The Effects of Temperature and Relative Humidity on Point-of-Care Glucose Measurements in Hospital Practice in a Tropical Clinical Setting

Abstract

Background:

Hospitals in tropical countries experience conditions that exceed manufacturer temperature and humidity limits for point-of-care (POC) glucose reagents. Our goal was to assess the effects of out-of-limits storage temperature, operating temperature, and operating humidity on POC glucose measurement reliability.

Methods:

Quality control measurements were performed monthly using glucose test strips stored under controlled conditions and in inpatient wards under ambient conditions. Glucose test strips were evaluated in groups organized by operating temperatures of 24-25 (group 1), 28-29 (group 2), and 33-34°C (group 3), and relative humidity (RH) of ≤70 (group A), ~80 (group B), and ~90% (group C).

Results:

Glucose results for different storage conditions were inconsistent. Measurements at higher operating temperatures had lower values with mean differences of −2.4 (P < .001) and −36.5 (P < .001) mg/dL (28-29 vs 24-25°C), and −3.6 (P < .001) and −37.4 (P < .001) mg/dL (33-34 vs 24-25°C) for low and high control levels, respectively. Measurements at higher RH had lower values with mean differences of −4.0 (P < .001) and −13.2 (P < .001) mg/dL (~80 vs ≤70% RH), and −5.8 (P < .001) and −16.6 (P < .001) mg/dL (~90 vs ≤70% RH) for low and high levels, respectively.

Conclusions:

High temperature and high RH decreased glucose concentrations for the POC oxidase-based system we evaluated. We recommend that individual hospitals perform stress testing, then determine if maximum absolute differences, which represent highest risk for patients, are clinically significant for decision making by using error grid analysis.

Inaccuracy of glucose monitoring may derive from test strip, patient, pharmacological, environmental, and other factors.¹ Incorrect results may cause serious harm and change clinical decisions.^2,3 To ensure reliable results, POC devices and test strips should be stored and operated according to manufacturer specifications. In tropical settings, temperature and relative humidity may exceed specified ranges.

Our goal was to assess the effects of storage temperature, operating temperature, and operating relative humidity (RH) when ambient conditions in patient ward areas naturally exceeded manufacturer limits because of the tropical setting of Siriraj Hospital. During ISO 22870^4,5 certification inspection, we discovered out-of-limit temperature and RH discrepancies that had to be investigated and corrected.

Methods

Glucose Meters and Test Strips

We evaluated the photometric SureStep®Flexx meter, SureStep™ Hospital glucose test strips, and control solutions (LifeScan, Milpitas, CA), which use glucose oxidase enzyme specific for D-glucose to report a plasma glucose-calibrated result. Glucose in blood interacts with glucose oxidase to produce gluconic acid and hydrogen peroxide. Peroxidase on the test strip causes the hydrogen peroxide to react with dyes and produce a blue color, the intensity of which is correlated with the concentration of plasma glucose. Test strips expired 4 months after opening. According to manufacturer specifications, test strips and control solutions should be stored outside a refrigerator in a cool and dry location below 30°C (86°F) and used at an operating temperature of 18-30°C (64.4-86°F) and RH of 30-70%.

The study was performed at Siriraj Hospital, a tertiary care teaching and public hospital in Bangkok during November 2007-March 2008, February-March 2012, and June-July 2011, for storage temperature, operating temperature, and operating humidity experiments, respectively. Siriraj Hospital, the largest university hospital of Thailand, has 2200 beds, 15 critical care units, 1800 physicians including residents, and 3000 registered nurses.

Approximately 500 000 POC glucose tests are performed per year in 147 sites. Personnel performing POC glucose tests comprise 2200 nurses, 230 residents, 660 medical students, 15 clinical pathologists, and 60 laboratory technicians, all trained to competency levels specified in ISO 22870.

Storage Temperature Experiment

Experiments were performed in actual ambient conditions over 2 consecutive 4-month periods, from November to February, considered winter in Thailand, and during summer and the early rainy season, March to June. In each period, test strips and low and high control solutions with the same lot number were used.

To assess the effects of ambient temperature on test stability, half of the strips were stored in the medical storage unit (controlled condition) of Siriraj Hospital, while the other half were kept in different non-air-conditioned inpatient wards: 5 wards (wards A-E) during November to February, and 6 wards (wards A-F) during March to June. The temperature of the storage room was controlled at 25°C (77°F).

Ambient temperatures in the medical storage unit and inpatient wards were measured every hour with 24-hour continuous temperature monitors (ESCORT Intelligent MINI data loggers, Cryopak, Edison, NJ). Because control solutions can be used for 3 months after opening, new bottles of control solutions (the same lot) were used at the end of the third month in each study period.

We performed glucose measurements using quality control solutions stored in the storage unit and inpatient wards by the same operator monthly to avoid introducing operator bias. The same lots of 2 levels of control solutions, and a single glucose meter were used for all sites. Glucose concentrations were measured by using 1 test strip for each control level from each vial stored in each storage site. Glucose concentrations obtained from unopened strip and control solution bottles at each site at the beginning of each study period were used as baseline.

Operating Temperature Experiment

Two levels of control solutions were used to obtain glucose measurements at 3 ambient temperature ranges: 24-25 (group 1), 28-29 (group 2), and 33-34°C (group 3) (75.2-77, 82.4-84.2, and 91.4-93.2°F), 1 day per temperature range (n = 20 per group). The same glucose meter and the same lot of QC reagents stored under controlled condition were used in this experiment. The first 2 temperature ranges were observed in air-conditioned settings and were within manufacturer specifications, while group 3 was in a setting that lacked air-conditioning and was out-of-limit. All test results were obtained by a single operator.

Operating Relative Humidity Experiment

Glucose concentrations in 2 levels of control solution reagents were measured at ambient RH ≤70% (control group or group A, n = 20). In the experimental groups, out-of-limit glucose concentrations were measured at ambient RH of approximately 80 (group B, n = 18) and 90% (group C, n = 20). The experiment was performed in 1 day per each RH range with 1 glucose meter and the same test strip lots stored under controlled condition by a single operator.

Statistical Analysis

Data for comparisons were normally distributed without outliers. Statistical analysis was performed using 1-way ANOVA Dunnett’s test for comparisons of glucose results in storage temperatures among different months and baseline, and different operating temperatures and operating humidity. Unpaired t-test for the means was used to compare mean glucose concentrations between the medical storage unit and inpatient wards each month. A P value < .05 was considered statistically significant. We used PASW version 18.0 for Windows (SPSS Inc, Chicago, IL).

Results

Storage Temperature Experiment

During November to January, no temperatures exceeded 30°C in the medical storage unit. Temperatures were elevated in up to 29.0% (481/1658 recordings) of the inpatient wards (Table 1). Wards had 1657-1658 records available for analysis during November to January except ward A, which had only 168 records due to the technical omissions. The temperature in the medical storage unit was monitored and verified to be within acceptable range during November to January, so we did not record it during March to June. Temperatures exceeding 30°C (86°F) were 0.8-71.6% of inpatient wards during this period (Table 1).

Table 1.

Storage Temperatures in the Medical Storage Unit and Inpatient Wards During 2 Periods: November to January (Winter) and March to June (Summer and Early Rainy Season).

	Medical storage unit	Ward A	Ward B	Ward C	Ward D	Ward E	Ward F
Nov-Jan
N	1658	168	1658	1658	1657	1658	NA
Mean ± SD (°C)	22.1 ± 0.8	27.9 ± 1.1	29.4 ± 1.4	28.6 ± 1.7	29.1 ± 1.6	27.2 ± 1.2	NA
Mean ± SD (°F)	71.8 ± 1.5	82.2 ± 1.9	85.0 ± 2.5	83.5 ± 3.0	84.3 ± 2.9	80.9 ± 2.1	NA
Min-Max (°C)	20-28.5	26-29.5	25-32.5	24-32	24-32.5	24-31	NA
Min-Max (°F)	68-83.3	78.8-85.1	77-90.5	75.2-89.6	75.2-90.5	75.2-87.8	NA
% T>30°C (86°F)	0	0	29.0	17.1	22.6	0.2	NA
Mar-Jun
N	NA	2771	2771	2771	2771	2771	2771
Mean ± SD (°C)	NA	28.5 ± 1.3	30.5 ± 1.0	30.6 ± 0.9	30.1 ± 1.3	27.3 ± 1.0	30.2 ± 0.9
Mean ± SD (°F)	NA	83.3 ± 2.4	86.9 ± 1.8	87.1 ± 1.6	86.1 ± 2.3	81.2 ± 1.8	86.3 ± 1.6
Min-Max (°C)	NA	24-32	24.5-34	24-32.5	24.5-33.5	24.5-31.5	25-33
Min-Max (°F)	NA	75.2-89.6	76.1-93.2	75.2-90.5	76.1-92.3	76.1-88.7	77-91.4
% T>30°C (86°F)	NA	11.3	58.7	71.6	46.6	0.8	46.4

N, number of data points for temperature record, which was done every hour; NA, not available; T, temperature.

Figure 1 shows glucose measurement results using low and high level control solutions for strips stored in the medical storage unit and inpatient wards. Glucose concentrations differed significantly in November, December, and February (P = .019, .026, <.001 for low level; P = .027, .025, .019 for high level). Only the low level was significantly different at baseline, in March, April, and May (P = .039, .043, .024, and .027, respectively).

Figure 1.

Comparison of glucose measurements from storage temperature experiments between 2 periods in the medical storage unit and inpatient wards. Results were obtained with glucose test strips stored in the medical storage unit (black solid lines) for 4 months and in inpatient wards (gray solid lines) for 4 months during 2 periods: November to February (A, C) and March to June (B, D). Frames A and B show glucose measurements obtained with the low control reagent, C and D with the high control reagent. Means are shown with 95% confidence interval (CI) cross bars. The dashed lines show the span of the minimum and maximum glucose concentrations. Note that in some cases, the range was quite high. In the first period, the number of replicates was 10, 10; 10, 8; 8, 8; 8, 8; and 8, 8 for baseline, November, December, and February, respectively. In the second period, the number of replicates was 12 for baseline and all 4 months (March to June). *P < .05. ***P < .001.

By ANOVA and Dunnett’s multiple comparison testing, glucose concentrations tested with strips stored in the medical storage unit from the following months were not different from baseline except for low level control solutions between March and baseline (P = .012), as well as April and baseline (P = .002). For glucose strips stored in inpatient wards, glucose concentrations were different in the low level control between February and baseline (P < .001). The range of glucose concentration in low level control tested by strips stored in inpatient wards was highest in April (10 mg/dL), the hottest month of the year. In addition, the range in high level control of strips in inpatient wards was highest during March and April (55 mg/dL) (Figure 1).

Operating Temperature Experiment

Table 2A shows actual temperature and RH. Operating temperature was divided into 3 groups: 24-25 (group 1), 28-29 (group 2), and 33-34°C (group 3); RH was within acceptable range for all temperature groupings. When the temperature was low, RH also was low. Mean differences in low control and high control solutions between group 2 and group 1 were −2.4 (95% CI −1.0, −3.8, P < .001) and −6.5 (95% CI −26.7, −46.3, P < .001) mg/dL, respectively. Mean difference between group 3 and group 1 were −3.6 mg/dL (95% CI −2.1, −5.1, P < .001) for low level and −37.4 mg/dL (95% CI −27.6, −47.1, P < .001) for high level control solution (Figure 2).

Table 2.

Actual Temperature and Relative Humidity in Different Groups of Operating Temperature (A) and Operating Humidity (B).

A.
Operating temperature experiment	Group 1, 24-25°C (75.2-77°F)	Group 2, 28-29°C (82.4-84.2°F)	Group 3, 33-34°C (91.4-93.2°F)
Actual temperature
Range (°C)	23.6-26.4	28.2-29	32.9-34.2
Range (°F)	74.5-79.5	82.8-84.2	91.2-93.6
Actual humidity (%)
Range	40-55	57-64	62-69
B.
Operating humidity experiment	Group A, ≤70%	Group B, ~80%	Group C, ~90%
Actual temperature
Range (°C)	26.4-27.9	27.8-28.5	27.3-28.5
Range (°F)	79.5-82.2	82.0-83.3	81.1-83.3
Actual relative humidity (%)
Range	49-70	79-83	89-90

Figure 2.

Glucose concentrations (mean, 95% CI) observed with low level (A) and high level (B) control solutions. Data were obtained at the operating temperatures of 24-25°C (75.2-77°F), 28-29°C (82.4-84.2°F), and 33-34°C (91.4-93.2°F); n = 20 for each group. The dashed lines represent the minimum and maximum glucose concentrations. Operating temperature specified by the manufacturer is 18-30°C (64.4-86°F). *P < .05. ***P < .001.

Operating Humidity Experiment

Table 2B shows the actual temperature and RH at RH ≤70 (group A), ~80 (group B), and ~90% (group C). The operating temperature was within acceptable limits for all RH groupings. Mean difference between groups B and A was −4.0 mg/dL (95% CI −2.8, −5.1, P < .001), and between groups C and A, −5.8 mg/dL (95% CI −4.7, −6.9, P < .001) for low level control solution. For the high level control solution, mean difference between groups B and A was −13.2 mg/dL (95% CI −5.9, −20.5, P < .001), and between groups C and A, −16.6 mg/dL (95% CI −9.5, −23.7, P < .001) (Figure 3).

Figure 3.

Glucose concentrations (mean, 95% CI) measured using low level (A) and high level (B) control solutions at ≤70% (n = 20), ~80% (n = 18), and ~90% (n = 20) relative humidity. The dashed lines represent the minimum and maximum glucose concentrations. Operating humidity specified by manufacturer is 30-70%. ***P < .001.

Clinical Impact

We assessed the clinical impact of the glucose measurement by using the surveillance error grid.⁶ We used the maximum value of glucose concentration in each experimental group as the reference blood glucose (BG) and the minimum value of glucose concentration in each group as the measured BG in the software by Kovatchev et al.⁷ We found that paired comparisons of maximum and minimum values of glucose concentrations in each experiment were within zone A, which was the allowable total error region usually associated with no harm.⁶

Discussion

The most impactful results in this research were higher operating temperature resulted in lower glucose concentrations, and higher operating RH also resulted in lower glucose concentrations. Not all inpatient wards maintained storage temperature below 30°C. Glucose concentration differences between the medical storage unit and inpatient wards were inconsistent with respect to time and levels of the control solution when assessed with respect to baseline. Storage temperature, operating temperature, or RH might be the cause of these differences.

When the operating temperature increased, glucose concentration decreased, as shown in Figure 2. Glucose concentrations measured at 28-29 and 33-34°C were statistically significantly lower than those measured at 24-25°C. However, glucose concentration measured at 28-29°C, as specified by the manufacturer, was not different from that measured at 33-34°C using the high level control. For the low level control, glucose concentration measured at 28-29°C was different statistically from that measured at 33-34°C, but the mean difference of 1.2 mg/dL was not clinically significant.

Increased RH generated lower glucose values as shown in Figure 3. Previous studies have demonstrated that high temperature and high RH affected glucose values from different manufacturers differently. For example, high operating temperature results in overestimates by some glucose meters,^8,9 while others underestimate.¹⁰ Lam et al showed that short-term exposure (15 minutes) of glucose test strips and meters to high temperature and high humidity resulted in elevated glucose results as high as 33 mg/dL, and that the combined effects of stressing both meters and test strips at the same time were synergistic.¹¹

Degradation of enzyme may explain the changes in test performance. With simulated disaster climates, such as Hurricane Katrina (set point range 20-45°C [68-113°F]; humidity, 31-96%), glucose values were lower with glucose-oxidase based test strips results, and higher with glucose-dehydrogenase based test strips.¹² Studies have examined the instability of glucose oxidase enzyme.^13-15 High temperature dissociates flavin adenine dinucleotide cofactors and destroys the secondary and tertiary structure of glucose oxidase.

Cembrowski et al performed BG measurements using SureStep®Flexx and found that the glucose concentrations for the quality control solutions and patient samples were consistently higher in the winter months. This finding is probably due to the very low indoor humidity associated with external subzero temperatures. Low humidity will be accompanied by rapid evaporation of blood samples. The average weekly temperatures in that study varied from −25°C in the winter to +20°C in the summer.¹⁶ Our study was performed in a clinical tropical setting where we found that higher operating temperature and RH decreased glucose values. In general, environmental factors can affect several POC tests and quality control reagents as well.^16-20

To our knowledge, this is the first article to use error grids⁶ for the assessment of environmental effects on bedside glucose meter performance in a tropical country. Using the maximum value in each group as the reference BG and the minimum value in each group as the measured BG in the software by Kovatchev et al,⁷ found that paired data of maximum and minimum values of glucose concentrations in all 3 experiments were within zone A, an allowable total error region usually causing no harm.⁶ Siriraj Hospital POC Committee,²¹ composed of physicians from several departments, agreed that while the effects of temperature and RH thus far did not appear to have affected medical decision making, maximally discrepant results had potential to do so. Hence, precautionary measures will be implemented to preserve the quality of bedside testing on clinical wards in the future.

Siriraj Hospital critical care areas are air-conditioned. We confirmed that temperature and RH are maintained within manufacturer specifications, so there was no need to study environmental stress effects in critical care units. One limitation of our study was that higher temperatures were accompanied by higher RH. Therefore, effects of temperature on test strip enzyme may be confounded by simultaneous alteration of RH. The use of QC reagents was a logical first step, since each glucose meter must perform within manufacturer specifications, but at the same time, represents a limitation of the study. Cembrowski et al observed that glucose concentrations for QC solutions were higher in the winter months and found the same results in human blood.¹⁶

Conclusions

High temperature and RH decreased the glucose values obtained with glucose oxidase-based meters. Therefore, we recommend performing quality control if out of range of operating temperature or RH occurs. Glucose test strips should be used only if quality control results are within acceptable ranges. Proper monitoring of temperature and RH during storage and test performance is necessary to ensure reliability. In tropical countries, risk management principles, for example, individualized quality control plans,²² can be implemented to ensure the quality of test results, which may be affected by environmental stresses. These precautions should facilitate ISO certification and long-term quality of patient care.