Clinical characteristics and a diagnostic model for high-altitude pulmonary edema in habitual low altitude dwellers

Introduction

Short-term exposure may cause non-acclimatized individuals to suffer from acute mountain sickness, high-altitude pulmonary edema (HAPE), and high-altitude cerebral edema (HACE) following a rapid ascent above 2,500 m (Sartori et al., 2002). HAPE represents a life-threatening form of lung injury resulting from exposure to hypobaric hypoxia at high altitudes. Typical symptoms include shortness of breath, fever, tachycardia, hacking cough, cyanosis, and occasionally pink sputum (Swenson et al., 2002). HAPE constitutes a form of hydrostatic pulmonary edema wherein the permeability of the alveolar capillaries is altered. There are some possible pathways for the development of early HAPE, one is a traumatic breakdown of the basement membrane of the alveolar capillary barrier caused by high pulmonary arterial pressure, allowing passage of plasma and blood cells, and another pathway may be of transcellular passage through vesicular channels that formed by high intracapillary pressure (Swenson et al., 2002). Altitude, ascending speed, and ascending mode, especially individual susceptibility, are the most important determining factors for the occurrence of HAPE (Bartsch, 1999).

A series of physiological responses across various body systems is involved in the process of HAPE. At high altitudes, hypoxia and inflammation can impair the function of the lungs, heart, and kidneys. Consequently, blood parameters and biochemical factors are reportedly as crucial to high-altitude acclimatization and disease management as other factors (Bhandari & Cavalleri, 2019; Huang et al., 2017; Sanchez et al., 2022). HAPE can trigger a severe inflammatory response or even death in later stages of its development if not appropriately and promptly treated (El Alam et al., 2022). However, current studies infrequently present a comprehensive analysis of the clinical characteristics and changes in blood biochemical indicators in patients with HAPE. Although chest X-rays and CT scans are the primary diagnostic methods for HAPE, many high-altitude regions have limited medical resources and lack specialized equipment. Additionally, HAPE can easily be confused with other lung injuries, leading to incorrect medication, a waste of medical resources, and complicating treatment. There are few statistical models for the integrated diagnosis of HAPE that incorporate physiological indicators, blood parameters, and biochemical factors. This study analyzed the laboratory indicator characteristics of HAPE patients, constructed a diagnostic model, and visualized it using a nomogram. The diagnostic goal is to detect or identify HAPE in its early subclinical stage. Investigators continuously monitored individuals for evidence of HAPE based on specific clinical manifestations of the disease.

Materials and Methods

Results

Clinical characteristics

From March 2015 to May 2022, 1,255 subjects were screened for eligibility (Fig. 1). The hospital-based sample consisted of 842 males and 413 females, aged 18 to 76 years. Of these, 604 subjects were admitted to the emergency room of Houbei Hospital a few days after arriving at 3,500 m above sea level, presenting significant symptoms of altitude sickness, and were later hospitalized with a diagnosis of HAPE. The control group comprised 651 subjects who were hospitalized at Houbei Hospital during the same period and met the inclusion criteria. Within the data framework, all participants self-identified as Han Chinese, having lived on the plains for an extended period. No indigenous Tibetans were included in the analyzed population. The confidence level for all sampled data was set at 95%. Except for age, chloride (Cl), globulin (GLO), total bilirubin (TBIL), and direct bilirubin (DBIL), the remaining indices differed between the HAPE and control groups. Subsequent analyses were conducted after excluding indicators that showed no differences. Of the 604 HAPE patients, 76.37% were men, 1.99% were aged over 65 years, and 47.85% were overweight (Table S1).

Development of an individualized diagnostic model

From a total of 34 features, 11 potential predictors were identified from 864 patients (Figs. 2A and 2B), characterized by having non-zero coefficients in a LASSO logistic regression model. The model, incorporating the independent predictors mentioned above, was developed (Table S3) and is presented as a nomogram (Fig. 3).

Validation of the HAPE nomogram

Discrimination: Figure 4 illustrates the area under the ROC curve (AUC). The nomogram demonstrated good discrimination, with AUC values of 0.787 (95% CI [0.757–0.817]) for the training dataset and 0.833 (95% CI [0.793–0.874]) for the validation dataset.

Calibration: The calibration curve, displayed in Figs. 5A and 5B, was evaluated using the U-test, yielding P-values of 0.793 for the training dataset and 0.629 for the validation dataset. The non-significance of the Hosmer-Lemeshow test indicates a high level of agreement. The results of the calibration curve demonstrate that the predicted probabilities from the model align closely with the actual observed outcomes, indicating strong consistency and reliability of the model’s predictions.The accuracy of the predictions was assessed using the Brier Score (BS), defined as the mean squared deviation between the predicted probabilities and the actual outcomes. The Brier Score was 18.7% for the training set and 18.0% for the validation set, as shown in Figs. 5A and 5B. Meanwhile, a BS of 24.3% was calculated through bootstrap resampling of 100 replicates (Fig. 5C). Confusion matrices, compiled for both datasets, indicated accuracies of 70.95% for the training set and 74.17% for the validation set (Table S4). The nomogram-based predictions showed good agreement with the observations from the calibration curve regarding the probability of HAPE.

Discussion

HAPE typically occurs and progresses rapidly. If not treated promptly and effectively, HAPE can deteriorate into multiple organ dysfunction syndrome, potentially complicating treatment or resulting in death (Choudhary et al., 2023). Prompt diagnosis is essential to prevent the condition from worsening. An accurate diagnostic model enables healthcare professionals to detect HAPE in its early stages or before symptoms emerge, allowing for earlier intervention, thus averting severe complications and lowering mortality rates. Diagnosis via laboratory markers is generally simpler and more convenient than imaging studies or other invasive tests. If the model can be based on standard blood and biochemical tests, it will significantly simplify and lower the cost of diagnosis. In this study, we analyzed changes in clinical characteristics and biochemical markers among Han Chinese who traveled to high-altitude destinations to determine if these significant changes are associated with HAPE. Our study further explored the diagnostic model for HAPE, utilizing demographic factors, physiological indices, hematological parameters, and biochemical indices. To our knowledge, this is a unique diagnostic model that reports the occurrence of HAPE in Han Chinese following rapid ascent to the Tibetan Plateau.

The results indicated that the most prevalent clinical and laboratory changes in HAPE patients, resulting from high-altitude exposure, were related to blood cells, immune cells, liver, and heart functions. Utilizing these clinical and laboratory changes, a diagnostic nomogram for individualized prediction of HAPE has been developed and validated. For the development of the HAPE signature, 34 candidate features were reduced to 11 potential predictors by assessing the associations between predictors and outcomes and shrinking the regression coefficients via the LASSO method. The nomogram incorporates eleven items, including gender, BMI, DP, MAP, WBC, LYMPH%, HCT, MCV, MCHC, PLT, and MPV. This predictive model demonstrated adequate discrimination in the training cohort (AUC: 0.787), with a surprising improvement in the validation cohort (AUC: 0.833). The chosen model, which boasts a low Brier score, was computed using the bootstrap resampling method and validated in both the training and validation sets. Given that lower Brier scores indicate higher model accuracy, the model with the lowest Brier score was ultimately selected as the final model (Seif et al., 2014).

Interestingly, only blood parameters, along with gender, age, BMI, and blood pressure, were needed to make predictions about HAPE. Health systems in high-altitude areas are frequently characterized by inequitable distribution, limited medical resources, transport difficulties, and the isolation of small regions (Li et al., 2012). As detailed in the results, the nomogram developed in our study predicts the likelihood of an individual developing HAPE by solely combining blood parameters and demographic factors. A nomogram integrates multiple predictive indicators, representing them on a single plane through scaled line segments to illustrate the interrelationships among variables in the predictive model. This visualization, typically arranged as column charts, simplifies complex regression equations, making the outcomes of prediction models more intuitive and accessible. Such design enhances understandability, allowing even non-professionals to easily interpret and utilize the information. Users can swiftly estimate the predicted outcomes for high-altitude pulmonary edema (HAPE) under specific conditions by simply aligning values and reading off the chart. Employing only routine blood tests for diagnosis, without excessive laboratory testing, reduces patient costs. This approach not only improves clinical decision-making for clinicians and patients but also results in a greater net benefit. In line with the current trend towards personalized medicine, this user-friendly scoring system enables both clinicians and patients to make individualized predictions of HAPE risk (Balachandran et al., 2015). Our diagnostic model enables early preventive interventions for patients with HAPE and, conversely, prevents the overtreatment or biased treatment of those without HAPE. Numerous retrospective observational analyses have described the complex interplay of various erythrocyte indices in acclimatization and adaptation to high-altitude hypoxia. Exposure to hypoxia at high altitudes is known to stimulate increased bone marrow production of red blood cells (erythropoiesis) and the formation of new blood vessels (angiogenesis) (Mallet et al., 2023). Research indicates that in the acute phase, blood cell counts such as RBC, HGB, and HCT significantly increase and continue to do so in the chronic phase, while platelet counts are significantly reduced, benefiting blood flow (Azad et al., 2017; Srihirun et al., 2012). In the chronic phase, the volume of red blood cells and MCH continue to increase, thereby enhancing the blood’s oxygen-carrying capacity (Sanchez et al., 2022). Additionally, neutrophils and platelets are simultaneously activated to regulate the inflammatory response during acute inflammation (Daseke et al., 2019). Although mechanistically HAPE is not triggered by inflammation, its occurrence can lead to inflammatory complications within the body (Idzko et al., 2014). In our study, both white blood cell count and total neutrophil percentage are critical indicators for predicting the inflammatory response that accompanies most cases of HAPE. Routine blood tests are instrumental not only in diagnosing HAPE but also in monitoring changes in a patient’s condition during clinical treatment. Regular assessments enable physicians to track critical parameters such as oxygen levels, red blood cell counts, inflammatory markers, and other indicators included in the diagnostic model. Variations in these metrics can reveal the patient’s response to treatment and provide prognostic insights. Timely adjustments to treatment strategies—such as modifying oxygen supplementation, pharmacotherapy, and supportive care—can significantly enhance patient outcomes and reduce the risk of complications.

We also found that HAPE occur more frequently in males compared with female. Some studies demonstrated that females are less susceptible to high-altitude sickness and are better adapted to high altitude than males (Hackett et al., 1998). The reason for this may be that estrogen dilates the renal microvascular system, increases the hematocrit in blood from small arteries, capillaries and veins, alters oxygen delivery per unit of erythrocyte mass and provides a mechanism for altering erythrocyte mass without compensatory erythropoiesis (Murphy, 2014). Considering potential uncorrected gender differences in logistic regression models, we conducted separate analyses for each gender. The results demonstrated that the area under the curve (AUC) for the three datasets—the total dataset, male dataset, and female dataset—were 0.797 (95% CI [0.773–0.822]), 0.796 (95% CI [0.766–0.827]), and 0.817 (95% CI [0.775–0.859]) respectively. These findings indicate that age and gender heterogeneity have minimal impact on the diagnostic outcomes, as detailed in Table S5. In addition, overweight or obese people are more prone to HAPE than slimmer people in this study. A few of studies have focused on the relationship between BMI and high-altitude acclimatization (Coello et al., 2000; Shen et al., 2017), but there is scarce study focus on the relationship between BMI and HAPE. It was indicated that subjects with higher BMI responses reluctant to hypoxia. Acutely, blood pressure, respiratory function and oxygen saturation increased less in people with higher BMI than in those with lower BMI (Peng et al., 2013). The reason for this phenomenon may be that in subjects with a higher BMI, there is little room for blood pressure or respiratory function increase in the acute phase because they were already higher at baseline. Also, blood pressure and red blood cell volume increased less in people with a higher BMI than in people with a lower BMI. Human blood pressure (BP) homeostasis is challenged by hypoxia at high altitudes. The effects of altitude exposure on patients with hypertension have also been the subject of specific studies. Acute exposure to hypobaric hypoxia at high altitudes induces an increase in blood pressure in humans with high blood pressure, both at rest and during exercise (Caravita et al., 2018). In our study, DP was included in the model, suggesting that blood pressure is also an indicator of great interest in studying HAPE in Han Chinese. Exposure to high altitude is associated with significant changes in cardiovascular function, and this is probably the reason.

This study retrospectively analyzed previous case data to explore the disease characteristics and predictive factors of high-altitude pulmonary edema. Although this research method is widely used in medical research, there are also some inherent limitations. Firstly, data relies on existing medical records and faces the problem of incomplete data. Some important information may not be recorded or recorded incompletely. Secondly, this study is based on patients in a medical center, and the results may affect the promotion and application of the research findings in other populations.

Conclusion

In conclusion, we have presented clinical characteristics and a nomogram based on hematological parameters and demographic variables, which can be conveniently used for individualized diagnosis of HAPE. In high-altitude areas with limited emergency environments or resources, diagnostic models based on laboratory indicators can provide fast and reliable diagnostic support for medical staff, helping them make better treatment decisions. Moreover, the symptoms of high-altitude pulmonary edema may be similar to other high-altitude diseases or respiratory system diseases, which can easily lead to misdiagnosis. An accurate diagnostic model can reduce the risk of misdiagnosis and improve the accuracy of clinical diagnosis. By utilizing machine learning technology to process and analyze a large amount of patient laboratory data, and establishing and optimizing such diagnostic models, we can better understand the pathogenesis and risk factors of HAPE, providing data support for future research and drug development.

Supplemental Information