Filter Paper Blood Spot Enzyme Linked Immunoassay for Insulin and Application in the Evaluation of Determinants of Child Insulin Resistance

Background In large-scale epidemiology, bloodspot sampling by fingerstick onto filter paper has many advantages, including ease and low costs of collection, processing and transport. We describe the development of an enzyme-linked immunoassay (ELISA) for quantifying insulin from dried blood spots and demonstrate its application in a large trial. Methods We adapted an existing commercial kit (Mercodia Human Insulin ELISA, 10-1113-01) to quantify insulin from two 3-mm diameter discs (≈6 µL of blood) punched from whole blood standards and from trial samples. Paediatricians collected dried blood spots in a follow-up of 13,879 fasted children aged 11.5 years (interquartile range 11.3–11.8 years) from 31 trial sites across Belarus. We quantified bloodspot insulin levels and examined their distribution by demography and anthropometry. Results Mean intra-assay (n = 157) coefficients of variation were 15% and 6% for ‘low’ (6.7 mU/L) and ‘high’ (23.1 mU/L) values, respectively; the respective inter-assay values (n = 33) were 23% and 11%. The intraclass correlation coefficient between 50 paired whole bloodspot versus serum samples, collected simultaneously, was 0.90 (95% confidence interval 0.85 to 0.95). Bloodspot insulin was stable for at least 31 months at −80°C, for one week at +30°C and following four freeze-thaw cycles. Paediatricians collected a median of 8 blood spots from 13,487 (97%) children. The geometric mean insulin (log standard deviation) concentrations amongst 12,812 children were 3.0 mU/L (1.1) in boys and 4.0 mU/L (1.0) in girls and were positively associated with pubertal stage, measures of central and peripheral adiposity, height and fasting glucose. Conclusions Our simple and convenient bloodspot assay is suitable for the measurement of insulin in very small volumes of blood collected on filter paper cards and can be applied to large-scale epidemiology studies of the early-life determinants of circulating insulin.

Important challenges in large-scale epidemiology include nonacceptance of venepuncture by children and/or their parents; the costs, safety and logistics of serum or plasma separation by centrifugation; and frozen storage and transport of aliquots. An alternative procedure involves capillary puncture of the finger pulp to collect bloodspots which are then dried on filter paper. [29,30] The major advantages of dried blood spot sampling are minimal training, lower cost than venepuncture, acceptability to parents and children, [30] negligible processing requirements (cards must be air dried), low biohazard risk because samples cannot leak and the ease of storage and transport of filter paper cards. [29,31].
The use of dried blood spots is currently limited to large-scale paediatric screening programs for rare inherited disorders. There is, however, interest in developing and validating a wide range of bloodspot assays, and developments include radioimmunoassay [32] and chemiluminescence [33] methods for insulin quantification in dried blood spots. Radioimmunoassays require regulatory compliance, because of the use of radioisotopes, and both radioimmunoassay and chemiluminescence methods require specialised and costly equipment (radiation-counting apparatus or a luminometer, respectively) that may not be present in many laboratories. We are not aware of any reports using enzyme-linked immunoassay (ELISA), which offers the important advantage that results can be read on universally available microtitre plate readers, without the need for regulatory approvals to use radioisotopes or more specialised and costly measuring equipment. Development and validation of an ELISA blood spot assay offers a simple, convenient and novel alternative method of measuring insulin in large-scale epidemiology studies in a wide range of settings.
We describe the development and validation of an ELISA for quantification of insulin from dried blood spots using a simple adaptation of an existing commercial kit originally designed for use on serum or plasma (Mercodia Human Insulin ELISA, catalogue number: 10-1113-01, Mercodia AB, Sweden). The kit quantifies insulin using a solid phase, two-site enzyme immunoassay, based on the direct sandwich technique in which two monoclonal antibodies are directed against separate antigenic determinants on the insulin molecule. We show the successful practical application of our dried blood spot assay in a large-scale, multicentre trial (the Promotion of Breastfeeding Intervention Trial, PROBIT [34]) by describing the distribution and correlates of fasting insulin levels measured on over 12,000 children aged 11.5 years from 31 polyclinics distributed across the Republic of Belarus.

Materials
Blood samples were collected onto Food and Drug Administration approved Whatman 903 filter paper cards, [35] pre-printed with eight 10 mm diameter circles, standardized to absorb blood in a homogeneous manner so that uniform punches from any section of the sample yield the same quantity of blood. [36] The Human Insulin ELISA kit was from Mercodia AB, Sweden, which has no detectable cross-reactivity against c-peptide or proinsulin (both ,0.01%) and showed excellent agreement against an isotope dilution-liquid chromatography/tandem mass spectrometry (IDMS) measurement procedure calibrated using purified recombinant insulin. [37] This kit was modified by Mercodia, in collaboration with the PROBIT study investigators, to produce a Mercodia Blood Spot Insulin ELISA kit, using the same antibodies as in the Human Insulin ELISA kit (catalogue number: 10-1113-01). Instruments used included an automatic filter paper card puncher (Wallac DBS (Dried Blood Spot) Puncher, product number: 1296-071, Perkin Elmer, USA), a microplate washing device (DELFIA Washer-Diskremove, product number: 1296-0010 Perkin Elmer, USA), a plate shaker (DELFIA Plateshaker, product number:1296-003, Perkin Elmer, USA) and a microplate reader (VICTOR3 Multilabel Reader, product number: 1420-012, Perkin Elmer, USA) with a 450 nm filter.

Preparation of Bloodspot Insulin Standards and Internal Quality Control Samples
To minimize matrix differences and maximize comparability between the standards, quality control samples and the trial samples, whole blood insulin standards (diluted to 1, 3, 10, 30 and 50 mU/L whole blood) and internal quality control samples were prepared by Mercodia using recombinant human insulin in insulin ELISA calibrator buffer, article number 20-2615, mixed with washed red blood cells (in proportions 60:40 for a 0.40 haematocrit), followed by application to Whatman 903 filter paper. Standards were calibrated against the Human Insulin ELISA kit, in turn calibrated against the 1st International Reference Preparation 66/304. The detection limit (1mU/L) in dried blood spots was determined using the methodology described in ISO11843-part 4. [38].

Assay Procedure
All frozen blood spot cards (see sample collection below) were brought to room temperature before being removed from the plastic bags for assaying. Two 3-mm discs (< 6 mL of blood) were punched from the blood spots (of the standards, quality control samples and the trial samples) using an automatic puncher into 96-well microtitre plates, precoated with mouse anti-human insulin monoclonals. Two 3-mm discs were used in each assay well to ensure an adequate sample volume and elution from the bloodspots. Insulin was extracted from the discs by incubation for at least 16 hours in a refrigerator with 100 mL peroxidase conjugated mouse monoclonal anti-insulin diluted with Mercodia enzyme conjugate buffer, article number 20-2627. Following incubation, the contents of the wells were discarded, any remaining filter paper discs carefully removed and the plates washed 6 times with 350 mL of wash solution (Tris-buffered saline) to remove unbound enzyme labelled antibody. The bound conjugate was detected by a 30 minute incubation with 200 mL of 3,39,5,59-tetramethylbenzidine (TMB) added to each well before adding 50 mL of stop solution of 0.5 mol/L of H 2 SO 4 to terminate the peroxidase/TMB reaction and give a colorimetric endpoint that was read spectrophotometrically at 450 nm. The run-specific absorbance of the standards was plotted against the assigned concentration in mU/L of the whole blood filter paper standard values (omitting calibrator 0) using the log scale on both x and y axes. The concentration of the samples was read from the smoothed cubic spline fitted standard curve. Samples with insulin levels above the highest standard (50 mU/L) were reanalyzed by using only one 3-mm disc instead of two (effectively diluting 1 in 2) and the results multiplied by 2.

Inter-and Intra-assay Imprecision
Inter-and intra-assay imprecision was assessed using internal quality control samples at two levels: low (6.7 mU/L) and high (23.1 mU/L). Intra-assay imprecision for low and high concentrations of insulin was established from duplicate analysis (n = 157). The inter-assay imprecision was determined from 33 separate runs over a period of 15 months. Both intra-and interassay imprecision include between-spot extraction variations.

Stability
Stability studies were conducted at both the Mercodia laboratory and the central analytical laboratory for the PROBIT trial based in Minsk. Mercodia conducted a short-term stability study in which 16 sets of bloodspot insulin standards were stored at 280uC for up to 2 months and exposed to up to four freeze-thaw cycles. Mercodia also assessed the stability of the samples when stressed at +30uC for one hour, 24 hours, 48 hours and one week. To assess stability at 280uC over the longer term, 3 sets of whole blood samples in Minsk were spotted onto Whatman 903 cards, cut into strips containing one blood spot per strip, and frozen at 280uC. Prior to freezing, baseline (pre-storage) values were established using 30 replicates of each fresh blood spot sample. Approximately monthly, for a period of 26 months, a different strip was thawed (to ensure insulin levels were based on only one thaw) and the insulin value measured. We also stored low (6.7 mU/L) and high (23.1 mU/L) internal quality control samples, prepared on the same occasion in the same batch by Mercodia, at 280uC and repeated the assay at regular intervals, after between 7 and 31 months of frozen storage (the IQCs were stored for longer than the Minsk samples).

Comparison between Sera and Dried Blood Spots
We compared fasting insulin concentrations in serum (measured using Mercodia's Human Insulin ELISA) with insulin concentrations measured from dried blood spots, collected simultaneously from 50 children in Minsk. Whole blood samples were allowed to stand at room temperature for 2-4 hours then centrifuged at 3000 rpm for 5 minutes before separating the serum. Blood spots  Table 1. Intra-and inter-assay imprecision of dried whole bloodspot insulin assays.

Mean standard deviation of repeated measures Coefficient of variation
Intra-assay (n = 157)     and sera were stored at 280uC and brought to room temperature before assaying.

Distribution of Insulin by Demographic and Clinical Characteristics
We followed up 13,879 children aged 11.5 years (interquartile range 11.3-11.8 years) who were participants in a multicentre trial of a breastfeeding promotion intervention involving 31 polyclinics (39 paediatricians) located throughout the Republic of Belarus. [34] Children who were eligible for the trial (healthy, term newborns whose mother initiated breastfeeding) were originally recruited at birth between 1996 and 1997, [34] and have been followed up intermittently since then, including the 11.5 year follow-up reported here between 2008-2010. At 11.5 years, dried blood spots were collected from the children as described below. The children also had whole blood fasting glucose measured by glucometer (Roche ACCU-CHEK Advantage meter system, F. Hoffmann-La Roche AG, Basel, Switzerland) and the following physical measurements (amongst others not reported here) were made by study paediatricians: Tanner pubertal stage, height, weight, body mass index (BMI, weight (kg)/height (m) 2 ), skinfold thicknesses, waist circumference and leg-to-leg bioimpedence (Tanita TBF 300GS, Tanita Europe BV, Netherlands) for calculation of fat-mass. Birthweight had been collected when the children were recruited into the trial at birth in 1996/1997. [34].

Sample Collection
Children participating in the PROBIT trial follow-up were asked to fast for at least 8 hours. Up to 8 drops of fingertip capillary blood were obtained by paediatricians and applied to the pre-printed circles to form one discrete spot per drop, following specific guidelines [35,39] and training (see below). The fingerprick was made using a sterile, single-use disposable lancet device with three depth settings designed to minimize pain (Roche ACCU-CHEK Safe-T-Pro Plus lancet, Roche Diagnostics Corp., Indianapolis, USA). To avoid contact of the fresh blood filterpaper sample with potential contaminants, the sample was airdried on bespoke drying racks for 3-4 hours and the dried blood spots placed in low gas-permeable zip-closure plastic bags with desiccant packages. [36] These were stored initially in -20uC freezers at the 31 polyclinics, prior to transfer in cool boxes packed with freezer packs to 280uC freezers at the central laboratory in Minsk. The samples were stored at 220uC for a median of 1.6 months (IQR 1.0-5.0 months) and at 280uC for a median of 16.9 (IQR 12.6-20.2) months.

Training in Blood Spot Sampling
Paediatricians received training at a 1-day workshop and were provided with a written manual of procedures and an illustrative DVD. [35,39] Training included: the importance of proper technique for ensuring adequate blood flow (warm fingers, held below lap); emphasizing the need to allow blood to well on the finger-tip (rather than apply excessive pressure) to minimize haemolysis or dilution of the sample by interstitial fluid; and the need to allow large drops of blood to pool on the finger before dropping them once onto the pre-printed filter circles (with no blotting or re-spotting) to ensure adequate and evenly distributed sample volume. Training was conducted by a consultant biochemist (Dr Gusina) with extensive experience of running a national neonatal screening programme based on filter-paper blood spot collections. Feedback on the quality of spots was provided to the paediatricians at workshops every 6-months during the study fieldwork period. As blood spot cards arrived at the central laboratory in Minsk, laboratory staff assessed and recorded the quantity and quality of blood spots. Any paediatrician noted to have consistent problems (e.g. blotting, spotting, fewer than 6 good spots on average) was contacted by Dr Gusina to discuss techniques to improve the quality of the blood spot collection.

Ethics
The study received ethical approval from McGill University Health Centre Research Ethics Board; the Human Subjects Committees at Harvard Pilgrim Health Care; and the Avon Longitudinal Study of Parents and Children Law and Ethics Committee. The conduct of the study was also approved by the Ministry of Health of the Republic of Belarus. A parent or legal guardian provided written informed consent and all children provided written assent. The trial registration number for Current Controlled Trials is ISRCTN37687716.

Statistical Analysis
Calibration curves were constructed with Multicalc software (Wallac, Turku, Finland), using the log scale for both x and y axes. To assess the relationship between insulin values from serum versus blood spot samples collected simultaneously, we computed the intraclass correlation coefficient (95% confidence interval) using one-way analysis of variance and constructed a Bland-Altman plot of the difference between the serum and bloodspot insulin values (y-axis) against the mean of these two values (x-axis). Linear regression accounting for clustering by polyclinic was used to investigate relationships of fasting insulin levels with storage time in the 280uC freezer, age, location of polyclinic (urban/ rural; East/West Belarus), puberty as measured by Tanner stage, measures of general (BMI, fat mass), peripheral (triceps skinfold thickness) and central (subscapular skinfold thickness, waist circumference) adiposity, height, birthweight and fasting glucose. We calculated sex-specific associations and performed a likelihood ratio test for interactions between sex and the demographic, glucose and anthropometric factors on insulin. Insulin concentrations were log-transformed (ln(x)) for the analyses and geometric means and log standard deviations are presented. Insulin values below the lowest standard (1 mU/L, n = 1,458) were assigned the observed value (mean = 0.5 mU/L, IQR = 0.3-0.7), after determining amongst a sample of 80 children that repeating the assay for these low values made no material difference to the observed values. A sensitivity analysis excluding those 1,458 children with an insulin concentration below 1 mU/L did not alter the observed  Figure 1. The precision of the dried blood spot extraction and consecutive determination by ELISA is summarised in Table 1. Mean intra-assay (n = 157) coefficients of variation were 15% and 6% for 'low' (6.7 mU/L) and 'high' (23.1 mU/L) values, respectively; the respective interassay values (n = 33) were 23% and 11%. Figure 2 shows a comparison of insulin values for 16 dried blood spot samples stored at 280uC for 8 weeks versus insulin concentrations measured in dried blood spots prior to frozen storage (mean: 9.1 mU/L; range: 1.7 mU/L to 23.2 mU/L). These data show a very strong correlation between insulin levels measured before (fresh sample), compared to those measured after, freezing (intraclass correlation coefficient = 0.99; 95% confidence interval, CI: 0.99 to 1.00). The mean difference between values obtained from fresh versus frozen samples was very small and the 95% confidence interval was consistent with no true difference (0.2 mU/L; 95% CI: 0.0 to 0.5). A Bland-Altman plot suggested that low values of insulin were slightly increased, and high values slightly reduced, after 8 weeks frozen storage ( Figure 3). However, the differences were small and the 95% limits of agreement were within 1 mU/L. When three separate samples were stored at +30uC, the values at 1 hour, 24 hours, 48 hours and 1 week were between 89% and 127% of the values at time 0 ( Figure 4).  Figure 7 shows that insulin concentrations for 3 samples remained stable to being subjected to 4 freeze-thaw cycles.

Associations of Insulin with Demographic and Clinical Variables
Out of a total of 13,879 children who were approached, a total of 13,487 (97.2%) provided at least one and 13,403(96.6%) provided at least two acceptable dried blood spot samples (two were sufficient for the insulin assay). Amongst the 13,487 children, the paediatricians collected a median of 8 blood spots, with 84.1% (11,347) of children assessed by the central laboratory in Minsk as having had 8 acceptable quality blood spots collected. We assayed a total of 12,812 children who fasted for at least 8 hours. Table 2 shows mean (log standard deviation) insulin levels in boys and girls by the following variables: age-group; location of the polyclinic (urban/rural and East/West); Tanner stage; BMI (normal weight, overweight and obese as defined by the International Obesity Task Force, IOTF BMI cut-offs [40]); waist circumference, fat mass, triceps and subscapular thicknesses (above or below the 90th centile); and quartiles of fasting glucose, height and birthweight. In both boys and girls, mean insulin levels increased with Tanner stage by pubic hair; all measures of adiposity (BMI, waist circumference, fat mass and skinfold thicknesses); concurrent glucose levels; and height. In girls, insulin levels were also positively related to Tanner stage by breast development.

Discussion
The advantages of fingerstick blood collection onto filter paper include ease of processing, storage and transport, thus providing a valuable research tool for large epidemiology studies, particularly those that are geographically dispersed. We have shown that insulin in dried blood spots is stable and can be reliably quantified on a very large-scale with modest infrastructure, making the method useful for population-based studies of insulin resistance, diabetes and cardiovascular risk in a variety of settings.
The dried blood spot methods showed acceptable analytical performance characteristics and good agreement with the conventional serum insulin assay across the distribution of insulin values. There are several sources of variation unique to bloodspot sampling and the assay that could explain differences between bloodspot versus serum. For example, proper placement of the whole blood sample on the filter paper is critical and errors can be introduced if blood is blotted or smeared rather than drawn onto the filter paper by capillary action. [32] We minimised this source of error by rigorous training and regular feedback. Pre-printed circles on the filter paper were used to help guide the paediatricians to position the bloodspot. To maximise stability, samples were placed in zip-closure bags packed with dessicant and frozen at 220uC promptly after drying to reduce the chances of degradation, although our stability study suggested insulin was stable in dried blood spots when exposed to +30uC for one week. These sources of variation are not seen with venepuncture  sampling. The haematocrit level was set to 0.4 in the standards but not adjusted for in each sample; any sample differences would introduce errors. The exact reason why there were outliers outside the 95% limits of agreement in Figure 9 is not clear, but may be related to the quality of the blood spots or haematocrit variation. The variation seen in Figures 5 and 6 is broadly in line with the imprecision suggested in Table 1, although in Figure 6, the CV for the high IQC samples used in the stability study was slightly higher at 18%, which could reflect sources of variation outlined above. Despite unique sources of error, we achieved a good correlation between dried blood spot versus serum methods of measurement. In addition, although bloodspot assays are inherently less precise than serum assays, the present ELISA method exhibited intra-and inter-assay coefficients of variation (see Table 1) that compare favourably with those reported for radioimmunoassay (14.0% and 25.0%, respectively) [32] and chemiluminescence (11.7-29.8% and 9.8-15.5%, respectively, from Table 3 of reference [33]) methods for insulin quantification in dried blood spots. This is the first study we are aware of to report the long-term stability of insulin in dried blood spots, and our data show stability for over 2 years when frozen at 280uC. Reassuringly, we have shown that samples can be left at very high room temperatures for up to a week (+30uC) and remain stable, providing some flexibility when samples are being collected in the field (e.g. in remote sites with limited immediate access to freezers).
A range of 1-50 mU/L was selected for the assay to cover the anticipated whole blood insulin concentrations of children. Our lower limit of detection (1 mU/L) and high intraclass correlation (0.90) between serum and dried blood spot measurements, collected simultaneously, compares favourably with the chemilu-minescent immunoassay assay developed by Butter et al (lower limit of detection = 0.8 mU/L; correlation = 0.89 between paired whole bloodspots and serum samples). [33] The close agreement between dried blood spot (measured against bloodspot standards) and conventional serum (measured against liquid buffer standards) assays strongly supports the validity of dried bloodspot insulin analysis. We have also demonstrated that insulin is stable when exposed to at least four freeze-thaw cycles, indicating that samples can be reanalysed without loss of insulin.
Red blood cells contain an insulin degrading enzyme that cleaves insulin. We did not investigate the effect of adding an insulin degrading enzyme inhibitor to the eluate, because Butter et al showed that this did not improve insulin concentrations measured from whole bloodspots, [33] speculating that since the action of applying whole blood to filter paper causes red blood cells to lyse, this may release insulin degrading enzymes before assay incubation.
In our fieldwork study, we showed that taking capillary blood samples from children is feasible on a large scale, since only 3% (n = 392) of 13,879 children approached did not provide a useable sample and the paediatricians were able to take a median of 8 acceptable blood spots amongst those who were sampled. Although we documented slightly greater imprecision than would be expected in serum samples, the mean values in girls and boys that we observed, and our analysis of established associations of insulin levels with adiposity measures, puberty, glucose and height, were in line with previously published results. [41].
The advantages of filter paper capillary blood sampling, combined with the acceptable analytical performance characteristics of the dried blood spot assays, makes the approach ideal for large-scale epidemiologic applications, including studies of chil- dren, whenever the acceptability, stability, cost and safety of liquid sample collection and transportation to distant laboratories may be limiting factors. There are caveats, as previously pointed out, including the need for adequate training for those taking the bloodspot samples onto filter paper, to avoid several sources of error, and the fact that we have developed the dried bloodspot assay for research and not clinical use, which would require additional investigation. [42].