The effects of changing temperature correction factors on measures of acidity calculated from gastric and oesophageal pH recordings

Authors


Dr J. D. Gardner, Science for Organizations, Inc., 156 Terrace Drive, Chatham, NJ 07928, USA.
E-mail: gardnerj@scifororg.com

Summary

Background  Recently, Medtronic notified customers that new correction factors should be used for their Slimline and Zinetics24 single-use, internal-standard pH catheters.

Aim and Methods  We selected 24-h recordings of oesophageal and gastric pH with the Zinetics24 from our archives for five healthy subjects and for five gastro-oesophageal reflux disease subjects who were studied at baseline and again after 8 days of treatment with a proton-pump inhibitor. All pH values obtained with the old correction factors were rescaled using the new correction factors. Values for median pH, integrated acidity and time pH ≤ 4 were then calculated from pH values with old and new correction factors.

Results  The new correction factors changed values for median pH, integrated acidity and time pH ≤ 4. Values for median pH and integrated acidity changed in a predictable, proportionate way, whereas values for time pH ≤ 4 did not.

Conclusions  The new correction factors will not change the interpretation of previously published results with median pH or integrated acidity. In contrast, values for time ≤4 cannot be converted in an obvious way with the new correction factors. Instead, the raw pH data will need to be rescaled and values for time pH ≤ 4 recalculated using the rescaled pH data.

Introduction

When gastric or oesophageal pH is recorded in humans using an antimony electrode, the recorded values need to be adjusted because of temperature-dependent variation in the relation between potential difference and pH. Gastric and oesophageal pH values are recorded at 37 °C, but the pH electrode is calibrated at ambient temperature (21 °C). As a result, the pH values need to be rescaled from the calibration relationship at 21 °C to the corresponding value at 37 °C using so-called ‘temperature correction factors’ (CFs) provided by the manufacturer. This rescaling process is usually performed automatically by the software used to process the pH recordings.

Recently, Medtronic notified customers that new CFs should be used for the Slimline pH catheter as well as for the Zinetics24 single-use, internal-standard pH catheter (a copy of this notice is in the Appendix). Furthermore, a study that measured the pH of orange juice in vitro at 37 °C with a glass pH electrode and a Medtronic Slimline pH catheter found that the average pH measured with the Slimline catheter was 0.6 pH units less than that measured with the glass electrode.1 As measurements of pH with a glass electrode are not temperature-sensitive, the authors concluded that this difference reflected a flaw in the Medtronic software used to process the data.1 We wanted to analyse the impact on values for different measures of acidity such as median pH, integrated acidity and time pH ≤ 4. Because some of our previous publications2–7 are based on gastric and oesophageal pH recorded with the Medtronic Zinetics24 single-use, internal-standard pH catheter, we also wanted to examine the extent to which these new CFs might impact our previous results.

Materials and methods

Rescaling refers to the process of converting values on one scale to corresponding values on another scale. A rescaling process that is probably familiar to most readers is converting temperature in °C to temperature in F, or vice versa.

Figure 1 illustrates how the Medtronic software uses CFs to rescale the measured pH. The rescaling process involves linear interpolation from values on one scale to those on another scale. The left panel illustrates the first step in a pH recording – calibrating the electrode at ambient temperature (21 °C). Two calibration solutions are used – one at pH 7.0 and one at pH 1.07. Medtronic has determined experimentally that the pH electrode functions differently at 37 °C (the body temperature at which pH is recorded) than at 21 °C; therefore, the calibration at 21 °C has to be rescaled using the CFs determined experimentally by Medtronic. The middle panel gives the experimentally determined CFs. In the right panel, the high CF is 1.0; therefore, the high CF is subtracted from pH 7.0 at 21 °C to rescale the calibration pH to 6.0 at 37 °C. The low CF is 0.70; therefore, the low CF is subtracted from pH 1.1 at 21 °C to rescale the calibration pH to 0.37 at 37 °C. The measured pH values are first located on the 21 °C calibration line and then linearly rescaled to 37 °C in proportion to the rescaled calibration values.

Figure 1.

Illustration of how correction factors are used to rescale the calibration pH.

Initially, Medtronic used the words ‘correction factor’ to refer to the value that is subtracted from the calibration pH to give the rescaled value as illustrated in Figure 1. This is the definition that will be used throughout the present study. Subsequently, however, Medtronic software designers changed the definition of ‘correction factor’ to refer to the value of the rescaled pH. This revised definition is the one that is currently used by Medtronic.

Figure 2 illustrates how we rescaled previously recorded values for gastric and oesophageal pH. The left panel gives the calibration pH values rescaled to 37 °C using the old CFs. These are the same pH values given in the right panel of Figure 1. The right panel gives the calibration pH values rescaled to 37 °C using the new CFs (high CF 0.78 and low CF 0.01).

Figure 2.

Illustration of how to rescale the pH for different correction factors.

In Figure 2, a new pH (P2) can be calculated from the corresponding old pH (P1) using the following equation.

image

where H1 and H2 refer to the original and rescaled high pH calibration values at 37 °C, and L1 and L2 refer to the original and rescaled low pH calibration values at 37 °C.

For example, if P1 is 4.0, the above equation gives a value for P2 of 4.4.

image

To assess the potential impact of the new Medtronic CFs on different measures of acidity calculated from pH recordings, we selected 24-h recordings of oesophageal and gastric pH from five healthy subjects and from five gastro-oesophageal reflux disease (GERD) subjects who were studied at baseline and again after 8 days of treatment with a proton-pump inhibitor (PPI; oral omeprazole or rabeprazole −20 mg once daily). The records from subjects with the five lowest subject numbers in each group were selected from the database we have used for previous publications.2–7 We chose these three different groups to examine a broad range of values for acidity. Each 24-h recording had 21 600 oesophageal pH values or 21 600 gastric pH values.

The pH recordings were all obtained using the Zinetics24 single-use, internal-standard pH catheter and were initially processed using polygram for Windows, Version 2.04, Medtronic Synectics and the old CFs illustrated in Figure 2. The data were then converted to ASCII files, and stored in an electronic archive. For each pH record, we rescaled all pH values using the new CFs as outlined in Figure 2. After excluding all values that were less than pH 0.5 or greater than pH 7.58 from each record, we calculated integrated acidity, time pH ≤ 4, and median pH from data that had been processed using the old CFs as well as the new CFs.

As omitted data and how they are dealt with will be an important factor for some of our results, we have summarized how the different versions of Medtronic software exclude pH data. Table 1 outlines the differences in the ways that different versions of Medtronic software exclude pH values. If users are unaware of these differences, they may have difficulty interpreting effects of changing CFs on different measures of acidity. One way to avoid excluding data is to create an ASCII file containing all values produced by a particular software version using a particular set of CFs. Each Medtronic software version has an ‘export’ function that allows the user to create an ASCII file of all values.

Table 1.  pH exclusion criteria for different versions of Medtronic software used to process data from gastric and oesophageal pH recordings
Software versionExclusion criteria
  1. Information was provided by Tom Bombeck (Medtronic Gastroenterology, Shorview, MN 55126-2983, USA).

DOSUser can exclude consecutive values by inserting ‘ignore’ periods in one or more sections of the recording. Percentage of values between cut points can be determined using the oscillatory index
polygram for WindowsUser can exclude consecutive values by inserting ‘ignore’ periods in one or more sections of the recording. By default, the software excludes pH values <0.0 or >9.0. These settings can be altered by the user. Percentage of values between cut points can be determined using the oscillatory index
polygram 98User can exclude consecutive values by inserting ‘ignore’ periods in one or more sections of the recording. It is not possible to exclude individual pH values based on the magnitude of the value. Percentage of values between cut points can be determined using the oscillatory index, but the lowest cut point allowed by the software is pH 1.0
POLYGRAM.NETUser can exclude consecutive values by inserting ‘ignore’ periods in one or more sections of the recording. By default, the software excludes pH values <0.0 or >9.0. These settings can be altered by the user. Percentage of values between cut points can be determined using the oscillatory index, but the lowest cut point allowed by the software is pH 1.0

Statistical analyses were performed using GraphPad for instat version 3.01 for Windows. Curve-fitting was performed using GraphPad Prism version 4.0 for Windows.

Results

Figure 3 illustrates that median pH values with the old CFs ranged from 0.85 to 6.8, and as would be expected, values for median gastric pH were lower than values for median oesophageal pH. The new CFs increased values for median pH and the magnitude of the increase was greater at low pH values than at high pH values. As changing the CFs alters the magnitude of a pH value, but not its relative position in a series of values ordered from highest to lowest, the results in Figure 3 can also be obtained by simply using the value for median pH with the old CFs plus the values for the new and old CFs illustrated in Figure 2.

Figure 3.

Median pH with old and new correction factors (CFs). Values are from 15 recordings of gastric pH and 15 recordings of oesophageal pH. Values for old and new CFs are illustrated in Figure 2. The solid diagonal line is the identity line. Values that are the same with old and new CFs will lie on the identity line. Values that are lower with the old CFs will lie below the line and vice versa.

Figure 4 (top) illustrates that values for integrated acidity ranged from 1 to 4317 mmol h/L, and that values for integrated acidity with the new CFs are clearly lower than with the old CFs. The data in Figure 4 (top) were best-fit by a straight line having a Y-intercept of zero and a slope of 4.8 indicating that the magnitude of the decrease in integrated acidity with the new CFs is proportional to the value of integrated acidity with the old CFs.

Figure 4.

Integrated acidity with old and new correction factors (CFs). Values are from 15 recordings of gastric pH and 15 recordings of oesophageal pH. Values for old and new CFs are illustrated in Figure 2. The solid diagonal line is the identity line. Values that are the same with old and new CFs will lie on the identity line. Values that are lower with the old CFs will lie below the line and vice versa. The bottom panel is a log–log plot of data in the top panel.

Plotting the data in Figure 4 (top) in a log–log format (Figure 4-bottom) makes it easier to distinguish values for oesophageal acidity from those for gastric acidity. In this log–log format, the points fall on a straight line that is parallel to the identity line. These data could be fit by a straight line with a slope of 1.0 and a Y-intercept of 0.60. The important practical issue is that values for integrated acidity obtained with the old CFs can be converted to values with the new CFs by multiplying the original values by a constant scaling factor (0.25).

As would be expected from the results in Figure 4, the new CFs produced no significant change in the values for PPI inhibition of integrated acidity (data not shown). Figure 5 illustrates that values for time pH ≤ 4 with the old CFs ranged from 0.1% to 82%, and that values for time gastric pH ≤ 4 were higher than values for time oesophageal pH ≤ 4. With new CFs, the magnitude as well as the direction of the change in values for time pH ≤ 4 varied depending on the value with the old CFs. That is, values that were relatively low with the old CFs decreased with the new CFs. In contrast, values that were relatively high with the old CFs increased with the new CFs. Moreover, as shown in the log–log plot in Figure 5 (bottom), the lower the value with the old CFs, the greater the magnitude of the decrease with the new CFs.

Figure 5.

Time pH ≤ 4 with old and new correction factors (CFs). Values are from 15 recordings of gastric pH and 15 recordings of oesophageal pH. Values for old and new CFs are illustrated in Figure 2. The solid diagonal line is the identity line. Values that are the same with old and new CFs will lie on the identity line. Values that are lower with the old CFs will lie below the line and vice versa. The bottom panel is a log–log plot of data in the top panel.

The results in Figure 5 indicate that values for time pH ≤ 4 obtained with the old CFs cannot be converted to values with the new CFs by simply multiplying the original values by a constant scaling factor, as was the case for integrated acidity calculated from the same data. Instead, values for time pH ≤ 4 must be recalculated using both a constant scaling factor and an exponential function of the value obtained with the old CFs.

We considered the possibility that the increase in values for time pH ≤ 4 with the new CFs might occur because values that were less than pH 0.5 were excluded from all analyses. All values that increased with the new CFs were for gastric pH; therefore, we examined the relationship between the percentage of values below pH 0.5 and the change in the value for time gastric pH ≤ 4 with the new CFs.

Figure 6 illustrates that there was a linear relationship between the increase in time gastric pH ≤ 4 with the new CFs and the percentage of pH values <0.5 with the old CFs. These results indicate that as values below pH 0.5 were excluded from the original analyses, the increase in pH with the new CFs produced values that were above pH 0.5 but yet still ≤4.0 thereby increasing the value of time pH ≤ 4. When values of gastric pH < 0.5 were <6%, values for time gastric pH ≤ 4 decreased with the new CFs. When values of gastric pH < 0.5 were >6%, values for time gastric pH ≤ 4 increased with the new CFs.

Figure 6.

Relationship between the change in time gastric pH ≤ 4 with old and new correction factors (CFs) and the percentage of gastric pH values <0.5 with the old CFs. Values are from 15 recordings of gastric pH. Values for old and new CFs are illustrated in Figure 2. The Y-axis plots the value with new CFs as a percentage of the value with the old CFs. The solid line is the least-squares fit of the data.

As some studies have compared values for time pH ≤ 4 with different PPIs (e.g. Ref.9), we evaluated the effect of changing the CFs on time pH ≤ 4 with a PPI (Figure 7-top). Values for oesophageal pH ranged from 1.1% to 10.2% with the old CFs, and each value was reduced with the new CFs so that the range of values was compressed to a new range from 0.7% to 6.2%. One possible consequence of such an effect is that values that were significantly different with the old CFs might not be significantly different with the new CFs. Figure 7 (top) also illustrates that values for time gastric pH ≤ 4 with a PPI may not all change in the same direction with new CFs. As a result, values that were not significantly different with the old CFs might become significant with the new CFs. On the contrary, the range of values may be sufficiently compressed so that values that were significantly different with the old CFs might not be significant with the new CFs. For example, for the four values that decreased with new CFs, the range with the old CFS was 34.6–56.1%, whereas that with the new CFs was only 29.0–44.7%. Finally, in some instances in Figure 7 (top) when values decreased, the line connecting one pair of values crossed that connecting another pair of values. This phenomenon indicates that the proportional decrease varied among the different values, with no apparent relationship to the magnitude of the value with the old CFs.

Figure 7.

Effect of changing correction factors (CFs) on time pH ≤ 4 with a proton-pump inhibitor (PPI). Values for old and new CFs are illustrated in Figure 2. Values in the top panel are from recordings of gastric or oesophageal pH from five gastro-oesophageal reflux disease (GERD) subjects during day 8 of treatment with a PPI, and values connected by a solid line are from the same record. Gastric values are indicated by open circles; oesophageal values by closed circles. Values in the bottom panel are for inhibition of time oesophageal or gastric pH ≤ 4 calculated as 100 × (Baseline value − Day 8 value)/Baseline value from five GERD subjects treated with a PPI. In the bottom panel, the solid diagonal line is the identity line. Values that are the same with old and new CFs will lie on the identity line. Values that are lower with the old CFs will lie below the line and vice versa.

Instead of simply comparing values for time pH ≤ 4 with a PPI, some might be interested in calculating the magnitude of PPI-induced inhibition (e.g. Ref.2). Figure 7 (bottom) illustrates that the new CFs usually increased the magnitude of PPI-induced inhibition of time pH ≤ 4; however, the magnitude of this inhibition was variable and did not depend on the magnitude of the value with the old CFs.

Discussion

The present results indicate that it is a straightforward process to estimate the impact of new CFs on values for median pH obtained from pH recordings using the old CFs. As the rescaling process does not change the order of the individual pH values, the value for median pH itself can be rescaled using the process described in Materials and methods, without resorting to the original individual pH values.

Because values for integrated acidity represent the time-weighted average of the acid concentration expressed in mm, values with the old CFs change substantially with the new CFs. On the contrary, the values change in a nearly constant, proportional manner so that values with the new CFs are 25% of corresponding values with the old CFs. Even though the proportionality factor varies slightly from low to high values of integrated acidity, the resulting differences are well within the range of replicate values for integrated acidity determined in the same subjects on different occasions. One important practical consequence of our findings with integrated acidity is that the interpretation of previous results with integrated acidity will not change,2–7 because all previously reported values will change by a constant factor.

Although the percentage change in values for time pH ≤ 4 is generally smaller than that in corresponding values for integrated acidity, the direction and magnitude of the change can vary depending on the value for time pH ≤ 4 with the old CFs. We were initially perplexed by this finding because we anticipated that the pattern of change in values for time pH ≤ 4 would be similar to that for median pH. Once we considered the potential impact of values that were excluded from our calculation of time pH ≤ 4 because they were below pH 0.5, the explanation for the increase in values for time pH ≤ 4 the new CFs became clear. That is, the number of pH values that were less than pH 0.5 with the old CFs and were increased to greater than pH 0.5 but yet below pH 4 with the new CFs exceeded those that were increased to above pH 4 with the new CFs. The net result was an increase in time pH ≤ 4. This phenomenon illustrates a general issue with values for time pH ≤ 4; namely, the magnitude and direction of change with the new CFs will depend on the distribution of pH values with old CFs. In this regard, we have analysed gastric pH recordings from other studies that excluded pH values that were below pH 0.5 or above pH 7.5. In a number instances, values for time gastric pH ≤ 4 with the old CFs did not increase with the new CFs, even though 15% of values were below pH 0.5 with the old CFs and no values were below pH 0.5 with the new CFs. These records tended to have relatively high values for time pH ≤ 4 with the old CFs (in the range of 80–90%), and the number of pH values that were increased to above pH 0.5 but below pH 4 was less than the number that were increased to above pH 4. The practical issue is that values for time ≤4 with the old CFs cannot be converted in an obvious way to corresponding values with the new CFs. Instead, the raw pH data will need to be rescaled using the new CFs and values for time pH ≤ 4 recalculated using the rescaled pH data.

Our findings raise other important issues regarding measures of time pH ≤ 4. For example, values with PPIs, particularly for time gastric pH ≤ 4, that are not significantly different with the old CFs may become significantly different with new CFs. Also, values for time pH ≤ 4 that are significantly different with the old CFs may not be significantly different with new CFs. A similar phenomenon might also occur with values for inhibition of time pH ≤ 4, because the magnitude of the change in values with the new CFs is variable and is not related to the magnitude of the values with the old CFs. These issues, for example, may be important for studies that have compared time gastric pH ≤ 4 with different PPIs or with different formulations of the same PPI.

With antimony electrodes, it is possible to circumvent the temperature correction issue by calibrating the electrode at 37 °C so that there will be no need for any adjustment to the data. This solution, however, would require a water bath at 37 °C and might be impractical for many clinical investigators. If one is only interested in gastric acidity, it is possible, although labour-intensive, to aspirate samples of gastric fluid and measure pH ex vivo using a glass electrode, which is not temperature-sensitive (e.g. Ref.10).

Acknowledgement

This study was supported by Science for Organizations, Inc. and Blossomtech Inc.

inline image

inline image

Ancillary