Effect of the definition of hypopnea on apnea/hypopnea index

C. A. Nigro, E. E. Rhodius

Laboratorio de Sueño, Servicio de Neumonología, Hospital Alemán, Buenos Aires

Direccion Postal: Carlos Nigro, Cachimayo 333 4º P, 1424,  Buenos Aires, Argentina.
Fax: (54-11) 4827-7000, int.2879. E-mail: cnigro@intramed.net

Abstract
The objective of this study was to determine whether different decreases in oxygen saturation (SaO₂) or the presence of electroencephalographic arousals (EEGA) in the definition of hypopnea modify hypopnea index and apnea/hypopnea index and the prevalence of obstructive sleep apnea/hypopnea syndrome (OSAHS). A total of 20 polysomnographies performed in patients with OSAHS were analyzed. There are four different definitions of hypopnea: > 30% reduction in airflow or 50% decrease in abdominal movement associated with ¯SaO₂ > 3% (type 1); ¯ SaO₂ > 3% or EEGA (type 2); ¯ SaO₂ > 4% (type 3); ¯ SaO₂ > 4% or EEGA (type 4). The prevalence of OSAHS was calculated for an apnea/hypopnea index (AHI) > 10 and > 15. Hypopnea index (HI) and AHI types 2 and 4 were higher than type 3 (HI: type 2: 20±10.6, type 4: 18.6±10, type 3: 11.4±10, p < 0.001; AHI: type 2: 23.3±11.6, type 4: 21.4±11.2, type 3: 14.7±11.6, p < 0.001). No differences were observed between HI and AHI types 1 and 2 (HI: type 1: 17.4±10, type 2: 20±10.6; AHI: type 1: 20.6±11.8, type 2: 23±11.6, p > 0.05). The prevalence of OSAHS was 30-55% in type 3, 70-85% in type 4 (p < 0.05), and 70-85% in types 1 and 2 (p > 0.05). In our patient´s population, the presence of EEGA in the definition of hypopnea significantly increased the HI, the AHI and the prevalence of OSAHS when associated with a > 4% decrease in SaO₂.

Key words: Diagnosis; Hypopnea; Sleep apnea syndrome.

Resumen
Influencia de la definición de hipopnea sobre el índice apnea/hipopnea. El objetivo del estudio fue evaluar si diferentes niveles de descenso de la saturación de O₂ (SaO₂) o la presencia  de micro-despertares electroencefalográficos (MDEEG) en la definición de hipopnea, modifican el índice de apnea/hipopnea y la prevalencia del síndrome apnea/hipopnea del sueño (SAHS). Se analizaron 20 polisomnografias de pacientes con SAHS. Hipopnea (H) se definió de cuatro formas: descenso del flujo aéreo >30% o caída del 50% del movimiento abdominal asociado a: ¯ SaO2  > 3% (tipo 1) ; ¯ SaO2 > 3% o MDEEG (tipo 2); ¯ SaO2 > 4% (tipo 3); ¯ SaO2 > 4% o MDEEG (tipo 4). La prevalencia del SAHS se calculó para un índice apnea/hipopnea (IAH) > 10 y > 15. El índice hipopnea (IH), el IAH tipo 2 y 4 fue mayor respecto al tipo 3 (IH: tipo 2: 20±10.6, tipo 4: 18.6±10, tipo 3: 11.4±10, p < 0,001; IAH: tipo 2: 23.3±11.6, tipo 4: 21.4±11.2, tipo 3: 14.7±11.6, p < 0.001). No hubo diferencias entre el IH y el IAH tipo 1 y 2 (IH: tipo1: 17.4±10, tipo 2: 20±10.6; IAH: tipo1: 20.6±11.8, tipo 2: 23±11.6, p > 0.05). La prevalencia de SAHS fue del 30-55% con el tipo 3, 70-85% con el tipo 4 (p < 0.05)  y del 70-85% con los tipos 1 y 2.(p > 0.05). En nuestra población de pacientes, la presencia de MDEEG en la definición de hipopnea aumentó significativamente el IH, el IAH y la prevalencia de SAHS cuando se asoció a un descenso de la SaO₂ > 4%.

Palabras clave: Diagnóstico; Hipopnea; Síndrome apnea del sueño.

Obstructive sleep apnea/hypopnea syndrome (OSAHS) is diagnosed due to the presence of frequent episodes of apnea/hypopnea during sleep, and is associated with symptoms such as excessive daytime sleepiness or cognitive disorder^{8, 13}.
Although apnea has been defined in the literature as the absence of oronasal airflow > 10 sec, controversy exists about the definition of hypopnea. The criteria used for the identification of hypopneas, such as magnitude of the decrease in airflow, level of oxygen desaturation and presence of electroencephalographic arousals differ considerably in the literature^2,10,11. Gould² demonstrated that a 50% fall in thoracoabdominal movements examined with calibrated inductance plethysmography better correlated with arousals and oxygen desaturation than 25% or 75% reductions in the dimension of respiratory movements. In a study carried out by 44 accredited sleep disorder centers in the United States¹, in addition to the reduction in ventilation, a decrease in oxygen saturation (SaO₂) was included as a criterion of hypopnea by 82% of the centers. The decrease in SaO₂ was higher than 4% in 30% of the centers, higher than 2% and lower than or equal to 4% in 22%, and any degree of oxygen desaturation was used in the remaining 42%. The presence of an electroencephalographic arousal associated with the end of hypopnea was included by 75% of the laboratories. These differences could play a significant role in the diagnosis of OSAHS and in the comparison of the results obtained from epidemiological studies.
Little data exist regarding the impact of different definitions of hypopnea on the apnea/hypopnea index (AHI). Using different levels of SaO₂ fall (0% to > 5%), Redline⁹ found an 11-fold difference in the median of AHI and a 16-fold difference in the prevalence of OSAHS. Tsai⁷ demonstrated that the addition of electroencephalographic arousals (EEGA) to the decrease in SaO₂ in the definition of hypopnea brought about no significant changes in AHI, but it did in the prevalence of OSAHS.
The objective of this study was to determine whether the use of EEGA as an alternative criterion of the decrease in SaO₂ in the definition of hypopnea modifies the HI, the AHI and the prevalence of OSAHS.

Material and Methods

Study subjects and design

Polysomnographies performed in patients with suspicion of OSAHS, who had been filed in a database from 12/01/98 to 07/31/00, were revised by a physician from the Sleep Laboratory in the Hospital Alemán. Patients selected for this study should meet the following inclusion criteria: 1) more than 50% of hypopneas upon respiratory compression analysis (oronasal airflow, abdominal movement and SaO₂), and 2) a proper signal for all the recorded variables. The criteria for exclusion from the study included: patients < 18 years of age and clinical diseases that might result in oxygen desaturacion and sleep respiratory alterations during the sleep, such as neuromuscular disease, severe chronic obstructive pulmonary disease (FEV1 < 50% of the baseline value), severe coronary disease (unstable angina, myocardial infarction or aortocoronary bypass in the previous 6 months), chronic sleeplessness, restless legs syndrome, psychiatric disorders and use of tranquilizers. A group of 20 patients were recruited. The present study was approved by the Hospital Alemán bioethics committee.

Methods

Polysomnographies were performed with an AKONIC version 8.0 computing polysomnographer. The following variables were registered: four channels of electroencephalogram (EEG: C3, C4, O1, O2), bilateral electrooculogram (EOG), chin electromyogram (chin-EMG), right and left anterior tibialis surface electromyogram (LAT AND EMG), electrocardiogram (EKG), oronasal airflow measured with thermistors, abdominal movements measured with a pressure transductor, and SaO₂ measured with a finger sensor. SaO₂ was measured with a NOVAMETRIX 505 oximeter, and an average frequency of 4 sec was used. Sleep stages were identified based on international guidelines⁴. We used the criteria of the American Sleep Disorders Association (ASDA) for EEGA⁵. Apnea was defined as the absence of oronasal airflow for > 10 sec. Hypopneas (H) were defined according to four different types: a discernible decrease in oronasal airflow or a 50% reduction in abdominal movements associated with one of the following criteria: a) a 3% reduction in SaO₂ (type 1), b) a 3% decrease in SaO₂ or EEGA (type 2), c) a > 4% reduction in SaO₂ (type 3), d) a > 4% decrease in SaO₂ or EEGA (type 4). A discernible decrease in airflow  was represented by a > 30% fall in thermistor signal amplitude. AHI was defined as the total number of apneas plus hypopneas divided by the total sleep time. Hypopnea index (HI) was defined as the total number of hypopneas divided by the total sleep time.
Polysomnographies were analysed by a skilled physician other than the observer who selected the tracings. The following algorithm was used for the classification of hypopneas: 1) identification of the decrease in ventilation; 2) identification of the highest decrease in SaO₂ (either 3% or > 4%). The latter was achieved given that the program constantly shows SaO2 values on the screen, thus allowing the operator to identify 1%, 2%, 3% or >4% falls in SaO₂; 3) the presence or absence of an EEGA associated with the end of hypopnea was determined. These data were used to design a table that allowed determining the number of hypopneas according to each definition.

Statistical analysis

A descriptive analysis was performed. AHI and HI were calculated according to each definition of hypopnea. Analysis of variance was used for repeated measurements and the Scheffe test for the differences in HI and AHI according to the different definitions of hypopnea. Two cut-off values were used as diagnostic criteria of OSAHS: a > 10 AHI and a > 15 AHI. Mc Nemar test was used in order to determine the differences in the prevalence of OSAHS between hypopnea type 1 vs. 2 and type 3 vs. 4. A p < 0.05 significance level was used. The strength of association between polysomnographic indexes based on different criteria of hypopnea was determined  by the intraclass correlation coefficient (ICC). Considering a relatively high correlation does not necessarily imply good agreement, Bland-Altman plots were constructed in order to assess the extent of agreement¹⁴.

Results

Table 1 summarizes patients’ anthropometric characteristics and the results of polysomnographies. According to the definition used, hypopneas accounted for 78% to 85% of the whole respiratory events (apneas plus hypopneas) (Table 2). The addition of EEGA to the definition contributed to the identification of 13% and 53% of hypopneas (type 1: 1572 vs. type 2: 1779; type 3: 1053 vs. type 4: 1608).

The ICCs were calculated for paired comparisons between the different  HIs and AHIs derived from four methods for scoring hypopneas. The ICC of HI between method 1 and method 2 was 0.91 (95% CI 0.78 to 0.96). The ICC of HI between method 3 and method 4 was 0.79 (95% CI 0.53 to 0.91). The ICC of AHI between method 1 and method 2 was 0.93 (95% CI 0.82 to 0.97). The ICC of AHI between method 3 and method 4 was 0.83 (95% CI 0.62 to 0.93). Figure 1 and 2 are Bland- Altman plots of  the agreement between HI  type 1 / HI type 2 and  AHI type 3 / AHI  type 4. The mean difference between HI type 1 / HI type 2 and AHI type 1 / AHI type 2 were -2.6±9/h (mean±2SD). Notwithstanding, substancial differences between HI type 3 / HI type 4 and AHI type 3 / AHI type 4 were observed (-6.75±13.3/h, mean±2SD).

Table 3 shows the different values of AHI and HI, according to the different definitions of hypopnea. HI and AHI types 2 and 4 were higher than type 3 (HI type 2 = 20 ± 11.6, HI type 4 = 18.6 ± 10 vs. HI type 3 = 11.4 ± 10, p < 0.001; AHI type 2 = 23.3 ± 11.6, AHI type 4 = 21.4 ± 11.2 vs. AHI type 3 = 14.7 ± 11.6, p < 0.001). HI and AHI types 1 and 2 were similar (HI type 1 = 17.4 ± 10, AHI type 1 = 20.6 ± 11.8 vs. HI type 2 = 20 ± 10.6, AHI type 2 = 23 ± 11.6, p > 0.05).

Table 4 shows the prevalence of OSAHS according to both AHI cut-off value and the definition of hypopnea. The diagnostic frequency of OSAHS in the population herein studied was 30% to 55% when a > 4% decrease in SaO₂ was used as criterion of hypopnea, and 70% to 85% when a > 4% fall in SaO₂ and/or EEGA was used (p < 0.05). Definitions 1 and 2 showed no differences in the frequency of OSAHS (type 1: 70% to 85%, type 2: 70% to 85%, p > 0.05).

Discussion

In the present study we compared apnea/hypopnea indices according to four different definitions of hypopnea: 1) type 1: ¯ SaO₂ 3%; 2) type 2: ¯ SaO₂ 3% or EEGA; 3) type 3: ¯ SaO₂ > 4%; 4) type 4: ¯ SaO₂ > 4% or EEGA. Results show that when the definitions including EEGA (H type 2 and 4) are used, the average number of hypopneas, as well as HI and AHI are significantly higher than the same variables in type 3. HI and AHI type 2 were 76% and 59% higher than HI and AHI type 3. Moreover, HI and AHI type 4 were 64% and 46% higher than HI and AHI type 3.
Our results reveal that HI or AHI type 1 and HI or AHI type 2 show a high level of correlation and agreement (ICC 0.9 to 0.93, mean diffrerence between HI /AHI type 1 and HI /AHI type 2 = - 2.6±9/h, Fig. 1). Therefore, the addition of an arousal requirement to the oxygen desaturation criterion for hypopnea causes only a small change in HI and AHI. On the other hand, the correlation and agreement between HI or AHI type 3 and HI or AHI type 4 was poor, showing  that a definition of hypopnea requiring an arousal associated to oxygen desaturation yields substancially different results than one based solely on oxygen desaturation (ICC 0.78 to 0.83, mean diffrerence between HI or AHI type 3 and HI or AHI type 4 = - 6.75±13.3/h, Fig. 2). Indeed, a statistically significant difference was found between the mean values of HI or AHI type 3 and 4 (p < 0.001).
The above mentioned definitions of hypopnea were chosen for several reasons. The "discernible" decrease in airflow was selected due to observations indicating that small decreases in thermistor signal amplitude could be associated with up to 50% reductions in the current volume measured with a pneumotachograph⁶. A > 50% fall in abdominal movements was herein included as an alternative criterion of ventilation reduction. The cut-off value chosen for this study has been largely reported in the literature⁷ and allows a higher correlation with oxygen desaturation and arousals². We chose either the 3% or > 4% SaO₂ decreases, since they are the most frequently used in the different sleep laboratories with a variability higher than the 2% shown by pulse oximetry.
The findings herein reported are similar from the study by Tsai et al⁷, who found a HI type B higher than HI type A (HI type B = 14, HI type A = 9, p < 0.004) in a group of 98 patients with OSAHS. These authors defined hypopnea as > 30% decrease in the dimension of thoracic or abdominal movements > 10 seg. associated with one of the following parameters: 1) type A: ¯ SaO₂ > 4%; 2) type B: ¯ SaO₂ > 4% or EEGA. However, the AHI  type B and type A were similar  in both definitions (AHI type A = 16, AHI type A = 19, p > 0.05). These differences could be explained because  the population of patients studied by Tsai, presented a lower rate of hypopneas (57% - 75%) than our population (78% - 85%). Thus, although HI in definition "type B" was higher than in "type A", AHI was similar in both definitions, due to a "dilution effect" derived from the higher rate of apneas in their population. Similarly to us, Redline et al⁹ revealed a 40% increase in AHI in hypopneas identified by oxygen desaturation or electroencephalographic arousals compared with definitions that only required SaO₂ fall. On the other hand, Manser et al¹² observed in a population of 48 patients with predominatly hypopneas, an increase in HI and AHÍ of 24 to 35 and 30 to 42 (p < 0.0005) when he used as criteria of hypopnea a decrease in the oxygen saturation > 3% vs oxygen desaturation or arousal. By contrast, we did not observe significant differences between HI type 1 (¯ SpO₂ > 3%) and type 2 (¯ SpO₂ > 3% or EEGA). These differences  could be explained by several reasons: 1) Although we observed a trend towards a higher HI type 2 than HI type 1 (HI type 2 = 20, HI type 1 = 17), it did not reach statistical significance. This possibly  was because of the few number of patients we studied (type II error) 2) The methodology we used to detect descent of airflow (thermistors) may underestimate the number of hypopneas when it is compared with the respitrace¹⁵ 3) Finally, we used an airflow reduction to define hypopnea higher (¯ > 50% AM) than Manser (¯ > 30% respitrace) therefore we may have failed to detect some hypopneas. In table 5 we compare our results with those of Tsai and Manser.

The prevalence of OSAHS depends on both the cut-off value used in AHI⁸ and the definition of hypopnea⁹. Although the prevalence of OSAHS with the definition type 2 was higher than type 1 (85% and 70%), it did not reach statistical significance. This was probably due to the smaller number of patients  in our study compared to Manser´s paper. Nevertheless, we found that the prevalence of OSAHS using a cut-off value of AHI > 10 increased from 55% (type 3 - ¯ SaO₂ > 4%) to 85% (type 4  - ¯ SaO₂ > 4% or EEGA), and using AHI > 15, from 30% to 70% (type 3 and 4, respectively, p < 0.05).
We conclude that, in our patients´ population with high frequency of hypopneas, the use of EEGA as an alternative criterion of a > 4% decrease in SaO₂ in the definition of hypopnea, significantly increased the number of identified hypopneas, HI, AHI and the prevalence of OSAHS.

Acknowledgements: The authors thank Silvia Aimaretti for her technical assistance with polysomnographies and Fundación René Barón for the translation of the manuscript.

References

1. Moser NJ, Phillips BA, Berry DTR, Harbison L. What is hypopnea, anyway? Chest 1994; 105: 426-8.         [ Links ]
2. Gould GA, Whyte KF, Rhind GB, et al. The sleep hypop-nea syndrome. Am Rev Respir Dis 1988; 137: 895-8.         [ Links ]
3. Whyte KF, Gugger M, Gould GA, Molloy J, Wraith PK,  Douglas NJ. Accuracy of respiratory inductive plethys-mography in measuring tidal volume during sleep. J Appl Physiol 1991; 71: 1866-71.         [ Links ]
4. Rechtstaffen A, Kales A. A manual of standardized terminology, techniques and scoring system for sleep stages of human subjects. Los Angeles: Brain Information Institute, University of California, 1968.         [ Links ]
5. EEG arousals: scoring rules and examples. A preliminary report from the sleep disorders. Atlas Task Force of the American Sleep Disorders Association. Sleep 1992; 15: 174-84.         [ Links ]
6. Nigro C, Dibur E, Gallego C, Gorogod C, Rhodius E. Thermistors accuracy for detection of changes in tidal volume. Eur Respir J 1998; 12: 394s.         [ Links ]
7. Tsai WH, Ward Flemons WW, Whitelaw WA, Remmers JE. A comparison of apnea-hypopnea indices derived from different definitions of hypopnea. Am J Respir Crit Care Med 1999; 159: 43-8.         [ Links ]
8. Young T, Palta M, Demsey J, Skatrud J, Weber S, Badr S. The occurrence of sleep disordered breathing among middle aged adults. N Engl J Med 1993; 328: 1230 -5.         [ Links ]
9. Redline S, Bonekat W, Gottlieb D. et al. Variation in the apnea hypopnea index (AHI) according to hypopnea definition. Am J Respir Crit Care Med 1997; 155: A128.         [ Links ]
10. Berry DTR, Phillips BA. Sleep disordered breathing in the elderly: review and methodologic comment. Clin Psych Rev 1988; 8: 101-20.         [ Links ]
11. Thorpy MJ . Glossary of terms used in sleep disorders medicine. In: Handbook of sleep disorders. New York: Marcel Decker, 1990; pp. 779-95.         [ Links ]
12. Manser RL, Rochford P, Pierve RJ, Byrnes GB, Campbell DA. Impact of different criteria for defining hypopneas in the apnea-hypopnea index. Chest 2001; 120: 909-14.         [ Links ]
13. Sala H, Nigro C, Rabec C, Guardia A, Smurra M. Consenso Argentino de  Trastornos  Respiratorios Vinculados al Sueño. Medicina (Buenos Aires) 2001; 61: 351-63.         [ Links ]
14. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical mea-surement. Lancet 1986; 1: 307-10.         [ Links ]
15. Series F, Marc I. Nasal pressure recording in the diagnosis of sleep apnea hypopnea syndrome. Thorax 1999; 54: 506-10.        [ Links ]

Received: 18 de diciembre 2001
Accepted: 18 de febrero 2003