes mellitus is a chronic condition which can cause a person’s blood glucose to go over the normal range and stay there.1 There are 2 common types of diabetes patients usually have and these are type 1 diabetes (type 1 diabetes mellitus) and type 2 diabetes (type 2 diabetes mellitus).2 There are also some other variations of diabetes mellitus such as gestational diabetes (high blood glucose in mother during pregnancy) and neonatal diabetes (form of diabetes that is diagnosed under the age of 6 months) which are more uncommon forms of diabetes.3
Type 2 diabetes mellitus refers to when the person’s beta-cells which are found in the pancreatic islets of Langerhans cannot produce enough insulin in the setting of the body cells not responding adequately to the insulin that has been produced. This is due to the failure of its cells to respond sufficiently to normal levels of insulin largely within the muscles, liver and fat tissues.2 The ratio of insulin resistance against beta cell dysfunction can differ from patient to patient, with some having a primarily insulin resistance defect and a minor defect in insulin secretion and others with slight insulin resistance and a primarily insulin secretion issue. This can be due to multiple aetiologies. Type 2 diabetes mellitus is mainly due to the person’s diet which can lead to obesity and it can also be due to the person’s genetics.4 You are more likely to develop type 2 diabetes if you have been leading a sedentary lifestyle for a long time and therefore become obese. The location of the body fat also makes a crucial difference as extra fat in the belly is largely linked to insulin resistance, type 2 diabetes mellitus, heart and blood vessel disease.4 Type 2 diabetes mellitus usually begins with insulin resistance where the liver, muscle and fat cells slowly cannot detect insulin and this causes a greater amount of insulin to be produced until there is a complete resistance to insulin in the liver, muscle and fat cells which leads to hyperglycaemia.4
Medication for type 1 diabetes mellitus is usually regular insulin injections for the rest of their life whereas in type 2 diabetes mellitus, treatment is usually based on the patient leading a more healthy lifestyle as it is a much more progressive condition compared to type 1 diabetes mellitus.2 There are many ways that modern society have tried to tackle the treatment and management of diabetes and one of the new areas of focus is the SGLT2 co-transporter.2
SGLT2 is the major co-transporter involved in the reabsorption of glucose in the kidney. In non-diabetic subjects all the filtered glucose is reabsorbed in the proximal tubule of the nephrons and virtually no glucose is present in the urine.5 SGLT2s are a group of drugs called gliflozins (Dapagliflozin, Empagliflozin, Sotagliflozin and Canagliflozin).1 The use of gliflozins leads to the kidney reabsorbing less glucose from the urine which means that more glucose is being excreted from the urine and that leads to a decrease in blood glucose levels.5 Additionally, the mechanism of Gliflozins are completely independent of the hormonal action of insulin and Beta-cell function.5 Therefore, SGLT2s are currently being used in the treatment of type 2 diabetes alongside the likes of Insulin and Metformin.6 SGLT2s are new therapeutic agents that tackle hyperglycaemia using a new mechanism. The normal dosage of SGLT2s (dapagliflozin) is 5mg daily in severe impairment and this is further increased according to response of the patient.7 It is also known that SGLT2 inhibition in human proximal tubular cells in nephrons exposed to high glucose is associated with a reduction in inflammatory and fibrotic markers.5
Due to the relatively new nature of this therapeutic agent, there is relatively few literature on its long-term safety as shown by reports of higher chance of getting urinary tract infections and increases in HDL and LDL cholesterol.1 Due to the diuretic nature of this drug, a common side effect patients can face is that they can become thirsty and sometimes also dehydrated due to extensive water loss through the kidneys.
However, one of the most complicated side effects of SGLT2s is its short-term and long-term effect on eGFR in the kidneys of patients who have diabetes. This is because SGLT2s show a small decline followed by a stabilisation in GFR and may also show a renoprotective effect shown by a decrease in proteinuria alongside a long-term maintenance of GFR.5 This is also the side effect that this systematic review will be focusing on.
The eGFR is short for estimated glomerular filtration rate and it is a key indicator of renal function as it assesses how well the kidney is filtering out waste products from the blood. It is a mathematically derived rate analysed from the blood sample of a patient and uses a patient’s serum creatinine level, sex, race and age to calculate the results. A variety of recognised and well researched formulae have been used for this purpose including the CKD-EPI equations. Normal GFR is usually >90 ml/min/1.73m2.
The aims:
1) To provide an evaluation of how Sodium-glucose co-transporter-2 (SGLT2) inhibitors affect eGFR in patients with type 2 diabetes.
2) To identify how different SGLT2s affect eGFR in patients with type 2 diabetes.
Methodology
This is a systematic review of literature and a meta-analysis on the effect of SGLT2s on eGFR in patients with type 2 diabetes mellitus. This was chosen to be a systematic review and meta-analysis because it will allow future healthcare professionals to access a single literature that will be a broad summary of a diverse collection of data on this topic that is currently available online. Even though there are relatively few studies available and published on this topic, when these studies are collected together, it gives a large amount of data that can be used for a systematic review and meta-analysis. This also makes a systematic review and meta-analysis the highest level of evidence.8 The method for this systematic review and meta-analysis has been developed in accordance to the PRISMA (Preferred reporting items for systematic reviews and meta-analyses) checklist found on the PRISMA-statement website which is an evidence-based minimum set of items required for reporting in systematic reviews and meta-analyses.9 Initially, websites such as diabetes.co.uk, NHS and BNF were used to learn more about the mechanism of action of SGLT2s and the different types of SGLT2s available to patients. Then, a comprehensive literature search was performed online using the databases PubMed, Cochrane and Scopus. These three databases were largely chosen since they are all very big databases and also unique to each other, so the literature search could cover all the relevant papers that are available. Cochrane was chosen as one of the databases as it is a non-profit, non-governmental organisation which will reduce any funding bias that may occur through the tendency of the study to favour the interests of the financial sponsors. It is a library with a collection of systematic reviews of healthcare intervention and diagnostic tests. PubMed is a great free resource because it is maintained by the United States national library of medicine which will help broaden the literature search. It comprises of over 29 million citations for biomedical literature from MEDLINE, life science journals, and online books usually from the field of biomedicine and health. Scopus was chosen because it is the largest abstract and citation database available online thus it would widen the literature search even further. It has scientific journals and is the only one out of the choices to have books and conference proceedings. There are many other competent databases that could have been used for the literature search such as PsycINFO, MEDLINE, Web of Science and EMBASE however their database overlapped with the chosen databases, so it would not increase the literature search by a considerable amount. After having decided on these three databases, medical subject headings were applied to filter the papers relevant to the topic. The key search terms used on all three databases were “Gliflozin” OR “Gliflozins” OR “Dapagliflozin” OR “Empagliflozin” OR “Canagliflozin” OR “Sotagliflozin” OR “Invokana” OR “Forxiga” OR “Jardiance” AND/OR “Sodium/glucose cotransporter 2” OR “SGLT2” OR “SGLT2 inhibitor” OR “SGLT2 inhibitors” OR “Sodium/glucose cotransporter 2 inhibitor” OR “Sodium-dependent glucose transporter 2 inhibitor” AND/OR “Diabetes” OR “type 2 diabetes” OR “T2DM” AND/OR “Kidney” OR “Renal” AND/OR “Treatment” OR “Inhibitors” AND/OR “EGFR” OR “GFR” on all title, abstract and key words of each database (PubMed, Cochrane and Scopus). Free-text search terms were also applied on each database to source out papers that may have been missed out from the key search terms. To ensure that only the most relevant data was collected from the databases, a few inclusion and exclusion criteria to the database search were applied that are listed below:
• Inclusion criteria
o Randomised controlled trials of SGLT2s in type 2 diabetes
o Must report eGFR during the study duration
o Population under study are aged 18 and over
o Population under study are humans
o Intervention in the study being a treatment (SGLT2s – dapagliflozin, sotogliflozin, empagliflozin, canagliflozin)
o Publication dates of up to the last 25 years even though the therapeutic agent being only available as a treatment for diabetes since 2013 as the last 25 years will include any pre-clinical trials and papers from the period before availability
o Papers from only English language as foreign language publications will need an interpreter service.
o Both genders included – as gender is a factor that affects eGFR calculation
• Exclusion criteria
o Exclude animal studies
o Exclude editorials and reviews
Using the English language as an inclusion criteria will open the study up to English language bias as studies are more likely to publish results in English if the results are positive due to most large, international journals being English language. However, this type of bias is slowly regressing as it is becoming an increasing trend to publish in English. Furthermore, there could also be publication bias due to the search not including unpublished studies. This is due to studies being less likely to be published if the results are negative or insignificant and this may create exaggerated results.
After implementing the free-text search, the key search terms, the inclusion and exclusion criteria on all the 3 databases, all the duplicates were removed by cross-checking all 3 databases. If a study had multiple publications, only the first publication was selected that reported eGFR outcome. All the references that are relevant to endnote were imported which also helped remove any duplications that may have been missed out. From the papers that were left, abstracts were screened to identify the most relevant studies according to predetermined criteria for further evaluation. After the most relevant studies were sourced out, analysis of the full text was done to assess for eligibility of data. After this, all the relevant data were collected, and certain outcomes were measured. These were:
• How SGLT2s affect eGFR in type 2 diabetes patients 1 month after first use
• How SGLT2s affect eGFR in type 2 diabetes patients at the end of the study
The time scale of the study is measuring eGFR at the end of the study and also assess eGFR of the same patients in the first month of the study. The data collected was the drug intervention and its dosage, population characteristics (age, sex, ethnicity) of the clinical trial, eGFR (mL/min/1.73m2) of the diabetes patient at the start of first treatment with SGLT2s, the eGFR of the patient 1 month after first use and the eGFR of the patient at the end of the study. For each study, the study characteristics (author, year, study design) were also collected. After having collected the data, a meta-analysis was performed.
There are different types of meta-analysis such as fixed effect and random effect model. The type used for the study was dependent on the size of the literatures and the similarity between them. A fixed effect model assumes that one true effect underlies all the studies and the differences between them are due to chance. This leads to ignoring the heterogeneity between each study and causes narrowing of the confidence intervals. Weighed means are used in this model so if one study is relatively bigger than the rest, the rest may be ignored in the overall treatment effect. In such a case, a random effect model would be better suited as in this model, the study effect is from a distribution of study effects and heterogeneity is incorporated. This makes the confidence interval much wider as well. For this reason, a random effect model was used in this meta-analysis.
Results
The papers that I have included in my systematic review and meta analyses were gathered between the dates 10th February 2019 and 14th February 2019. After applying the same key search terms, inclusion and exclusion criteria on all three databases, PubMed provided by far the largest amount of relevant papers, alongside Scopus, whereas Cochrane provided the least amount of relevant papers. PubMed yielded 1100 initial results, Scopus yielded 110 initial results, and Cochrane yielded 1034 initial results. The search results from each database was independently screened and sourced out for further evaluation. I had also found 2 reference lists of similar studies to this being undertaken which produced 12 results in total. In total this resulted in 104 papers. Then, the search results from each database were independently screened for duplicates. After duplicates were removed, the three databases yielded a total of 89 papers. The abstract of these papers was analysed for relevancy to the study with the help of the pre-determined inclusion and exclusion criteria. Majority of papers that have been excluded at this stage had either an SGLT2 inhibitor being trialled that wasn’t Dapagliflozin, Canagliflozin, Empagliflozin or Sotagliflozin such as ertugliflozin, LX4211 or Luseogliflozin, or had Type 1 diabetes mellitus instead of Type 2 diabetes mellitus. After analysis of the abstracts of the 89 papers, 38 papers had been selected to have their full-text analysed and out of these 38 papers, 21 papers were finally sourced out to be included in the systematic review and meta-analysis. The other 17 papers were excluded due to not reporting the eGFR as a primary or secondary endpoint of the study. All the resulting papers included in the systematic review and meta analyses were randomised control trials which had either a placebo or drug controls such as Glimepiride or Linagliptin to compare against the SGLT2s intervention. A total of 16 studies had solely placebo control whereas the other 5 studies included drug controls. The studies included in the systematic review and meta analyses ranged from 5 weeks to 192 weeks. 3 studies also had co-morbidies of the renal system such as moderate renal impairment or chronic kidney disease.
The process of narrowing down of the results have been displayed in its entirety below on the PRISMA flowchart.
The table below contain all the relevant information from the studies selected regarding how SGLT2s have affected eGFR 1 month after use and/or at the end of the study from the 21 papers that have been used in the systematic review and meta analyses.
Study Population age expressed as the mean Control group Control group population Intervention
(medication with dosage in mg) Intervention population Baseline eGFR (mL/min/1.73m2) expressed as the mean +/- SD eGFR (mL/min/1.73m2) 1 month after intervention expressed as the mean +/- SD
eGFR (mL/min/1.73m2) of intervention at the end of the study expressed as the mean +/- SD eGFR (mL/min/1.73m2) of control at the end of the study expressed as the mean +/- SD Study duration (weeks)
Heerspink et al (2017) 56.4 +/- 9.5 Glimepiride 6-8mg N = 482 Canagliflozin 100mg N= 483 89.7 +/- 19.3 N/A 89.8 +/- 1.1 87.1 +/- 1.2 104
Heerspink et al (2016) 54.8 +/- 8.6 Placebo N= 189 Dapagliflozin 10mg N = 167 82.1 +/- 19.7 77.3 +/- 1.8 82.8 +/- 2.0 81.2 +/- 2.0 13
Januzzi et al (2017) 64 +/-6.3 Placebo N= 216 Canagliflozin 100mg,300mg N= 450 78.2 +/-16.9 75.2 +/- 1.0 75.2 +/- 1.1 71.0 +/- 1.4 104
Yale et al (2013) 69.5 +/- 8.2 Placebo N= 90 Canagliflozin 100mg, 300mg N= 90 39.7 +/- 6.9 – 36.1 +/- 6.9 38.3 +/- 6.9 26
Dekkers et al (2018) 62 +/- 8.1 Placebo N= 15 Dapagliflozin 10mg N= 16 72.0 +/-22.0 66.9 +/- 3.0 71.0 +/- 19.0 – 6
Cefalu et al (2013) 55.8 +/- 9.2 Glimepiride N= 482 Canagliflozin 300mg N = 485 89.7 +/- 19.3 – 86.0 +/- 19.3 87.1 +/- 1.2 52
Kohan et al (2014) 66.0 +/- 8.9 Placebo N= 84 Dapagliflozin 5mg N= 83 44.2 +/- 8.8 41.82 +/- 0.84 42.49 +/-1.23 43.22 +/- 1.01 104
Kohan et al (2014) 68.0 +/- 7.7 Placebo N= 84 Dapagliflozin 10mg N= 85 43.9 +/- 10.6 39.1 +/- 0.82 40.4 +/- 1.02 43.22 +/- 1.01 104
Ptaszynska et al (2014) 55.1 +/- 10.1 Placebo N = 1393 Dapagliflozin 10mg N =1193 82.53 +/- 0.0 – 84.73 +/- 0.0 83.19 +/- 0.0 102
Kohan et al (2016) 56.5 +/- 10.1 Placebo N= 1393 Dapagliflozin 5mg N= 1145 81.9 +/- 19.5 – 84.42 +/- 0.81 84.41 +/- 0.89 102
Johnsson et al (2016) 59.3 +/- 9.7 Placebo N = 1956 Dapagliflozin 10mg N = 2026 81.0 +/- 19.1 – 81.02 +/- 0.98 79.9 +/- 1.0 102
Mudaliar et al (2015) – Placebo N = 90 Canagliflozin 100mg N = 90 39.8 +/- 0.0 – 36.2 +/- 0.0
38.6 +/- 0.0 26
Heerspink et al (2013) 53.7 +/- 9.4 Placebo N = 25 Dapagliflozin 10mg N = 24 100.6 +/- 14.3 – 100.1 +/- 4.1 102.6 +/- 3.3 12
Jobori et al (2017) 57.0 +/- 2.0 Placebo N = 15 Empagliflozin 25mg N = 15 132.0 +/- 8.0 129 +/- 10.0 129.0 +/- 10.0 126.0 +/- 10.0 2
Lavelle-Gonzalez et al (2013) 55.5 +/- 9.4 Placebo N = 139 Canagliflozin 100mg N = 296 89.7 +/- 0.0 – 88.3 +/- 12.8 86.3 +/- 18.2 52
Lavelle-Gonzalez et al (2013) 55.3 +/- 9.2 Placebo N = 139 Canagliflozin 300mg N = 295 90.2 +/- 0.0 – 88.7 +/- 12.9 86.3 +/- 18.2 52
Qui, Capuano & Meininger et al (2014) 58.6 +/- 8.9 Placebo N = 93 Canagliflozin 50mg N = 93 87.2 +/- 18.0 – 80.2 +/- 12.1 84.5 +/- 13.0 18
Sha et al (2014) 63.3 +/- 4.0 Placebo N = 18 Canagliflozin 300mg N = 17 86.1 +/- 14.0 – 83.9 +/- 1.7 95.9 +/- 1.6 12
Strojek et al (2011) 60.2 +/- 9.73 Placebo + Glimepiride N = 146 Dapagliflozin 5mg + Glimepiride N = 145 83.5 +/- 19.6 82.3 +/- 1.03 83.4 +/- 1.02 80.20 +/- 0.94 24
Tikkanen et al (2015) 60.6 +/- 8.5 Placebo N = 271 Empagliflozin 10mg N = 276 83.01 +/- 16.43 82.81 +/- 8.99 86.07 +/- 10.05 84.20 +/- 17.06 12
Tikkanen et al (2015) 60.2 +/- 9.0 Placebo N = 271 Empagliflozin 25mg N = 276 83.97 +/- 17.85 81.37 +/- 9.98 86.72 +/- 9.71 84.20 +/- 17.06 12
Wanner et al (2016) 61.7 +/- 8.5 Placebo N = 1726 Empagliflozin 10mg, 20mg N = 3473 83.1 +/- 17.1 – 82.91 +/- 0.11 81.03 +/- 0.13 192
Lewin et al (2015) 53.9 +/- 10.5 Linagliptin 5mg N = 133 Empagliflozin 10mg N = 132 88.4 +/- 19.0 – 92.1 +/- 12.4 91.0 +/- 11.5 52
The forest plot below shows the effect of SGLT2 inhibitors on eGFR (mL/min/1.73m2) of a total of 10685 patients compared to their placebo control.
Due to the high heterogeneity of the studies, a sub-group analyses had been performed dividing the studies into 3 sub-groups, according to the study duration; under 27 weeks, 27-52 weeks, 53-192 weeks. The results are shown below
Although the only subgroup that had a significantly lower heterogeneity score (42%) was the week 27-52 weeks subgroup, and subgroup of less than 26 weeks had also lowered its heterogeneity score (99%), it shows that the study durations had an impact on the heterogeneity, alongside other factors such as ethnicity, co-morbidities and the method of mean eGFR calculation.
The 4- and 6-parameter modification of diet in renal disease equation using serum creatinine was the most common method of collecting mean eGFR as a total of 8 studies used it. (Heerspink et al 2017), (Heerspink et al 2016), (Yale et al 2013), (Kohan et al 2014), (Kohan et al 2016), (Johnsson et al 2016), (Sha et al 2014), (Wanner et al 2016). Only one study used clearance of Iohexol as the method of eGFR calculation. (Jobori et al 2017).
Discussion
The aims of this study were to provide an evaluation of how SGLT2 inhibitors affect eGFR in patients with type 2 diabetes and to identify how different SGLT2s affect eGFR in patients with type 2 diabetes. The major findings of this study are that, compared to control, use of SGLT2 inhibitors initially decreases eGFR in T2DM patients (n=917) 26 weeks after use by -1.72 (-4.43, 0.99), however, it provides a renoprotective effect in patients (n=9766) long-term i.e. post 26 weeks. Overall, SGLT2 inhibitor use provides a 0.13 (-0.41, 0.67) increase in eGFR in patients with T2DM however there was no significant increase in eGFR which may have been to do with the initial decrease in eGFR with SGLT2 inhibitor use and substantial heterogeneity being detected amongst studies (I 2 = 100%, p < 0.001) making the results less reliable.
SGLT2 inhibitor use is widely known to lower cardiovascular risk, predominantly due to its natriuretic and osmotic diuretic properties that reduces plasma volume and thus blood pressure. (Briasolis et al 2018) It can cause excretion of up to 200-600mL of urine volume per day and cause 2-5mmHg drop in systolic blood pressure. (Szalat et al 2018) This is likely to be a contributing factor to the initial reduction in eGFR after SGLT2 inhibitor use found in this study due to reduced plasma volume reducing renal perfusion and eGFR. Reported increase in bone fractures have also been linked to SGLT2 inhibitor induced reduction in plasma volume leading to adverse events i.e. falls. (fillipas et al 2018) Reduction in blood pressure also contributes to increased haematocrit in the blood in patients which is mainly attributed to haemoconcentration and an increased erythropoiesis rate via increased erythropoietin levels. (fillipas et al 2018) This SGLT2-induced increase in viscosity of blood and increased haematocrit has been associated with an increased risk of stroke. (fillipas et al 2018) Venous thromboembolism is also another risk associated with increased haemoconcentration via reduction in plasma volume. This occurs due to stasis of blood flow caused by increased viscosity of blood. However, current literature suggests that there is no significant increase in risk of venous thromboembolism with SGLT2 inhibitor use in patients with T2DM. (Ueda et al 2018)
One of the biggest risks of SGLT2 inhibitor is acute kidney injury (AKI). This can occur due to the reduction of trans-glomerular hydraulic pressure and reduced eGFR that occurs with SGLT2 inhibitor induction. (Szalat et al 2018) These occur due to the reduced renal blood flow that occurs alongside the volume depletion. (Szalat et al 2018) Furthermore, diabetic patients are commonly administered RAAS inhibitors as first-choice antihypertensives via being prescribed Angiotensin converting enzyme inhibitors (ACEi) or Angiotensin receptor blockers (ARB) which can further reduce trans-glomerular hydraulic pressure and lead to AKI. (Szalat et al 2018) Also, due to the diuretic nature of SGLT2 inhibitors, patients who are already on diuretics, or patients who are elderly frail individuals and/or have congestive heart failure may be more susceptible to clinically significant pre-renal failure due to volume depletion. (Szalat et al 2018)
One common benefit of the osmotic diuretic effect of SGLT2 inhibitors was found in hypertensive patients with stage 3 CKD as canagliflozin provided a numerically greater reduction in systolic and diastolic BP compared to placebo. (Yale et al 2013). However, due to the limited amount of urinary glucose excretion that has been observed with canagliflozin use in patients with more severe renal insufficiency, SGLT2 inhibitors are not predicted to be effective for stage 4 or 5 CKD (eGFR<30ml/min/1.73m2) or dialysis patients (Yale et al 2013).
This study, however, has its limitations. There were certain types of bias noted in a few of the trails that had been included in the systematic review and meta-analyses. In (Heerspink et al 2017), lack of placebo meant that it is inconclusive if canagliflozin was renoprotective or Glimepiride worsens progression of kidney disease relative to canagliflozin. In (Heerspink et al 2016), a post hoc study was performed, and the original study was not designed for assessing dapagliflozin on renal variables thus making the results of that study less reliable. Some studies also had exclusion of patients with eGFR<50 therefore the results are less generalizable to patients with worse renal function. It was stated that this omission was due to usage of metformin in older patient population (Januzi et al 2017). However, this exclusion criteria did reduce confounding effects of one of the outcome measures of the study, the biomarker concentrations of the patient.
It is important that further research is carried out into the safety of SGLT2 inhibitors as the results are inconclusive whether there are significant risks of adverse effects associated with volume depletion caused by the initiation of SGLT2 inhibitors. This is crucial as the renoprotective effects of SGLT2 inhibitors are long-term and can help in control of blood pressure in hypertensive CKD patients, however, decrease in eGFR initially can cause fatal adverse effects such as venous thromboembolism and stroke.
Conclusion
In conclusion, I have partially achieved the aims of the study that I had set out to fulfil as the study provided an evaluation of how SGLT2 inhibitors causes a renoprotective effect in eGFR at the end of the study however it showed an initial decline in eGFR largely down to its osmotic diuretic effect leading to volume depletion. However, I did not manage to achieve the aim of evaluating the different types of SGLT2 inhibitors as there were not enough trials on each individual SGLT2 inhibitors which had the same controls to make the comparison reliable. Further studies are required in this topic as there is potential therapeutic value not only in the management of T2DM but also cardiovascular risk and CKD.