Geographical areas above the sea level are associated to adverse conditions for the Life. For example, although the oxygen concentration remains constant with the trophosphere, the barometric pressure falls, which determines a condition called hypobaric hypoxemia. Indeed, the humidity and temperature, factors that may contribute to airway reactivity, insensible water losses, ventilatory changes and alterations in pulmonary hemodynamics,1 also decrease in these areas. Fast ascend to elevated areas is associated to well know entities such us mountain sickness, High Altitude Pulmonary Edema (HAPE), High Altitude Cerebral Edema (HACE) and cardiovascular collapse.2 However, those who stay for a long period of time above the sea level develop mechanisms to compensate this adverse environment; therefore previously mentioned entities are unusual. Acclimatization to altitude is a term use to describe physiological changes that occur over several weeks with the aim to increase the endurance against the deleterious effect of staying above the sea level.3–5

Invasive mechanical ventilation is the cornerstone intervention for supporting severe respiratory failure. Unfortunately, despite 385 million people permanently residing at elevations above 1500m6 and adverse effects of chronic exposure to high altitude are well demonstrated3,5,7; the knowledge regarding the effect of altitude above the sea level in acclimatized patients under invasive mechanical ventilation is scarce. Furthermore, several worldwide surveys regarding mechanical ventilation have been performed,8–11 but none of them have considered the effect of altitude above the sea level. It is paramount importance to highlight that mechanical ventilators are non pressurized systems and they function at the same atmospheric pressure as the environment. This mean that, despite the patient is under invasive mechanical ventilation, their alveolar pressure and their arterial pressure is lower than at sea level for the same inspiratory pressure.

On the other hand, with the aim to maintain the biological significance of one variable (e.g. diagnostic criteria for a disease or prognosis factor) sometimes it is necessary to adjust their value according to another variable. For example, it is possible to use the partial oxygen pressure (PaO2) as a surrogate biomarker of the gas exchange if everybody receives the same oxygen inspiratory fraction (FiO2). However, as it is possible to prescribe different FiO2, it is common to relativize the PaO2 value according to the FiO2 (PaO2/FiO2). Partial oxygen pressure is also dramatically influenced by the atmospheric pressure, thus in high altitudes above the sea level the biological significance of a specific PaO2/FiO2 could differ than at sea level. A mathematical equation has been proposed to adjust the PaO2/FiO2 according to the altitude (PaO2/FiO2*[barometric pressure/760]).12,13 However, as far as we know, it is an empiric equation and none have validate it.

Studying the effect of the altitude above the sea level is not merely an academic exercise as, if the outcome of lowlanders differs from highlanders, it may imply that one group are exposed to different risks and require special interventions than the other. Indeed, mechanisms for adaptation to hypobaric hypoxia could be useful in other hypoxemic diseases such as chronic obstructive pulmonary diseases (COPD), myocardial ischemia or ARDS.3,12,14–16 On the other hand, validating the equation for adjusting the PaO2/FiO2 according to the altitude is also of paramount importance as is the variable used to define, stratify and select the best treatment for patients with ARDS.12,17–19 Likewise, due to the fact that PaO2/FiO2 is associated to the outcome in general population under invasive mechanical ventilation, predicting their outcome would be inaccurate in patients living in high altitudes.10,20

We hypothesize that altitude above the sea level affects the outcome of acclimatized patients undergoing invasively mechanically ventilated.

The aim of this study is (i) to validate the traditional equation for adjusting PaO2/FiO2 according to the altitude and (ii) to analyze the effect of altitude above the sea level on the mortality rate in patients undergoing invasive mechanical ventilation.

Design: prospective, observational, multicenter and international study conducted during August 2016.

Inclusion criteria: age between 18 and 90 years old, admitted to intensive care unit (ICU) situated at the same altitude above the sea level in which the patient has stayed, at least, during the previous 40 days before ICU admission and received invasive MV for at least 12h. The full list of centers can be appreciated in supplementary material. Two hospitals (“Obrero N°1” Hospital, Bolivia and Donostia University Hospital, Spain) retrieved the data retrospectively. Exclusion criteria include: home oxygen therapy, metastatic cancer, pregnancy, and absence of informed consent.

Thirty medical-surgical ICUs with 249 patients from 7 countries (Bolivia, Brazil, Colombia, Ecuador, Mexico, Peru and Spain) were analyzed (Table S1). Patients were assisted at the following altitudes: <1500m n=55; 1500–2500m n=20; 2500–3500m n=155 and >3500m n=19. Reason for ICU admission was: pulmonary sepsis n=111; extrapulmonary sepsis n=42; trauma n=16; acute respiratory failure n=14; stroke n=8; post-operative n=6; acute heart failure n=4; burned n=3, cardiac arrest n=3, hemorrhagic shock n=3, pancreatitis n=3 and miscellaneous n=36. Characteristics of all patients can be appreciated in Table 1. Comparison between highlanders and lownlanders can be appreciated in supplementary Table 1.

Adjusting PaO2/FiO2 according to the altitude

The correlation between all continuous variables, including PaO2/FiO2 (non adjusted) and PaO2/FiO2 (adjusted), is shown in Fig. 1 and supplementary Fig. 1. Adjusted and non-adjusted PaO2/FiO2 were correlated with PEEP, arterial systolic blood pressure and creatinine concentration below and above 1500m. However, respiratory rate, peak pressure, plateau pressure, PaCO2, arterial pH, tidal volume per kilogram, length of mechanical ventilation, body mass index and bilirubin concentration were correlated with both PaO2/FiO2 only in high altitude patients. PCR concentration and APACHE II score were correlated with both PaO2/FiO2 at sea level but not in altitude patients. None discordances between non adjusted and adjusted PaO2/FiO2 were identified.

Fig. 1.

Correlation between continuous variables at <1500m and >=1500m.

(0.65MB).

Mortality

Seventy-nine patients died during the ICU stayed (32%). The mortality curve was not affected by the altitude above the sea level (Fig. 2, log-rank 0.986). However, when the mortality is stratified according to the PaO2/FiO2 (non adjusted and adjusted) and the altitude dichotomized in the two main groups (<1500m and >=1500m), the observer mortality in each PaO2/FiO2 range was different between both altitudes (Table 2).

Fig. 2.

Survival analysis according to the altitude above the sea level.

(0.08MB).

Table 2.

Cochran–Mantel–Haenszel test (PaO2/FiO2 and mortality stratified according to altitude above sea level).

PaO2/FiO2		>=1500m		<1500m
	All patients	Survivours	Non surviours	Survivours	Non surviours
(a) Not adjusted
<100	59	30	24	2	3
100–149	59	31	23	5	0
150–199	52	36	7	6	3
200–300	53	30	6	8	9
>300	26	6	1	16	3
(b) Adjusted
<100	100	52	43	2	3
100–149	66	48	12	6	0
150–199	40	22	5	7	6
200–300	27	11	0	10	6
>300	16	0	1	12	3

(a) Not adjusted PaO2/FiO2

Cochran–Mantel–Haenszel M^2=16.428, df=4, p-value=0.002495.

(b) Adjusted PaO2/FiO2

Cochran–Mantel–Haenszel M^2=19.097, df=4, p-value=0.0007523.

The univariate analysis demonstrated that age, haemato-oncology disease, COPD, chronic renal failure, creatinine, platelets count, arterial systolic blood pressure, APACHE II, PaO2/FiO2 (non-adjusted), PaO2/FiO2 (adjusted), peak pressure and PEEP were associated to mortality (Table 1). All these variables except PaO2/FiO2 (non-adjusted) and APACHE II were included in the maximum logistic regression model. Finally, PEEP (OR 1.17; 95% CI 1.06; 1.31; p: <0.01), age (OR 1.02; 95% CI 1.01; 1.04; p: <0.01), systolic arterial blood pressure (OR 0.98; 95% CI 0.96; 0.99; p: <0.01) and platelet count (OR 0.9999974; 95% IC 0.9999949; 0.9999997, p: 0.03) were independently associated to mortality (Table 1). The AUROC of the logistic model was 0.72 (95% CI: 0.65; 0.79), the sensibility 0.27 (0.17; 0.38), specificity 0.91 (0.85; 0.95), positive likelihood ratio 2.82 (1.56; 5.11) and negative likelihood ratio 0.81 (0.70; 0.93) (Fig. 3). The final regression model did no change when PaO2/FiO2 (non-adjusted) was substituted by PaO2/FiO2 (adjusted).

Fig. 3.

Receiver operating characteristic curve of mortality (logistic regression model).

(0.05MB).

Main results of this pioneer study are: (i) the recommended formula for adjusting the PaO2/FiO2 ratio according to the altitude above the sea level seems to be inaccurate and (ii) the altitude above the sea level not influence the mortality risk.

According to our results, in term of mortality prediction or correlation with physiological variables, the traditional equation for adjusting the PaO2/FiO2 according to the altitude above the sea level does not add any advantage over the raw variable. This fact was expected, as the equation that links both variables is lineal. However, it is important to highlight that some correlations identified in patients at <1500m were not observed in patients at >=1500m and visceversa. Furthermore, the mortality rate, stratified according to PaO2/FiO2, also differ in both groups of patients. We hypothesize that these finding evidence that it would be necessary to adjust the PaO2/FiO2 according to the altitude; but the current equation is not accurate. Future studies with more patients and altitude levels should clarify this issue.

On the other hand, one of the hallmark studies performed in non-specific population under invasive mechanical ventilation was conducted by Esteban et al.10 It include 5183 patients from 361 ICU of Europe, North and Latin America and reported that several variable including PaO2/FiO2 <200 are independently associated to mortality.10 Likewise, Sudarsanam et al.20 analyzed 200 consecutive patients admitted to an medical UCI and reported that type 1 respiratory failure is also independent predictor of mortality. On line with them, we reported that PEEP, age, systolic blood pressure and platelet count were associated to mortality risk. However, in our study neither in the regression model nor in the survival curve the mortality rate was affected by the altitude above the sea level. This result, that should be validated in future studies, rejected the hypothesis of this study and could reflect the efficacy of acclimatization mechanisms to altitude. Based on this result, we speculate that people acclimatized to high altitude have physiologic mechanisms to compensate the deleterious effect that high altitude bear. Considering this, one natural question arises: “At sea level, Is the outcome of patients adapted to high altitude similar to lowlanders?”. Exploring the differences between humans exposed and not exposed to hypobaric hypoxia has the potential to identify mechanisms important in critical illness and perhaps to alter our therapeutic focus toward increasing the efficiency of oxygen utilization rather than improving delivery.24

We have to accept that this study has some limitations. Firstly, the number of patients in extreme altitude is very small which may be explained by the low number of ICUs at those geographical areas. This may influence the result because this subgroup of patients and the clinical practice in these ICUs could differ from those at sea level. Likewise, it could be possible that specific subgroups of patients (e.g. ARDS, surgical, etc.) carried specific features that were not identify in this study. Secondly, despite enrolling a large number of ICUs from South America and Spain, our sample does not include all possible altitudes. Thirdly, the study was conducted for one month and may not represent year-round. Fourth, two centers collected the data retrospectively. Although this is an evident protocol deviation, it is very improbable that may influence the results as booth ICUs have an informatics medical records that allow them to include patients and retrieve their data. Fifth, the mechanical ventilation between different hospitals was not standardized. This limitation is common to all observational study. Sixty, recently new variables such as driving pressure and mechanical power have demonstrated a close relation with the outcome. Future studies will have to address the effect altitude above the sea level on these variable. Finally, as this is a pioneer study, the knowledge previously available to interpret its result could be insufficient or inaccurate. Contrarily, this study has several strengths. Firstly, we used an objectively and previous published definition for considering a patient adapted to a specific altitude.21 Unfortunately, analytical or genetic biomarkers for defining adaptation to a specific altitude are not available or are inaccurate. Secondly, the outcome we have address (mortality) is objective. Thirdly, we included several procedures (e.g., website, quality-control, etc.) to reduce the risk of bias. Likewise, the proportion of patient excluded from the final analysis, as their data were incomplete reflect the strict of methodology we applied.

In acclimatized patients undergoing invasive mechanical ventilation, the traditional equation for adjusting PaO2/FiO2 according the elevation above the sea level seems to be inaccurate and the altitude above the sea level does not affect the mortality risk.

This study was partially supported by Escuela de Medicina de la Universidad Internacional del Ecuador (UIDE).

PC, GO, MJ, participated in the design of the study and performed the statistical analysis. PC, GO, MJ, MG conceived of the study, and participated in its design and coordination and helped to draft the manuscript. PC, GO, MJ, MG, FG, FM, JM, JV, OB, SS, FV, FM, JT, CI BV, FZ, AL participate in coordination of their centers and critical care units. All authors read and approved the final manuscript.

The authors declare no conflict of interest.