Time-based V-slope for determination of anaerobic threshold Authors: Hirotaka Nishijima, Kazuyuki Kominami, Toko Katsuragawa, Masatoshi Akino Published: 2026 ABSTRACT BackgroundThe standard method for determining the noninvasive anaerobic threshold (AT) is the V-slope (VS) method. It is often supplemented by the ventilatory equivalent (VEQ) and end-tidal (ET) methods. However, a comparison issue arises: VS plots VCO2 against VO2 (x-axis), whereas VEQ and ET conventionally plot each variable against elapsed exercise time. This study aimed to develop a time-based VS graph to align all three methods on the same x-axis.MethodsV-slope is drawn as a square with equal sides, with the diagonal line representing the respiratory exchange ratio (R) = 1. V-slope generally follows the R = 1 line. This provides a framework in which the vertical difference (d) between a VCO2 data point and the corresponding point on the R = 1 line may be used and plotted against time; that is, d equals the measured VCO2 minus VCO2 on the R = 1 line. Since VCO2 equals VO2 on the R = 1 line, d equals VCO2–VO2. This new measure, d (mL/min, VCO2), combines the relation between VCO2 and VO2 into a single value, which now can be plotted against time. In this graph the diagonal of the conventional V-slope (R = 1) becomes the horizontal x-axis (time). The plot using this modification of VS identifies AT as the first upward inflection point against a flat baseline. Data from 127 subjects (age: 51 ± 21; 66 healthy, 20 treated for cardiovascular risk factors, 41 with cardiac disease) who underwent maximal symptom-limited exercise were analyzed. Each of the four AT methods (d, VS, VEQ, ET) was independently assessed in a blinded manner. Reliability was evaluated using Altman’s coefficient of repeatability (CR). Statistical comparison of variability of each detection method was performed after the squared difference was log transformed because the data was not normally distributed.ResultsThe mean AT (mL/min) and detection rates for d, VS, VEQ, and ET were 857 ± 414 (93%), 939 ± 407 (96%), 946 ± 176 (82%), and 862 ± 379 (85%), respectively. The d method identified AT significantly earlier than the conventional VS did (p < 0.001). The CR of d, VS, VEQ, and ET was 117, 115, 152, and 175 mL/min, respectively. The mean squared difference of d method was significantly smaller than those of VS, VEQ or ET (p < 0.001), indicating improved reliability of AT detection by d. The collective reviewing of all three time-aligned graphs (d, VEQ, ET) revealed that delayed or suppressed ventilatory responses to CO2 (d vs VEQ) in 48 subjects (38%).ConclusionsThe time-based V-slope was created by plotting changes in VCO2 from the R = 1 diagonal line against time; it aligned all three AT detecting methods on the same time axis, allowing precise detection of differences between methods. It also made AT detection more reliable, probably because the upward inflection breakpoint arises from the flat baseline and does not have the overlapping data points seen in V-slope, making it easier to identify a breakpoint. The d graph also identified AT significantly earlier than V-slope did. Introduction The anaerobic threshold (AT) was originally defined as the point during incremental exercise testing at which blood lactate begins to increase above the resting level (hence the term lactate threshold). AT was initially believed to represent the onset of anaerobic metabolism (Wasserman et al., 2012); however, subsequent studies did not support an anaerobic origin of increased level of lactate (Poole et al., 2021). Nevertheless, the breakpoint at which lactate begins to rise has continued to be regarded as a physiological marker or reference point indicating the boundary between moderate and severe exercise (Iannetta et al., 2020). AT is also used as a reference point for prescribing exercise intensity in cardiac rehabilitation (Hansen et al., 2022). Elevated lactate is considered to be a response to increased glycolytic flux elicited by an increased work rate. The term AT also continues to be used despite criticism. Early on, it was observed that the blood lactate elevation was accompanied by a disproportionate increase in CO2 output versus VO2 uptake. This was attributed to the buffering of hydrogen iron (H+) from lactic acid by bicarbonate (HCO3−), producing additional CO2. Consequently, a non-invasive method of detection of AT using respiratory gas measurements became popular, supported by the rapid advancement of technology (such as cardiopulmonary exercise testing). AT is also referred to as gas exchange threshold (GET), ventilatory threshold (VT), or lactate threshold (LT). The standard noninvasive method for determining AT is the V-slope (VS) method. It is recommended that VS be supplemented with two other methods: the ventilatory equivalent (VEQ: VE/VO2 and VE/VCO2) and end-tidal (ET) methods (Wasserman et al., 2012; European Respiratory Society: ERS Task Force on Standardization of Clinical Exercise Testing, 1997; American Thoracic Society; American College of Chest Physicians, 2003; Balady et al., 2010; ERS Task Force on Standardization of Clinical Exercise Testing. European Respiratory Society, 1997). An increase in VCO2 should be accompanied by an increase in ventilation, reflected by changes in VEQ and ET, and these ventilation-based parameters also serve as a check for non-specific hyperventilation. The VS with VEQ and ET approach has two main problems. First, in the VS method, carbon dioxide production (VCO2) is plotted against oxygen uptake (VO2) on the x-axis), whereas in the VEQ and ET methods, each variable is plotted against elapsed time (or work rate; x-axis). When discrepancies between VS and VEQ/ET appear to be present, it is difficult to pinpoint the precise location of the discrepancy as the data points do not share the same time-based x-axis. Second, even though VS is noninvasive, it is somewhat difficult and awkward to perform because the AT breakpoint must be identified as a point of disproportionate rise of VCO2 relative to VO2 on an already obliquely ascending baseline slope of approximately 45 degrees. Therefore, the aim of this study was to modify the standard VS graph to (1) construct a time-based VS representation to enable direct comparisons with VEQ and ET plots, and (2) restore the baseline to a horizontal line so that the breakpoint emerges from a flat x-axis rather than from an obliquely ascending baseline. A previous version of this work has also been deposited at SportRxiv (DOI: https://doi.org/10.51224/SRXIV.551). Materials and Methods Developing a d plot The V-slope is drawn as a square with equal sides, with the diagonal line representing the respiratory exchange ratio (R) = 1 (1). The V-slope generally follows the R = 1 line. This provides a framework in which the vertical difference (d) between a VCO2 data point and the corresponding point on the R = 1 line may be used and plotted against time; that is, d equals the measured VCO2 minus VCO2 on the R = 1 line. Since VCO2 equals VO2 on the R = 1 line, d equals VCO2–VO2. This new measure, d (mL/min, VCO2), combines the relation between VCO2 and VO2 into a single value, which now can be plotted against time. In this graph, the diagonal of the conventional V-slope (R = 1) becomes the horizontal x-axis (time). Every VCO2 change around R (d) is graphically synchronized with VEQ and ET. This method identifies AT as the first progressively upward inflection point of d, similar to the VEQ and ET methods. As an example, consider point A in the graph (Fig. 1A). A perpendicular line is drawn from A to the x-axis. Its intersections with the R = 1 and the x-axis are indicated as B and C, respectively. On any point on the R = 1 line, VCO2=VO2. Therefore, AB(d)=AC–BC=VCO2–VO2. An alternative explanation is: since the V-slope is plotted on equal scales, OBC is an isosceles triangle. Thus, BC (VCO2) = OC (VO2). Therefore, the difference (d) is calculated simply as VCO2–VO2. When a data point is below the R = 1 line (e.g., point E), d becomes negative (EF –DF). Figure 1B shows the transformed V-slope with d plotted against VO2. Figure 1C shows d plotted against time. When VO2 is used on the x-axis, data points overlap, whereas using time as the axis yields a single data point at each time interval. The aligned AT is indicated by a dotted line. (More examples of the V-slope aligned with the d-plot on the same VO2 x-axis are shown in Supplemental File 2). Evaluating the d-method in a population A database of 127 subjects (age 51 ± 21 years) who performed maximal symptom-limited exercise was used to assess the utility of the d plot (Table 1). Of these 127 subjects, 93 were men, 66 were sedentary to low-active healthy subjects, 20 were under treatment for cardiovascular risk factors (hypertension, hyperlipidemia, or diabetes mellitus), and 41 had cardiac disease (31 of ischemic origin). The exercise ramp protocol was 5 W/min (n = 22), 10 W/min (n = 48), 15 W/min (n = 19), 20 W/min (n = 17), 25 W/min (n = 20), or 30 W/min (n = 1). The database was the same one that was used for our previous study to develop a mathematical derivation of AT (Nishijima et al., 2019), except for one case of duplication, which was removed. Cardiopulmonary exercise testing was performed using a stationary bicycle (StrengthErgo 8; Mitsubishi Electric Engineering, Tokyo) and a breath-by-breath gas analyzer (AE-300S; Minato Ikagaku Co., Tokyo). The gas analyzer was calibrated for volume and O2 and CO2 concentrations before each test. Written informed consent was obtained from each participant. The breath-by-breath gas change was transformed into an eight-breath moving average using Excel, and the AT was visually read on the respective Excel graphs. AT was always read on a data point (never between data points). (A–E) Case 1 (73-year-old healthy female). Representative example plots of VS, d vs VO2, d vs time, VEQ vs time, and ET vs time. Five plots are displayed from top to bottom: (A) v-slope (VCO2 versus VO2), (B) d versus VO2, (C) d versus time, and (D) ventilatory equivalent versus time (VEQ: VE/VO2, VE/VCO2) and, (E) end-tidal O2, CO2 concentration versus time. A (V-slope) shows the framework of VS to time-based VS (d). Please see Materials and Methods (‘Developing a d plot’) for detailed explanation using the points (A, B, C, D). The arrow labeled “ramp” indicates the start of ramp exercise. In (B) d versus VO2 is shown just for reference to demonstrate that d can be aligned with the V-slope on the same VO2 axis as VS (dotted AT line aligned). In (C) d versus time is displayed. Solid black dots represent d, while gray dots represent VO2. (for reference only). In (D) VEQ versus time is displayed. Solid black dots and gray dots represent VE/VO2 and VE/VCO2, respectively. VE/VCO2 is shifted downward by 20 units (shown as VE/VCO2–20) to avoid overlap with VE/VO2. In (E) end-tidal VO2 and VCO2 are plotted against time. AT is identified at the same timeline, aligned across (C, D and E) (dotted arrow). The subsequent figures follow the same basic format. Each of the four AT detecting methods (VS, d, ET, and VEQ) was applied independently and blinded to the other methods. VS AT was determined by the modified V method described by Sue et al. (1988), rather than the more complex mathematical method described by Beaver, Wasserman & Whipp (1986), which requires computerized analysis. ATS/ACCP also suggests this simpler approach. Wasserman et al. (2012) exclusively used this modified method in all 80 of their case presentations. This method was also used in a large multicenter reliability study of AT (Myers). This modified V method proposes that the pre-AT segment of the V-slope is parallel to the respiratory exchange ratio (R) =1 line (diagonal of V-slope plot). Once this line is visually set, AT is the point that breaks off from this parallel line. This method of identifying AT was used for this study because a breakpoint on the V-slope is easy to observe online during exercise testing. The ET and VEQ parameters were used mainly to check for nonspecific hyperventilation. When the VS AT appeared to be different from the ET/VEQ AT, V-slope AT was adopted. AT determination by d, ET, and VEQ was performed by identifying a progressively ascending breakpoint. All readings were conducted by the same investigator (H.N.). Each AT was also read twice in a random order within a span of 4 months to assess reliability. All AT readings were coded for a level of confidence at the time of each reading as (1) confident, (2) less confident, or (3) unable to detect AT. The last AT reading where the d, VEQ, and ET plots were simultaneously viewed (Combined) was performed two months after the above, with one week separating two readings. At the time of the Combined reading, differences between breakpoints on the d-plot and VEQ plot were noted and coded as follows: code-0 as VEQ earlier than d, code-1 as VEQ within 1 min of d, code-2 as VEQ delayed > 1 min of d, code-3 as VEQ delayed until the respiratory compensation point (RCP), and code-4 as no detectable breakpoint. 10.7717/peerj.20847/table-1Table 1Demographic and clinical characteristics.Healthy(n = 66)Those with CV risks(n = 20)Patients with cardiac disease(n = 41)Age, y38 ± 21*59 ± 1167 ± 9Sex (male/female)47/1920/026/15Body weight, kg62 ± 10*72 ± 759 ± 9Height, cm167 ± 8167 ± 5160 ± 8Body mass index, kg/m222.0 ± 3.125.8 ± 2.923.3 ± 3.1Heart disease etiology: ischemic (%)NANA75Medications (n)NAACE or ARB916Diuretic27Beta-blocker017Inotropics00Ca channel blocker177Anti-lipidemic316Anti-diabetic24 Notes. *Significant at p < 0.001 (one-way analysis of variance). CVcardiovascular ACEangiotensin-converting enzyme inhibitor ARBangiotensin II receptor blocker Cacalcium Demographic and clinical characteristics. Notes. Significant at p < 0.001 (one-way analysis of variance). CVcardiovascular ACEangiotensin-converting enzyme inhibitor ARBangiotensin II receptor blocker Cacalcium cardiovascular angiotensin-converting enzyme inhibitor angiotensin II receptor blocker calcium On the VEQ plot, VE/VCO2 is displaced downward by 20 units (see the legend to the Figures: VE/VCO2–20: VE/VCO2 minus 20) to avoid overlapping plots by VE/VO2. Comparison of means between all five AT detecting methods (VS, d, VEQ, ET, and Combined) was conducted using repeated measures analysis of variance (ANOVA). Calculations used the mean of the first and second reading values of each method (AT had to be determined on both occasions). Reliability of the method was assessed by using the limits of agreement (Bland-Altman) and was expressed as the coefficient of repeatability (CR), calculated as 1.96 times the standard deviation of the differences between the two measurements (Bland & Altman, 1986). A p-value less than 0.05 indicated a significant difference. The variability of two readings of each AT method was also statistically compared. The difference between the first and second readings of AT was squared. First, each method was compared by the nonparametric Wilcoxon signed-ranks test (six pairwise comparisons). Multiple comparisons were corrected by multiplying each p-value by six (Bonferroni correction). Second repeated measures ANOVA was conducted after the log transformation of the squared difference (MedCalc, MedCalc Software, Ostend, Belgium) because the squared difference was not normally distributed. Zero values (perfect agreement) were replaced by the lowest value in the series. Intra-observer reliability was also assessed by intraclass correlation coefficient, ICC (1,1). Inter-observer reliability from our group has been reported (Minagawa et al., 2025). The percent of predicted peak VO2 and AT was calculated from a regression equation from healthy Japanese subjects, based on age and sex (Ohta et al., 1999). This study was conducted in accordance with the principles of the Declaration of Helsinki. The research protocol was approved by the institutional review board of Kitano Hospital (formerly Sapporo Ryokuai Hospital, No. 18-2). Results The demographic and clinical information of the study group is summarized in Table 1. The basic cardiopulmonary exercise testing variables are shown in Table 2. The mean AT and the AT detection rate of each method are presented in Table 3. The mean d-AT was significantly less than either the mean VS or VEQ AT (p < 0.001), but was not significantly different from the mean ET; that is, the d-AT breakpoint appeared earlier than that of either VS or VEQ. The mean AT detected using the Combined method also agreed with the mean AT values detected using d and ET. The detection rate of AT was approximately equal using d and VS (93%, 96%), but was significantly lower using VEQ (82%, p = 0.017 vs d; p = 0.004 vs VS). The simultaneous combined use of d, VEQ, and ET did not improve the AT detection rate. However, the combined use of d, VEQ, and ET (all plotted against time) revealed a group of 48 subjects who manifested a delayed or suppressed ventilatory response (d vs. VEQ; code 2, VEQ delayed >1 min: n = 35: code 3, VEQ delayed until RCP: n = 11; and code 4, no VEQ breakpoint: n = 2). An example of code 2 (VEQ delayed > 1 min of d) is shown in Fig. 2. A-D Case-2, an example of code 3 (VEQ delayed until RCP) is shown in Figs. 2E–2H Case-3, and an example of code 4 (no detectable VEQ-AT) is shown in Figs. 2I–2L Case-4. The subjects with a delayed ventilatory response were significantly older (56.2 ± 21.6 vs. 46.6 ± 21.6, p = 0.015) and manifested a significantly lower peak VO2 (23.4 ± 9.9 vs. 28.1 ± 10.3 mL/kg/min, p = 0.014) and AT (15.1 ± 4.5 vs 12.4 ± 3.8, p = 0.001) than those who did not show this abnormality. However, body weight did not differ significantly between these two groups (64.3 ± 10.9 vs 62.3 ± 1 0.1, p = 0.308). Clinical classification (healthy, risk factors, heart disease) was also not related with this abnormality. In a comparison of d vs. ET, the delayed or suppressed ventilatory response was observed in 21 subjects. 10.7717/peerj.20847/table-2Table 2Cardiopulmonary exercise testing in three groups of subjects.Healthy subjects(n = 66)Patients with CV risks (n = 20)Patients with heart disease (n = 41)p-valueRamp duration, s554 ± 141586 ± 136503 ± 1500.072Peak work rate (W)171 ± 65125 ± 3773 ± 29<0.001Borg scale (/20)16.7 ± 1.316.8 ± 1.415.4 ± 1.6<0.001Peak VO2, mL/kg/min33 ± 922 ± 616 ± 4<0.001% predicted peak VO2104 ± 1987 ± 1576 ± 18<0.001Peak VO2, mL/min2,041 ± 7311,593 ± 395989 ± 332<0.001Peak VCO2, mL/min2,396 ± 8711,824 ± 4781,102 ± 398<0.001Peak R1.17 ± 0.101.14 ± 0.091.11 ± 0.090.004AT (by VS), mL/kg/min16.9 ± 3.912.4 ± 3.310.6 ± 2.3<0.001% predicted AT98 ± 1982 ± 1878 ± 16<0.001AT, % peak VO254 ± 1254 ± 766 ± 11<0.001Peak HR, bpm164 ± 23136 ± 28117 ± 19<0.001Peak BP-systolic, mmHg187 ± 25196 ± 25174 ± 290.006 Notes. p-value by one-way analysis of variance. ATanaerobic threshold BPblood pressure CVcardiovascular HRheart rate Rrespiratory exchange ratioVCO2minute carbon dioxide productionVO2minute oxygen uptake VSV-slope Wwatt Cardiopulmonary exercise testing in three groups of subjects. Notes. p-value by one-way analysis of variance. ATanaerobic threshold BPblood pressure CVcardiovascular HRheart rate Rrespiratory exchange ratioVCO2minute carbon dioxide productionVO2minute oxygen uptake VSV-slope Wwatt anaerobic threshold blood pressure cardiovascular heart rate respiratory exchange ratio minute carbon dioxide production minute oxygen uptake V-slope watt 10.7717/peerj.20847/table-3Table 3Comparison of five AT detection methods.All AT readings (confidence code = 1 and 2)Confidence code = 1 onlyAT n %MeandVSVEQETCombCR n %CRd11893857 ± 414– * * 117917286VS12196939 ± 407 * – * * 115816489VEQ10482946 ± 376 * – * * 1528063124ET10985862 ± 379 * * –1757861147Comb11993869 ± 414 * * –119483859 Notes. *p < 0.001 (Bonferroni, after repeated measures ANOVA). Combcombined CRcoefficient of repeatability dVCO2 minus VO2 (mL/min) ETend-tidal VEQventilatory equivalent Comparison of five AT detection methods. Notes. p < 0.001 (Bonferroni, after repeated measures ANOVA). Combcombined CRcoefficient of repeatability dVCO2 minus VO2 (mL/min) ETend-tidal VEQventilatory equivalent combined coefficient of repeatability VCO2 minus VO2 (mL/min) end-tidal ventilatory equivalent Three representative cases demonstrating the usefulness of d plot. (A–D) Case-2 (63-year-old male with heart disease). Delayed AT detection using VEQ and ET compared with d. The ascending break point (AT-d) is clearly observed in the d plot, whereas it is less distinct and appears later in the ET and VEQ plots (At-ET, AT-VEQ). Two broken vertical lines highlight the difference. Although the AT breakpoint occurs from a descending baseline rather than a horizontal one, this pattern was observed only in 18% of cases. Note also the increased VCO2 signal gain in the d plot compared with the V-slope. (E–H) Case-3 (63-year-old healthy male). No ET or VEQ ascending break points observed until respiratory compensation point (RCP). No clear ascending break points are observed in the ET and VEQ plots until the appearance of RCP. A broken vertical line highlights the break point (AT) in the d plot and the absence of breakpoints in the ET and VEQ plots. (I–L) Case-4 (65-year-old male with heart disease). No AT breakpoints apparent in ET or VEQ plots. The AT breakpoint is difficult to identify in the ET plot, and is not discernible in the VEQ plot, despite a clear VCO2 response in the d and V-slope plots. This suggests an inadequate ventilatory response to CO2. The broken line (AT) highlights the difference. The appearance of AT in this case was very early during ramp exercise. The d plot often displayed a more distinct break point than did the VS plot (Figs. 3A–3D Case-5). An example of hyperventilation confounding the identification of AT is shown in Figs. 3E–3H Case-6. Three more representative cases demonstrating the usfulness of d plot. (A–D) Case-5 (58-year-old healthy male). Easier identification of a breakpoint in d rather than in V-slope plot. A clearly defined AT is more difficult to identify in the V-slope but is easier in the d plot, where it coincides with the ET and VEQ breakpoints (indicated by the broken line). (E–H) Case-6 (82-year-old female with heart disease). Hyperventilation (HV) obscuring the AT breakpoint. The first ascending breakpoint in the d plot appears to be confounded by hyperventilation, as confirmed by an increase in ET-O2 and decrease in ET-CO2, and a simultaneous increase in VEQ-O2 and VEQ-CO2 (first broken line). However, whether the second breakpoint (AT?) represents the true AT is debatable, as it is confounded by the hyperventilation artifact. (I–L) Case-7 (48-year-old male with cardiovascular risk factors). Slope changes rather than ascending breakpoints in ET and VEQ plots. There appears to be two inflection points on V-slope (AT1? And AT2?). The d breakpoint supports AT1? and it appears to coincide with a slowly ascending breakpoint or a slope change in ET plot. However, in the VEQ plot it appears to coincide with the junction point where the descending and the plateau phases meet (slope change). A later ascending break point (AT2?) coincides with the RCP in the ET and VEQ plots, which tends to support AT1? as the true AT. The appearance of AT by d was very early. The mean CR tended to be lower with VCO2 derived measures, d (117) and VS (115), and higher with VE derived measures, VEQ (152) and ET (175). When only the AT with the confidence level of code 1 (confident) was used, CR decreased markedly in all methods (Table 3). Although the mean CR values were not different between the four AT methods using repeated measures ANOVA, it was noted that more data points were clustered around zero with d than with the other four methods. Also, as the data did not show normal distribution, the squared difference was log transformed and the repeated measures ANOVA was again applied. The result showed that the mean squared difference of d was significantly smaller than those of the other three AT methods (Fig. 4). The variability of the AT reading was significantly small with the d method. Pairwise comparison of the original data (before log transformation) by the nonparametric method Wilcoxon signed-ranks test also showed that the squared difference was significantly small with the d method (d vs VS; p = 0.007, vs VEQ; p = 0.001, vs ET; p = 0.001). ICC was 0.98, 0.98, 0.90, and 0.96 for d, VS, VEQ, and ET, respectively. Comparison of AT detection variation among the four methods (d, VS, VEQ, ET). The x-axis values of this box-and-whisker plot is a squared difference of detected AT values. The value was log-transformed because the data were not normally distributed. The d method demonstrated the best reliability among the four methods of AT detection. The combined use of all three graphs suggested a changing slope of VEQ-O2 and ET-O2 against VEQ-CO2 and ET-CO2 as an alternative method of detecting AT (Figs. 3I–3L Case-7). However, this method exhibited poor reliability (for details see Supplemental File 1). Discussion We devised a new, simple time-based VS method in which VCO2 changes are plotted against time. We calculated a new parameter, d, as the difference between VCO2 and VO2 and then plotted d against time. The new parameter, d (mL/min, VCO2), represented a vertical difference between the measured VCO2 and the VCO2 on the diagonal line (R = 1) in the V-slope graph. This enabled a direct comparison of breakpoints and AT derived from using the VS method against ET and VEQ, the other two methods commonly used to detect AT. With the conventional VS method, the y-axis variable, VCO2, is continuously increasing (initially at about 45 degrees to the x-axis), so detecting AT requires discernment of a further disproportionate rise. In contrast, the d-method requires a simple judgement of an ascending point against the level x-axis. This modification is simply another way of plotting the V-slope. Therefore, the basic physiological basis for detecting AT using this method is exactly the same. Traditionally, three methods of AT detection have been employed: VS, ET, and VEQ. The V-slope method is the standard because it detects the disproportional elevation of VCO2 vs VO2, which is a direct result of buffering lactic acid. The ET and VEQ changes occur as a secondary ventilation response to an increase in VCO2. The ET and VEQ also serves as a check against nonspecific hyperventilation. The VS method plots VCO2 against VO2, whereas the ET and VEQ methods plot respective values against time. The lack of a common reference axis makes it visually difficult to assess exactly how a breakpoint on VS corresponds to a breakpoint on ET or VEQ; thus, any existing differences are difficult to pinpoint. There are instances in the literature where ET and VEQ are plotted against VO2, making the x-axis the same as that used for VS (ERS Task Force on Standardization of Clinical Exercise Testing. European Respiratory Society, 1997). This method has not been studied systematically as a method for determining AT. Furthermore, with VO2 as the x-axis, multiple overlapping y-values of ET or VEQ may exist. This may be a reason the method of aligning parameters has not been popular. Supplement File 2 contains aligned examples of VS and corresponding d plots. Conversely, a time-based method allows only one data point for a time line point on the x-axis, which is more advantageous for pinpointing breakpoints. In the past, excess CO2, calculated either as [VCO22/VO2− VCO2] or (VCO2–0.75 × VO2), has been used most often as a time-based measure of VCO2 change (Gaskill et al., 2001; Issekutz Jr & Rodahl, 1961). The value (VCO2–VO2)/heart rate has also been used (Kisaka et al., 2015). In the present study, we used the reverse term, VO2–VCO2, to describe the phenomenon of a rightward shift of VS (Nishijima et al., 2017). Except for one report, all of the existing literature focuses on the physiological study of AT using excess CO2. Only Gaskill et al. (2001) used excess CO2 to systematically determine individual AT in 185 subjects (athletes, active, and sedentary adults), comparing three methods of AT: VS, excess CO2 (time-based), and VEQ (time-based). They reported that the average AT values were statistically similar between the three methods. The acceptance rate (AT detection rate) was 78.4%, 80.5%, and 73% by VS, excess CO2, and VEQ, respectively, with the excess CO2 method having the highest detection rate. Gaskill et al. also reported that combining all three methods significantly decreased the CR of AT to about 230 mL/min with an increased acceptance rate of 90.3%. There are only few published reports in which different methods of AT detection are compared because most studies use the three methods simultaneously. Shimizu et al. (1991) compared three AT methods: VEQ, ET, and VS, each determined by three independent blinded reviewers in 23 subjects (17 with heart disease and 6 healthy men). The mean VEQ AT was higher than the mean V-slope AT, which was higher than the mean ET AT (statistical significance between VEQ vs ET, and V-slope vs. ET). The percentage of tests that were uninterpretable was 6% using VEQ, 7% using ET, and 12% using V-slope. Intraclass correlation coefficient (between reviewers across three bicycle ramp protocols) was 0.63–0.77 for VEQ, 0.49–0.67 for ET, and 0.85–0.95 for V-slope. Variance appeared larger for ventilation derived methods than the V-slope method. The authors cautioned that the method relying on VE may not be reliable in certain conditions such as obesity, air flow obstruction, or chemoreceptor insensibility. Caiozzo et al. (1982) compared five noninvasive ATs against blood lactate threshold (criterion measure) in 16 healthy subjects. They reported AT by VEQ best agreed with lactate threshold. The mean AT by VEQ was higher than the mean AT by V-slope (statistical significance not given). The AT detection rate and reliability were not reported. The findings of the lower rate of AT detection by VEQ (Gaskill et al., 2001) and the lower intraclass coefficient with VEQ (Shimizu et al., 1991) are in agreement with our findings. The mean d-method AT occurred significantly earlier than that in VS AT. This may be because the d-method detects the leading edge of a VCO2 increase on the time axis. With the VS plot, VCO2 data points often overlap in close vicinity on the same VO2 value. Therefore, the AT break point serves as an approximate average. In our study, the VEQ method yielded the lowest rate of AT detection and the lowest coefficient of reliability. Furthermore, on comparing d and VEQ on the time axis, differences on the appearance (or the absence) of breakpoints were readily apparent in 38% of cases. The cause for the delay of VEQ-AT (detecting a rising break point of VEQ-O2) is probably in part related to the fact that it often occurs against the descending baseline of VEQ-CO2, and an upward inflection is more difficult to identify on the descending baseline rather on the flat baseline (see Supplemental File 1 for more detailed discussion). Another reason for the VEQ-AT delay may be the insensitivity of chemoreceptors to CO2 (Shimizu et al., 1991). The pathophysiological significance of this phenomenon is not clear from the present study. Somewhat surprisingly, the d-method did not result in a higher AT detection rate compared to VS. This was probably because the assessor had years of experience of detecting AT primarily using VS, whereas d was developed within the past few years. Reliability in AT detection requires a certain amount of experience (Dolezal et al., 2017; Kaczmarek et al., 2019). Considering this factor, the learning curve of the d-method may be much steeper than that of VS. Knowing whether the AT reading was blinded against an intervention is essential when using AT. The reliability (repeatability) of AT is also essential (Vetter & Schober, 2018; Abbott et al., 2018; Dubé et al., 2016). Because of this, the d AT modification of the V-slope was superior to all the other methods, including the original V-slope. This advantage is likely due to the clearer visualization of a progressively rising breakpoint in the d-graph. In contrast, with the V-slope method, the VCO2–VO2 plots often overlap, making the identification of a distinct breakpoint more difficult. The earlier detection of AT with d AT (i.e., lower mean d AT values compared with V-slope AT) may be attributable to this visual clarity. Further investigation is needed to examine this point. VEQ and ET plots also use point-by-point data referenced to time; however, the leading edge of the breakpoint is generally less distinct. This is likely because changes in VEQ and ET do not directly reflect excess CO2 production but rather represent physiological responses to it. The issue of CO2 insensitivity is related to this limitation. The simple appearance of the new d-plot—featuring a breakpoint rising from the horizontal baseline—may be more amenable to AI-based detection of AT (Zignoli et al., 2022; Tomaszewski et al., 2024; Watanabe et al., 2024). A trade-off exists between AT identification rate and the level of confidence of identifying AT. If the reviewer’s confidence level is high with AT detection (less error), the detection rate of AT falls, but reliability increases. With less confidence (more error), detection rate increases. AT reliability using the CR of d, VS, and Combined were comparable to that reported by Myers et al. (2010), who conducted a multicenter evaluation of AT reliability in 428 patients with heart failure. Some of our cases (Case 4, Case 7) exhibited very early appearance of AT. Gaskill et al. (2001) found that when AT appeared < 4 min into the exercise test, it was considered indeterminate. However, there seems to be no consensus on this point. This phenomenon of very early appearance of AT needs to be further investigated. Our study has limitations. First, the d-method was tested by a single assessor and in only one population, so this method needs to be validated by other investigators in a variety of populations. Second, the d-method detected AT earlier than the VS method; the reason for this requires further investigation. Third, this method should be evaluated against changes in lactate and/or bicarbonate, the original sources of VCO2 changes. Fourth, based on the d-method, we observed delayed or absent AT using the VEQ and ET methods. This finding must be validated in a larger population and by additional investigators. Conclusion The V-slope method is the gold standard for the non-invasive detection of AT. Ventilatory equivalent and end-tidal gas concentration methods have been used simultaneously to rule out nonspecific hyperventilation and also to confirm an AT breakpoint because physiological VCO2 changes (detected by V-slope) should be accompanied by ventilatory volume and gas concentration changes. However, these measures have not historically been plotted on the same x-axis: VCO2 isplotted against VO2, and ventilatory equivalent and end-tidal gas concentrations are plotted against time or work rate. This has made the precise comparison of these measures difficult. Another problem is that using the standard V-slope makes it difficult to detect a breakpoint because of the oblique presentation of the plots. The newly developed d-modification of V-slope method aligns all three parameters on the same time axis. Furthermore, the d plot corrects the oblique presentation of plots by introducing a new horizontal baseline. The direct comparison of the three methods revealed delayed or absent AT by the equivalent/end-tidal methods in 38% of subjects, indicating CO2 chemoreceptor insensitivity. The d AT method showed the best reliability of all four methods and was able to detect AT earlier than the V-slope method, although the exact mechanism for this needs to be further studied. Supplemental Information AT read at the point of slope cahnge The definition of slope change pattern and the results of reading AT based on the slope change are described. Examples of d vs VO2 plot (for reference only) Six cases included in Figure 2 and 3 are presented with both V-slope and d vs VO2 plots. Multiple y-axis datapoints are notable in comparison with d vs time plots (Figure 2 and 3). Clinical and exercise testing raw data of all 127 subjects The data set includes clinical data, cardiopulmonary exercise data, and AT determined by various methods.