Metformin and insulin treatment of gestational diabetes: effects on inflammatory markers and IGF-binding protein-1 – secondary analysis of a randomized controlled trial

Background Gestational diabetes mellitus (GDM) is characterized by disturbed glucose metabolism and activation of low-grade inflammation. We studied whether metformin treatment has favorable or unfavorable effects on inflammatory markers and insulin-like growth factor-binding protein 1 (IGFBP-1) in GDM patients compared with insulin, and whether these markers associate with major maternal or fetal clinical outcomes. Methods This is a secondary analysis of a previous randomized controlled trial comparing metformin (n = 110) and insulin (n = 107) treatment of GDM. Fasting serum samples were collected at the time of diagnosis (baseline, mean 30 gestational weeks [gw]) and at 36 gw. Inflammatory markers serum high-sensitivity CRP (hsCRP), interleukin-6 (IL-6), matrix metalloproteinase-8 (MMP-8) and glycoprotein acetylation (GlycA) as well as three IGFBP-1 phosphoisoform concentrations were determined. Results In the metformin and insulin groups combined, hsCRP decreased (p = 0.01), whereas IL-6 (p = 0.002), GlycA (p < 0.0001) and all IGFBP-1 phosphoisoforms (p < 0.0001) increased from baseline to 36 gw. GlycA (p = 0.02) and non-phosphorylated IGFBP-1 (p = 0.008) increased more in patients treated with metformin than those treated with insulin. Inflammatory markers did not clearly associate with pregnancy outcomes but non-phosphorylated IGFBP-1 was inversely associated with gestational weight gain. Conclusions Metformin had beneficial effects on maternal serum IGFBP-1 concentrations compared to insulin, as increased IGFBP-1 related to lower total and late pregnancy maternal weight gain. GlycA increased more during metformin treatment compared to insulin. The significance of this observation needs to be more profoundly examined in further studies. There were no evident clinically relevant relations between inflammatory markers and pregnancy outcome measures. Trial registration The trial comparing metformin and insulin treatment was registered in ClinicalTrials.gov (NCT01240785) November 3, 2010. Retrospectively registered.

30 gestational weeks [gw]) and at 36 gw. Inflammatory markers serum high-sensitivity CRP (hsCRP), interleukin-6 (IL-23 6), matrix metalloproteinase-8 (MMP-8) and glycoprotein acetylation (GlycA) as well as three IGFBP-1 24 phosphoisoform concentrations were determined. 25 Results: In the metformin and insulin groups combined, hsCRP decreased (p = 0.01), whereas IL-6 (p = 0.002), GlycA 26 (p < 0.0001) and all IGFBP-1 phosphoisoforms (p < 0.0001) increased from baseline to 36 gw. GlycA (p = 0.02) and 27 non-phosphorylated IGFBP-1 (p = 0.008) increased more in patients treated with metformin than those treated with 28 insulin. Inflammatory markers did not clearly associate with pregnancy outcomes but non-phosphorylated IGFBP-1 29 was inversely associated with gestational weight gain. 30 Conclusions: Metformin had beneficial effects on maternal serum IGFBP-1 concentrations compared to insulin, as 31 increased IGFBP-1 related to lower total and late pregnancy maternal weight gain. GlycA increased more during 32 metformin treatment compared to insulin. The significance of this observation needs to be more profoundly 33 examined in further studies. There were no evident clinically relevant relations between inflammatory markers and 34 pregnancy outcome measures.
(Continued on next page) 131 and outcomes of the randomized trial have been re-132 ported elsewhere [24]. 133 For the present analysis Clinical data and serum sam-134 ples from the previous randomized trial were available 135 from 109 and 107 patients of the metformin and insulin 136 groups, respectively. Those patients in the metformin 137 group who received additional insulin are included in 138 the metformin group unless otherwise specified. 139 Biochemical methods and clinical variables 140 Fasting blood samples were drawn at baseline after the 141 GDM diagnosis had been confirmed (mean 30 [20-34] 142 gw) and at 36 gw. Serum concentrations of hsCRP and IL-143 6 were measured using ELISA [human C-reactive protein 144 (CRP) ELISA kit, R&D Systems, Minneapolis, USA; 145 interleukin-6 (IL-6) ELISA kit, R&D Systems, Minneap-146 olis, USA]. MMP-8, non-pIGFBP-1, low-pIGFBP-1, high-147 pIGFBP-1 were determined using ELISA and an immu-148 noenzymometric assay, as described earlier [15,25], and 149 GlycA according to a high-throughput proton ( 1 H) 150 nuclear magnetic resonance spectroscopy protocol [26]. 151 Glucose values of the 2 h 75 g OGTT were available at 152 the time of GDM diagnosis. C-peptide, HbA1c, age and 153 pre-pregnancy BMI were assessed as risk factors for GDM 154 and insulin resistance, to examine the relationship with 155 the risk factors, the inflammatory markers and IGFBP-1's. 156 HbA1c was determined using high pressure liquid chro-157 matography and fasting serum C-peptide by an electro-158 chemiluminescence immunoassay. Both analytes were 159 measured at baseline and HbA1c also at 36 gw.

160
Associations between inflammatory markers, IGFBP-1 161 phosphoisoforms and the following clinical outcomes 162 were studied, A) maternal outcomes: GWG, preeclampsia 163 or gestational hypertension, gestation length, induction of 164 labor, incidence of cesarean section, and B) fetal out-165 comes: birth weight, neonate admission to NICU and neo-166 natal intravenous glucose given for any indication. Total 167 GWG was defined as the last measured weight at the ma-168 ternity clinic minus self-reported weight before pregnancy, 169 and late GWG as the weight gain from the initiation of 170 antihyperglycemic medication. Birth weight was expressed 171 in grams and in SD units (deviation from the mean value 172 of the Finnish general population adjusted for gestation 173 duration [27]). Birth weight > 90th percentile was used as 174 an additional indicator of large for gestational age and a 175 birth weight below <10th percentile was used to calculate 176 the incidence of children of small for gestational age.

210
The study population characteristics are given in Table   T1 1.    Table   T2 2.
269 Gestational week 36 270 Similarly to baseline, non-pIGFBP-1 measured at 36 gw 271 was associated with lesser total and late GWG, and after 272 adjustment for pre-pregnancy BMI also low-pIGFBP-1 273 was associated with total GWG. In the metformin group, 274 MMP-8 was related to higher late GWG (0.74 kg/SD, 275 p = 0.35) and hsCRP with longer gestation (0.40 weeks/ 276 SD, p = 0.046), and these associations were unaffected by 277 adjustment for pre-pregnancy BMI. A high non-pIGFBP-1 278 concentration was related to a lower incidence for 279 cesarean section (OR: 0.49, p = 0.043), but this association 280 was no longer significant after adjustment for pre-281 pregnancy BMI. A high MMP-8 was associated with lower 282 birth weight (− 0.17 SD-units/SD, p = 0.022), and this as-283 sociation was not affected by pre-pregnancy BMI.

284
When the regression p-values at each time point were 285 adjusted using Bonferroni method, the associations be-286 tween non-pIGFBP-1 and GWG remained significant at   Conversely, hsCRP has been related to BMI [32], which 318 was higher in MiG than in our cohort; this emphasizes the t3:1 Table 3 Regression models with significant (p < 0.05) association of inflammatory markers and IGFBP-1 concentrations with maternal t3:2 and neonatal outcomes t3:3 Independent variable In line with previous reports in non-diabetic subjects, 322 IL-6 increased during the last trimester of pregnancy 323 [30]. IL-6 is secreted to a large extent by adipocytes and 324 correspondingly higher serum concentrations are associ-325 ated with higher BMI [30]. However, IL-6 has also anti-326 inflammatory effects [33] and considering the lack of as-327 sociations with any adverse outcomes in our data, the 328 complexity of IL-6 signaling in pregnancy remains in-329 completely understood. Still, we have demonstrated that 330 compared with insulin metformin treatment of GDM 331 does not appear to affect serum IL-6. 332 Previously it has been shown that, in the presence of 333 premature rupture of membranes, maternal serum IL-6 334 predicts preterm delivery at 72 h before delivery [34]. In 335 our data there was an inverse, albeit statistically non-336 significant association between IL-6 at 36 gw and gesta-337 tion length.

338
Serum GlycA increased in both treatment groups but 339 more in response to metformin treatment. This is in 340 contrast to a previous study in non-diabetic individuals 341 where metformin did not affect serum GlycA [35]. How-342 ever, the serum concentrations of some glycoproteins, 343 such as α-1-acid glycoprotein and α-1-antitrypsin, 344 change in normal pregnancy [36], and this confuses the 345 interpretation of GlycA. In general, pregnancy is associ-346 ated with activation of the innate immune system and 347 with an increase in the concentration of acute phase 348 proteins in the serum. An overall increase of GlycA dur-349 ing pregnancy has been reported previously in a popula-   duction of labor. This may reflect a better overall 427 metabolic health of patients with high serum IGFBP-1 428 while having a lower overall risk for pregnancy compli-429 cations (of which induction of labor was the most fre-430 quent). The induction rate of labor was marginally 431 higher in patients treated with insulin. This might reflect 432 the physicians' perception that GDM treated with insulin 433 is more severe than GDM without insulin treatment.

434
In our study, at baseline the inflammatory markers 435 hsCRP, IL-6 and GlycA, and IGFBP-1 phosphoisoforms 436 correlated stronger with fasting C-peptide and pre-437 pregnancy BMI than with fasting or postprandial glucose. 438 Hence, inflammatory markers and IGFBP-1 phosphoiso-439 forms seem to indicate obesity related insulin resistance. 440 We have demonstrated that metformin affects serum 441 GlycA and non-pIGFBP-1 in GDM, and that the associa-442 tions between these markers and clinical outcomes are 443 similar irrespective of the antihyperglycemic treatment 444 used. Based on this data it is unlikely that metformin, at 445 least when started this late in pregnancy, has any signifi-446 cant impact on the systemic low-grade inflammation 447 that is present in GDM [9-12] or reflects morbidity later 448 in life [13]. Follow-up studies are needed to assess the 449 long term safety of metformin treatment of GDM on 450 children. Further on, it needs to be studied whether pos-451 sible long term consequences are associated with the 452 changes in serum inflammatory markers or IGFBPs. 453 Strengths and limitations of the study 454 We have included two relatively novel inflammatory 455 markers, MMP-8 and GlycA, and provide longitudinal 456 data of their changes during the last trimester of preg-457 nancy. The study design was a randomized controlled 458 triala setting that improves the reliability of results. 459 Even so, there are some limitations to our study.

460
Our sample size was designed to prove non-inferiority 461 of metformin or insulin in birth weight in the previously 462 published primary randomized trial (24). Thus, although 463 the study population is fairly large, it was underpowered 464 to reveal or exclude all studied associations between in-465 flammation markers and IGFBP-1 s and outcome vari-466 ables. There may also be confounding factors that 467 slightly affect both maternal and neonatal outcomes, but 468 the statistical power of multiple adjusted regression 469 models to examine each outcome closely is limited. The