The GAP Model for Predicting Mortality Risk

Introduction

In the absence of effective treatments, the prognosis of idiopathic pulmonary fibrosis (IPF) has been very poor, with a median survival after diagnosis from 2 to 3 years based on several longitudinal studies.^1-4 Nonetheless, among individual patients there is wide heterogeneity.⁵ Some patients have an accelerated decline and die within months of diagnosis while many experience a slow, gradual progression over many years.⁵ This inherent uncertainty in prognosis may be frustrating to physicians who would like to provide patients with appropriate counseling in terms of life planning, treatment, and timely referral to lung transplant centers.⁶

In recent years, risk assessment systems for IPF have been developed.⁶ The baseline "GAP model" (so called for the variables it uses, Gender, Age, and Physiology) consists of 2 complementary prognostic tools that provide physicians with a framework for discussing prognosis and evaluating stage-specific management options with patients. The first GAP index and staging system provides a simple screening method to determine the average risk of mortality of patients with IPF by GAP stage. The second, the GAP calculator, provides an estimation of risk of mortality for those patients in whom a more precise estimation of risk may change management. The most recent addition to this series of models, the longitudinal GAP model, includes 2 additional variables which significantly improved the discriminative performance of the model: a history of respiratory hospitalization and 24-week change in FVC.⁷

Methods

To create the original GAP index and staging system, Ley et al⁶ used the records of 228 patients with IPF enrolled in the University of California, San Francisco, Interstitial Lung Disease Program’s longitudinal cohort study between 2001 and 2010 as a derivation cohort to develop the model. Then 330 patients entered in the Mayo Clinic at Rochester (Rochester, Minnesota) (n=208) and the Morgagni-Pierantoni Hospital (Forli, Italy) (n=122) databases between 2000 and 2010 were used as a validation cohort. By using competing-risks regression modeling, potential predictors of mortality were retrospectively screened. Potential predictor variables were required to be commonly measured and available at the time of initial consultation. Those considered included sex, body mass index, smoking status (those who had ever smoked versus those who had never smoked), use of long-term oxygen therapy, FVC, FEV₁, total lung capacity, and DL_CO (corrected for hemoglobin level when available).

Results

Point-Score Model (GAP Index)

Model screening selected 4 predictors and points were assigned to variable categories to create a point-score model (the GAP index) (Table 1).⁶ Then, total point scores were grouped into 3 stages (the GAP staging system) (Table 2).⁶

Table 1.
The GAP index. Adapted from Ley B, et al. Ann Intern Med. 2012;156:684-691.⁶

	Predictor	Points
G	GenderFemaleMale	0 1
A	Age, y≤60 61-65 >65	0 1 2
P	PhysiologyFVC, % predicted>75 50-75 <50DLCO, % predicted>55 36-55 ≤35 Cannot perform	0 1 2 0 1 2 3
	Total Possible Points	8

DL_CO, diffusing capacity for carbon monoxide; FVC, forced vital capacity; GAP, gender, age, and lung physiology variables (FVC and DL_CO).

Table 2.
The GAP staging system. Adapted from Ley B, et al. Ann Intern Med. 2012;156:684-691.⁶

Stage	I	II	III
Points	0-3	4-5	6-8
Mortality
1-year	5.6	16.2	39.2
2-year	10.9	29.9	62.1
3-year	16.3	42.1	76.8

In order to obtain a total point score (range, 0–8), points are assigned for each variable of the scoring system.⁶ DL_CO should be corrected for hemoglobin level when available. If patients’ symptoms or lung function prohibits performance of the DL_CO maneuver, they should be scored in the “Cannot perform” category for DL_CO. The model cannot be applied if DL_CO results are not available because it was not done or not completed because of nonrespiratory restrictions. The total point score is used to classify patients as stage I (0–3 points), stage II (4–5 points), or stage III (6–8 points). Table 2 lists model-predicted 1-, 2-, and 3-year mortality by stage.

The Continuous Model (GAP Calculator)

The same 4 predictors as the point-score model were selected by model screening for the continuous model (the GAP calculator): gender, age, and 2 lung physiology variables (FVC and DL_CO).⁶ Various statistical maneuvers were used to create a continuous model which was then updated by utilizing coefficients and cumulative subdistribution hazards based on the combined cohort to yield the most generalizable estimates for clinical application (Figure 1).

Figure 1.
The GAP calculator.⁶

In order to obtain estimates of 1-, 2-, and 3-year mortality, the patient gender is selected and actual values for age, FVC percent predicted, and DL_CO percent predicted are entered into the calculator.⁶ DL_CO should be corrected for hemoglobin level when available. As with the GAP index, if patients’ symptoms or lung function prohibits performance of the DL_CO maneuver, they should be scored in the “Unable to perform” category for DL_CO. The model cannot be applied if DL_CO results are not available because it was not done or not completed because of nonrespiratory restrictions.

Longitudinal GAP Model
Recently, Ley and colleagues published an update to the GAP model to include longitudinal variables.⁷ Of the variables investigated, a history of respiratory hospitalization and 24-week change in FVC significantly improved the predictive performance of the GAP model. Similar to the original GAP model,⁶ points are assigned for each variable of the scoring system to obtain a total point score (Table 3).⁷ The total point score is then used to estimate mortality risk for that particular patient (Table 4).

Table 3. The longitudinal GAP categorical and point index models. Adapted from Ley B, et al. Eur Respir J. 2015;45:1374-1381.⁷

Predictors	1-Year Mortality			2-Year Mortality			All Follow-Up
	HR	p-value	Points	HR	p-value	Points	HR	p-value	Points
GenderMale Female	1.06 -	0.810 -	1 0	1.11 -	0.610 -	1 0	1.13 -	0.528 -	1 0
Age, years>65 61-65 ≤60	1.52 1.07 -	0.123 0.839 -	4 1 0	1.49 1.18 -	0.078 0.576 -	4 2 0	1.54 1.19 -	0.056 0.558 -	4 2 0
Baseline FVC, % Predicted<50 50-75 >75	4.27 3.32 -	0.002 0.001 -	15 12 0	3.29 2.35 -	0.001 0.001 -	12 9 0	3.14 2.37 -	0.002 0.001 -	11 9 0
Baseline DLCO, % PredictedUnable to perform ≤35 36-55 >55	9.5 2.89 1.77 -	<0.001 0.054 0.282 -	23 11 6 0	7.37 2.77 1.74 -	<0.001 0.018 0.170 -	20 10 6 0	8.10 3.05 1.83 -	<0.001 0.009 0.135 -	21 11 6 0
∆FVC, % Predicted≤-10 -10 to -5 >-5	3.46 1.65 -	<0.001 0.074 -	12 5 0	2.68 1.57 -	<0.001 0.050 -	10 4 0	2.78 1.63 -	<0.001 0.031 -	10 5 0
History of Respiratory HospitalizationYes No	3.94 -	<0.001 -	14 0	3.84 -	<0.001 -	13 0	3.83 -	<0.001 -	13 0

DL_CO, diffusing capacity for carbon monoxide; FVC, forced vital capacity; GAP, gender, age, and lung physiology variables (FVC and DL_CO); HR, hazard ratio.

Table 4.
The longitudinal GAP point index models: predicted 1- and 2-year mortality risks. Adapted from Ley B, et al. Eur Respir J. 2015;45:1374-1381.⁷

Predicted Risk	1-Year Mortality	2-Year Mortality
<2%	0-10	-
2-5%	11-19	0-6
5-10%	20-26	7-13
10-20%	27-34	14-20
20-30%	35-38	21-25
30-40%	39-42	26-29
40-50%	43-45	30-32
50-60%	46-48	33-35
60-70%	49-51	36-37
70-80%	52-54	38-40
≥80%	55-69	41-60

Summary

The baseline and longitudinal GAP indices and staging systems, and GAP calculator discussed here, are validated tools that use commonly measured clinical and physiologic variables to predict mortality in IPF.^6,7 The use of these tools in clinical practice may help guide discussions with patients about life planning, facilitate treatment decisions, and provide a framework for timely referral to lung transplant centers.

References

1. Flaherty KR, Toews GB, Travis WD et al. Clinical significance of histological classification of idiopathic interstitial pneumonia. Eur Respir J. 2002;19(2):275-283.
2. Nicholson AG, Colby TV, du Bois RM, Hansell DM, Wells AU. The prognostic significance of the histologic pattern of interstitial pneumonia in patients presenting with the clinical entity of cryptogenic fibrosing alveolitis. Am J Respir Crit Care Med. 2000;162(6):2213-2217.
3. Rudd RM, Prescott RJ, Chalmers JC, Johnston ID. British Thoracic Society Study on cryptogenic fibrosing alveolitis: Response to treatment and survival. Thorax. 2007;62(1):62-66.
4. Bjoraker JA, Ryu JH, Edwin MK et al. Prognostic significance of histopathologic subsets in idiopathic pulmonary fibrosis. Am J Respir Crit Care Med. 1998;157(1):199-203.
5. Raghu G, Collard HR, Egan JJ et al. An official ATS/ERS/JRS/ALAT statement: idiopathic pulmonary fibrosis: evidence-based guidelines for diagnosis and management. Am J Respir Crit Care Med. 2011;183(6):788-824.
6. Ley B, Ryerson CJ, Vittinghoff E et al. A multidimensional index and staging system for idiopathic pulmonary fibrosis. Ann Intern Med. 2012;156(10):684-691.
7. Ley B, Bradford WZ, Weycker D, Vittinghoff E, du Bois RM, Collard HR. Unified baseline and longitudinal mortality prediction in idiopathic pulmonary fibrosis. Eur Respir J. 2015;45(5):1374-1381.