Predicting the need for supplemental oxygen during airline flight in patients with chronic pulmonary disease: A comparison of predictive equations and altitude simulation

Abstract

BACKGROUND:

Patients with chronic pulmonary diseases are at increased risk of hypoxemia when travelling by air. Screening guidelines, predictive equations based on ground level measurements and altitude simulation laboratory procedures have been recommended for determining risk but have not been rigorously evaluated and compared.

OBJECTIVES:

To determine the adequacy of screening recommendations that identify patients at risk of hypoxemia at altitude, to evaluate the specificity and sensitivity of published predictive equations, and to analyze other possible predictors of the need for in-flight oxygen.

METHODS:

The charts of 27 consecutive eligible patients referred for hypoxia altitude simulation testing before flight were reviewed. Patients breathed a fraction of inspired oxygen of 0.15 for 20 min. This patient population was compared with the screening recommendations made by six official bodies and compared the partial pressure of arterial oxygen (PaO₂) obtained during altitude simulation with the PaO₂ predicted by 16 published predictive equations.

RESULTS:

Of the 27 subjects, 25% to 33% who were predicted to maintain adequate oxygenation in flight by the British Thoracic Society, Aerospace Medical Association or American Thoracic Society guidelines became hypoxemic during altitude simulation. The 16 predictive equations were markedly inaccurate in predicting the PaO₂ measured during altitude simulation; only one had a positive predictive value of greater than 30%. Regression analysis identified PaO₂ at ground level (r=0.50; P=0.009), diffusion capacity (r=0.56; P=0.05) and per cent forced expiratory volume in 1 s (r=0.57; P=0.009) as having predictive value for hypoxia at altitude.

CONCLUSIONS:

Current screening recommendations for determining which patients require formal assessment of oxygen during flight are inadequate. Predictive equations based on sea level variables provide poor estimates of PaO₂ measured during altitude simulation.

The growing prevalence of chronic pulmonary disease ensures that clinicians will increasingly be called on to evaluate their patients’ fitness to travel by air. Travel by modern airliner exposes patients to the reduced ambient oxygen tensions equivalent to those experienced at 2438 m above sea level for several hours, which can cause significant hypoxemia in individuals with chronic lung disease.

In the physician’s office, determining the impact of high-altitude exposure involves two steps. The first step is to screen the large population of patients with chronic lung diseases for the likelihood of developing hypoxemia during flight, who, therefore, require more careful evaluation. The recommendations made by various official bodies are discrepant on whom to test. The British Thoracic Society (BTS), American Thoracic Society (ATS) and European Respiratory Society (ERS), Canadian Thoracic Society (CTS), the Aerospace Medical Association (AMA) and the Department of Veterans Affairs (VA) have published recommendations for the evaluation of patients at risk (1–5). The BTS recommends a hypoxia altitude simulation test (HAST) for patients with a baseline saturation of peripheral oxygen (SpO₂) of between 92% and 95%, and an additional risk factor (1). Values outside of this range imply safety or the need for supplemental oxygen. The ATS/ERS guidelines and VA guidelines recommend a HAST for chronic obstructive pulmonary disease (COPD) patients with comorbidities, previous in-flight symptoms, recent exacerbation(s), hypoventilation on oxygen administration or borderline estimates from a regression equation (2,5). The AMA and CTS consider a partial pressure of arterial oxygen at ground level (PaO_2gr) of less than 70 mmHg to be an indication of a patient’s need for altitude simulation (3,4).

The clinician’s second task is to decide whether the identified patients are able to tolerate the conditions of air travel. Here also, there are discrepant guidelines concerning the choice of test and how to interpret the results. The HAST artificially reduces inspired oxygen to similar levels experienced at 2438 m for 20 min by either giving a fraction of inspired oxygen of 0.15 or by reducing atmospheric pressure to 565 Torr (75 KPa) in a hypobaric chamber; both methods are considered to be equivalent (6,7). The expected PaO₂ at altitude (PaO_2alt) is then determined from measurements of arterial blood gases (ABGs). The HAST is considered to be the ‘gold standard’; however, opinions differ as to whether a PaO₂ of 55 mmHg or less (AMA) or 50 mmHg (ATS, ERS and VA) warrants the prescription of oxygen for flight. The HAST simulates the maximum drop in pressure expected during air travel, and has been demonstrated to produce oxygenation comparable with that observed during actual air travel (7–9).

Several investigators have developed predictive equations that estimate PaO_2alt using measurements made at sea level (6,10–15); PaO_2gr appears in all equations. Other variables include cabin altitude, inspired oxygen pressure at altitude, partial pressure of arterial carbon dioxide (PaCO₂) at sea level, health status, forced expiratory volume in 1 s (FEV₁) and forced vital capacity (FVC). Guideline documents refer to equations as either a screening tool for recommending a HAST (ATS, ERS), or a replacement for HAST when it is unavailable (BTS, VA).

Finally, some airline medical departments and the AMA recommend the 50 m walk test as a sufficient estimate of fitness to fly (1,4).

The current study aimed to evaluate the latter methods of determining fitness to fly in patients with chronic lung disease. We propose that predictive equations and the 50 m walk test must be both highly specific and sensitive to be useful clinical tools. Our objectives were to determine the adequacy of screening recommendations at identifying patients at risk of hypoxemia at altitude, to evaluate the specificity and sensitivity of published predictive equations compared with the HAST, and to analyze other possible predictors of the need for in-flight oxygen.

METHODS

Charts were reviewed at two university affiliated teaching hospital pulmonary function laboratories that performed altitude simulation testing. Consecutive patients who were not on supplemental oxygen were tested during a 10-month period. The study was approved by the Research Ethics Board of the University Health Network, Toronto, Ontario.

Altitude simulation was performed as previously described (11). For altitude simulation, patients inhaled a fraction of inspired oxygen of 0.15 via a nonrebreathing mask for 20 min with concurrent pulse oximetry. Measurements of ABGs were taken from the radial artery after subjects had inhaled the hypoxic mixture for 20 min whereupon they recovered with supplemental oxygen. Only one centre took baseline ABGs for the HAST. For the other patients, baseline ABGs were taken within three months of the HAST, from outpatient visits only.

PaO_2alt estimates derived from 16 predictive equations (Table 1) were compared with the PaO_2alt measured during the HAST and each equation was evaluated for sensitivity, specificity, positive and negative predictive value, positive and negative likelihood ratio and accuracy that considered a PaO_2alt of 50 mmHg or less a positive result.

TABLE 1.

Predictive equations evaluated

Equation		Author (reference)
1	PaO2alt = 0.410(PaO2gr) + 17.652	Dillard et al (10)
2	PaO2alt = 0.417(PaO2gr) + 17.802	Dillard et al (6)
3	PaO2alt = 22.8 – 2.74*feet + 0.68*PaO2gr	Gong et al (11)
4	PaO2alt = 20.38 – (3*cabinalt) + (0.67*PaO2gr)	Henry et al (12)
5	PaO2 FIO2 0.15 = PaO2 FIO2 0.21*0.54 + 4.7	Seccombe et al (13)
6	PaO2alt = 1.59 + 0.98*PaO2gr + 0.0031*Alt – 0.000061*PaO2gr*Alt – 0.000065*PCO2gr*Alt + 0.000000092*Alt2	Muhm (14)
7	PaO2alt = 46.23 + 3.02*PCO2gr – 0.0038*Alt + 30.57*HS – 0.48*PCO2gr*HS = 0.000064*Alt*HS – 0.05*PCO22	Muhm (14)
8	PaO2alt = 2.14 + 0.97*PaO2gr + 0.006*Alt – 0.000081*PaO2gr*Alt – 0.000076*PCO2gr*Alt + 0.00043*Alt*HS	Muhm (14)
9	PaO2alt = 10.61 + 0.87*PaO2gr + 0.0004*Alt – 0.000033*PaO2gr*Alt – 0.000033*PCO2gr*Alt + 0.000027*Alt*HS	Muhm (14)
10	PaO2alt = 0.519 × PaO2gr + 11.855 × FEV1 (L) – 1.760	Dillard et al (10)
11	PaO2alt = 0.453 × PaO2gr + 0.386 × FEV1 (% predicted) + 2.44	Dillard et al (10)
12	PaO2alt = 0.294(PaO2gr) + 0.086(FEV1%) +23.211	Dillard et al (6)
13	PaO2alt = 0.245(PaO2gr) + 0.171(FEV1/FVC%) + 21.028	Dillard et al (6)
14	PaO2alt = 0.238(PaO2gr) + 20.098(FEV1/FVC) + 22.258	Dillard et al (6)
15	PaO2alt = [PaO2gr] [(−1)(0.02002-0.00976FEV1)(PIO2gr – PIO2alt)]	Dillard et al (15)
16	PaO2alt = [PaO2gr] [(−1)(0.01731-0.00019FEV1%)(PIO2gr – PIO2alt)]	Dillard et al (15)

The equation number is used to identify each equation within the text. Alt Altitude; cabin_alt Cabin pressure at altitude; HS Pulmonary health status: 1 = normal, −1 = chronic obstructive pulmonary disease; FEV₁ Forced expiratory volume in 1 s; FIO₂ Fraction of inspired oxygen; FVC Forced vital capacity; PaO_2alt Partial pressure of arterial oxygen at altitude; PCO₂ Partial pressure of carbon dioxide at sea level; PIO_2alt Inspired oxygen pressure at altitude; PaO_2gr Partial pressure of arterial oxygen at ground level

If patients in our population did not meet the criteria specified by the equation’s authors, the equation was evaluated in the subset of our population who did meet the criteria. The following inclusion criteria relate to equations 3, 4, 5, 15 and 16 respectively (Table 1): FEV₁ less than 80% of predicted and FEV₁/FVC ratio less than 70%; COPD; PaO₂ greater than 70 mmHg, SpO₂ greater than or equal to 95%; and FEV₁ greater than 60% of predicted.

Data from a 6 min walk test was obtained within three months of the HAST to evaluate the airline criteria of a successful 50 m walk for determining fitness to fly. The inclusiveness of the various guidelines was evaluated by determining how many patients would not have been tested according to each guideline, and of those, how many were hypoxemic after undergoing a HAST.

Receiver operating characteristic (ROC) curves of both the PaO₂ and the SpO₂ at sea level were plotted to evaluate fitness to fly cut-off values. Finally, a multiple stepwise regression analysis was performed to determine if any values of FEV₁, FVC, FEV₁/FVC, residual volume or diffusion capacity (DLCO) correlated strongly with PaO₂ after HAST. Results were considered significant at P<0.05.

RESULTS

The patients (n=27) had typical chronic lung disease of moderate severity. Three patients had cystic fibrosis, 22 had COPD, and two had cystic fibrosis and COPD. Demographic data are presented in Table 2. After a HAST, four patients had a PaO₂ value of less than 50 mmHg and six had values of between 55 mmHg and 50 mmHg. The 50 mmHg cut-off was used for the purpose of the present paper; however, all 10 of these patients were advised to consider in-flight supplemental oxygen.

TABLE 2.

Demographic data

Characteristic	Mean ± SD	n
Age, years	64±18	27
Men, %	57	27
Height, cm	163±10	25
Weight, kg	70±16	25
Chronic obstructive pulmonary disease, %	93	27
FVC, L	2.48±1.02	20
FVC, % predicted	76±25	20
FEV1, L	1.41±0.71	20
FEV1, % predicted	58±22	20
FEV1/FVC	58±17	20
FEV1/FVC, % predicted	79±23	20
Residual volume, L	2.01±1.16	14
Diffusion capacity, mL/min/mmHg	12.19±5.73	13
Baseline blood gases
SpO2, %	95±3	27
PaCO2, mmHg	38±5	27
PaO2, mmHg	75±10	27
Post-HAST blood gases
SpO2, %	89±3	27
PCO2, mmHg	37±4	27
PO2, mmHg	58±7	27

FEV₁ Forced expiratory volume in 1 s; FVC Forced vital capacity; HAST Hypoxia altitude simulation test; PaCO₂ Partial pressure of arterial carbon dioxide at sea level; PCO₂ Partial pressure of carbon dioxide; PaO₂ Partial pressure of arterial oxygen; PO₂ Partial pressure of oxygen; SpO₂ Baseline saturation of peripheral oxygen

Table 3 presents an evaluation of each equation that was tested. Equations showed better or equivalent results when tested on their intended population than with the entire sample set. Equations 1 to 9 underestimated the PaO_2alt, leading to high sensitivity but low specificity. In contrast, equations 15 and 16 overestimated the PaO_2alt, resulting in high specificity but low sensitivity. These two equations did not identify any patients as needing in-flight oxygen. Equations 10 and 14 were low in both sensitivity and specificity.

TABLE 3.

Equation and 50 m walk test evaluation

Equation	Sensitivity	Specificity	Positive predictive value	Negative predictive value	Positive likelihood ratio	Negative likelihood ratio	Accuracy	Area under ROC curve	n
1	1.00	0.32	0.22	1.00	1.47	0.00	0.48	0.66	27
2	1.00	0.44	0.25	1.00	1.79	0.00	0.56	0.72	27
3	1.00	0.57	0.31	1.00	2.33	0.00	0.67	0.78	27
3*	1.00	0.55	0.17	1.00	2.20	0.00	0.58	0.77	12
4	1.00	0.32	0.22	1.00	1.47	0.00	0.48	0.66	27
4*	1.00	0.41	0.07	0.41	1.69	0.00	0.43	0.71	23
5	1.00	0.07	0.16	1.00	1.08	0.00	0.22	0.53	27
5*	1.00	0.13	0.07	1.00	1.14	0.00	0.18	0.56	17
6	1.00	0.32	0.22	1.00	1.47	0.00	0.48	0.66	27
7	1.00	0.07	0.16	1.00	1.08	0.00	0.22	0.53	27
8	1.00	0.00	0.15	1.00	1.00	0.00	0.15	0.50	27
9	1.00	0.38	0.24	1.00	1.61	0.00	0.52	0.69	27
10	0.25	0.50	0.11	0.73	0.50	1.50	0.45	0.38	20
11	0.00	0.88	0.00	0.78	0.00	1.14	0.70	0.44	20
12	0.25	0.25	0.80	0.57	0.33	3.00	0.25	0.25	20
13	0.00	0.63	0.00	0.72	0.00	1.59	0.50	0.31	20
14	0.50	0.50	0.20	0.80	1.00	1.00	0.50	0.50	20
15	0.00	1.00	0.00	0.80	0.00	1.00	0.80	0.50	20
15*	0.00	1.00	0.00	0.43	0.00	0.00	0.43	0.50	7
16	0.00	1.00	0.00	0.80	0.00	1.00	0.80	0.50	20
16*	0.00	1.00	0.00	0.43	0.00	0.00	0.43	0.50	7

A cut-off value of partial pressure of arterial oxygen (PaO₂) ≤ 50 mmHg was used to define positive cases.

^*Each equation was tested on the full sample as well as the portion of the sample that fit the inclusion criteria of the original author, if applicable. ROC Receiver operating characteristic

The 11 patients who performed a 6 min walk test covered (mean ± SD) 359±101 m, all passing the airline requirements for fitness to fly. Evaluations based on SpO₂ during the 6 min walk test did not improve this test’s predictive value. Five patients had an SpO₂ below 85% while walking, but of these, only two were identified by the HAST as needing supplemental in-flight oxygen. One interesting trend was that the four patients for whom the HAST recommended supplemental oxygen walked the shortest distance in the patient population (330 m or less in 6 min).

Of the 12 patients who the BTS considered fit to fly (SpO₂ greater than 95%), three had a PaO₂ of less than 50 mmHg after the HAST and one had a PaO₂ of 52 mmHg. Four of the 16 subjects presumed fit to fly by the AMA (PaO_2gr of 70 mmHg or higher) also benefited from the HAST (PaO_2alt 55 mmHg or less), with one below 50 mmHg. Thus, 25% to 33% of the population presumed fit to fly required further investigation. Figure 1 shows the drop in PaO₂ experienced by each of these six patients between room air and the HAST. Diagnosis of these patients revealed that three had COPD, two had cystic fibrosis and one patient had both. The drops in PaO₂ observed in the two cystic fibrosis patients were not distinguishable from those of the three COPD patients.

Figure 1).

Partial pressure of arterial oxygen (PaO₂) drop in six patients whom guidelines considered fit to fly. The patients were considered fit to fly by at least one of the criteria set by the British Thoracic Society (BTS; baseline saturation of peripheral oxygen [SpO₂] >95% and the Aerospace Medical Association (AMA; PaO₂ at ground level ≥70 mmHg). Three patients had a posthypoxia altitude simulation test (HAST) PaO₂ of less than 50 mmHg, while the other three subjects had a PaO₂ of between 50 mmHg and 55 mmHg

ROC curves are shown in Figure 2. Baseline PaO₂, as validated against HAST outcome, gave an area under the ROC curve of 0.696±0.095 (P<0.13) and suggested that the cut-off yielding greatest accuracy was a PaO₂ of 72 mmHg or less, at which point sensitivity was 1.00 and specificity was 0.61. Baseline SpO_2, as validated against HAST outcome, gave an area under the ROC curve of 0.402±0.112 (P<0.55) and suggested a cut-off value for greatest accuracy of 96%, at which point sensitivity was 1.00 and specificity was 0.74. Each equation was also subjected to ROC analysis in comparison with the HAST results (Table 3). Based on the area under the ROC curve, equation 3 was the most accurate predictor evaluated but had poor overall predictive characteristics.

Figure 2).

Receiver operating characteristic (ROC) analysis of baseline partial pressure of arterial oxygen (PaO₂) and resting baseline saturation of peripheral oxygen (SpO₂) related to the hypoxia altitiude simulation test (HAST) outcome (n=27). (A) ROC curve of PaO₂ at ground level and HAST outcome gave an area under the curve of 0.696 (P=0.13). At a specificity of 100, PaO₂ ≤72 mmHg. (B) ROC analysis of SpO₂ at rest related to HAST outcome, gave an area under the curve of 0.402 and a SpO₂ cut-off for recommending a HAST of 95% or lower

Regression analysis determined three variables that correlated with the ratio of PaO_2alt to PaO_2gr (r=0.45; P<0.009), DLCO (r=0.56; P<0.05) and FEV₁% predicted (r=0.57; P<0.009). All other variables tested (age, FVC, FEV₁/FVC and residual volume) showed no statistically significant correlations with PaO_2alt.

After controlling for PaO_2gr, DLCO remained significantly and independently associated with PaO_2alt (r=0.60; P<0.04), as did FEV₁% predicted (r=–0.48; P<0.04). Because the sample size allowed for the examination of two-variable interactions, the following new prediction equations were derived:

$${\text{PaO}}_{2\text{alt}}=0.41\hspace{0.17em}({\text{PaO}}_{2\text{gr}})+0.54\hspace{0.17em}(\text{DLCO})\hspace{0.17em}+\hspace{0.17em}22.08$$

$${\text{PaO}}_{2\text{alt}}=0.34 ({\text{PaO}}_{2\text{gr}})-0.05 ({\text{FEV}}_1\%) + 24.22$$

DISCUSSION

Our data suggest that current guidelines concerning the need for detailed assessment of oxygen requirements at altitude need review. In the patients tested, neither SpO₂ or PaO_2gr was sufficiently sensitive or specific to determine the patients most in need of altitude simulation or further evaluation. Our results, like those of Christensen et al (16), show that using cut-offs to identify patients who are fit to fly without further examination misses up to one-third of individuals who will desaturate to a PaO₂ of less than 50 mmHg. Furthermore, equations are poor predictors of PaO₂ measured during altitude simulation procedures and, thus, are an unreliable means of determining a patient’s fitness to fly. Because no individual equation could consistently distinguish patients who needed oxygen from those who did not, we recommend against their use in preflight evaluation. Concern over this practice has been raised previously by authors reporting poor reproducibility of results (17–20). These equations have been accepted as interchangeable with other methods, despite the lack of validation. We suggest that the HAST should be used to evaluate all patients suspected to be at risk of becoming hypoxemic at altitude. Using the equations as a criteria for HAST selection is also a risky practice because our results showed that their predictions depended largely on which equation was used rather than on patient characteristics.

We demonstrated a correlation between FEV₁ and PaO_2alt. This second-order correlation has been reported several times (6,10,15) but the finding has not been consistently reproduced (21). Interestingly, our study suggests that DLCO and FEV₁ are better predictors of PaO_2alt than PaO_2gr. This finding is consistent with the predictors of arterial desaturation in COPD patients during exercise (22).

We also found the 50 m walk test to be an unreliable predictor of fitness to fly because all of the evaluated patients who had available walk test data greatly exceeded this distance. We did find, however, that the patients in our pool who walked 330 m or less during the 6 min walk test all had a PaO₂ of less than 50 mmHg after the HAST.

Several questions remain regarding the generalizability of the HAST results. Because the simulation lasts only 20 min, it is unclear whether a further drop in PaO₂ can be expected on longer flights. While there is some evidence that this does not occur, patients may compensate for the reduced oxygen by increasing their respiratory rate (18). This could possibly become fatiguing in longer flights, resulting in a delayed decrease in respiratory rate and, consequently, a further drop in PaO₂.

There is a need for more evidence on which to base guidelines in this field. Current guidelines vary, and among individual clinicians, the practice is even more diverse (23). For example, PaO₂ in flight is considered safe if it is greater than 50 mmHg. This value was reported in the paper that first described the HAST test (11), and was subsequently incorporated into guideline documents and other publications (1,6,17,24,25). We find this value to be substantially lower than necessary, and speculate that patients with a PaO₂ of between 50 mmHg and 55 mmHg may also benefit from in-flight oxygen.

Our ROC analysis of PaO₂ and SpO₂ criteria for HAST produced mixed results. Although both cut-offs coincided with guidelines, these findings were not statistically significant, and the areas under the ROC curves indicated poor predictive ability of these variables. We suggest more analyses such as these be performed to form a solid evidence base from which to draw guidelines.

Some limitations to our study must be noted. First, our sample size was small, reflecting the infrequent use of altitude simulation procedures to assess fitness to fly. We doubt that a larger sample size would significantly alter our major conclusions given the poor performance of screening recommendations and predictive equations. Second, our patients’ referral for this test indicates they were at a geographical advantage and had time to perform the test before travel; consequently, their results may only approximate those of the larger population. Although it is possible that they were at particular risk of hypoxemia in flight, this was not evident from any demographic, clinical or physiological characteristics at baseline, and the population was sufficiently diverse in HAST outcome to test the screening recommendation and predictive equations. Third, although the HAST is the gold standard for estimating in-flight hypoxia, it does not take into account the effects of additional stress and exercise performed at altitude. Finally, our population was not limited to patients with COPD but also included patients with restrictive processes. However, when these were excluded, our conclusions remained unchanged. Avenues for future research include consideration of underlying pulmonary diagnosis as a baseline variable in the clinical outcomes among disease entities, and the implications of increasing the use of the HAST from a cost perspective.

SUMMARY

Current screening recommendations do not identify all patients at significant risk for hypoxemia at altitude and predictive equations based on ground level variables cannot replace altitude simulation procedures when evaluating patient need for oxygen supplementation during airline flight.