Factors related to respiration influencing survival and respiratory function in patients with amyotrophic lateral sclerosis: a retrospective study
Abstract
Background
Various breathing abnormalities (Neurology 2009; 73: 1218) have been proposed as indicators for the introduction of non-invasive positive-pressure ventilation (NIV) in patients with amyotrophic lateral sclerosis (ALS). We were interested in the usefulness of symptoms of respiratory insufficiency and abnormal results of daytime arterial gas analyses (AGA) as predictors of survival and the effect of NIV on respiratory volumes and pressures.
Methods
Reported symptoms, respiratory subscore of the ALS Functional Rating Scale (ALSFRS-r), Norris scale (Norris-r), and AGA were retrospectively analyzed in 189 ALS patients. Longitudinal follow-up of forced vital capacity (FVC), maximal inspiratory and expiratory pressure (MIP, MEP), and sniff nasal pressure (SNP) were analyzed with regard to the introduction of NIV.
Results
Respiratory symptoms were a bad prognostic sign (P = 0.007). Abnormalities in Norris-r, ALSFRS-r, pO2, pCO2, and oxygen saturation tended to be associated with a shorter survival, although they were not statistically significant. NIV prolonged survival and reduced the decline in FVC (P = 0.007), MIP, MEP, and SNP (the last three were not statistically significant). Symptoms, abnormal FVC, and AGA do not always coincide, and they can appear in a different sequence.
Conclusions
Any respiratory abnormality should prompt the clinician to start discussing NIV with the patient. NIV prolongs survival and improves respiratory function.
Introduction
Amyotrophic lateral sclerosis (ALS) is a rapidly progressive, invariably fatal neurodegenerative disease that attacks the motor nerve cells responsible for controlling voluntary muscles. The gradual degeneration and death of motor neurons lead to death from respiratory insufficiency 1. Non-invasive positive-pressure ventilation (NIV) and invasive mechanical ventilation (IMV) are used to alleviate respiratory symptoms, improve quality of life, and prolong survival 2. The guidelines of the European Federation of Neurological Societies and the American Academy of Neurology (AAN) recommend regular follow-up of symptoms and clinical signs of respiratory insufficiency, forced vital capacity (FVC), maximal inspiratory pressure (MIP), sniff nasal pressure (SNP), and nocturnal oximetry 2, 3. However, no single test is considered to be the most appropriate for detecting impending respiratory failure 4. In the present retrospective study, we were interested in the usefulness of symptoms of respiratory insufficiency and abnormal results of daytime arterial gas analyses (AGA) as predictors of survival and whether the introduction of NIV decreases a pre-NIV decline of FVC, MIP, maximal expiratory pressure (MEP), and SNP in terms of their slopes of decline before and after the introduction of NIV.
Methods
Patients
We retrospectively analyzed 189 of 206 patients with ALS who were followed by our ALS clinic from October 2002 to December 2010. All patients were diagnosed according to the revised El Escorial criteria 5. Amongst the 17 excluded patients, three were on IMV, two committed suicide, one had a pulmonary embolism, three had an additional severe disease such as cancer, cerebrovascular insult, multiple sclerosis, that might have independently shortened survival or influenced respiratory functions, and eight patients were seen only once so their follow-up data were incomplete. The study was part of the research project P3-0338 and was approved by the National Medical Ethics Committee.
Investigations
Patients were regularly followed in an outpatient clinic at 3-month intervals. They were questioned whether at least one symptom of respiratory insufficiency was present (dyspnea, orthopnea, or morning headaches). In further analyses, we used respiratory symptoms as binary data, that is, as either present or not. Their respiratory function was assessed using the ALS Functional Rating Scale (ALSFRS-r) four-point subscore as either 4 (normal), 3 (shortness of breath with minimal exertion), 2 (shortness of breath at rest), 1 (intermittent ventilatory assistance required), or 0 (ventilator dependent) 6, and the respiratory subscore of the Norris scale (Norris-r) 7 as either 3 (normal), 2 (impaired), 1 (trace), or 0 (no use).
Respiratory muscle function was assessed by measurement of FVC, SNP, MIP, MEP, and daytime AGA. FVC was measured using a dry wedge spirometer (Vitalograph ALPHA, Ennis, Ireland). MIP, MEP, and SNP were measured using a respiratory pressure meter (MicroRPM, CareFusion, San Diego, CA, USA). FVC, MIP, and MEP were measured using a pressure transducer, flanged mouthpiece and nose clip with attention being paid to obtaining a good lip seal. All pulmonary function tests were performed in the sitting position. All patients had to repeat each maneuver three times, with the best result being recorded and used for analysis. FVC was expressed as the deficit with respect to the minimal predicted value for matched age, height, smoking status, and gender. Arterial blood samples were taken using heparin-containing syringes (Gas Lyte sampler, Vital signs, Totowa, NJ, USA). Oxygen saturation, pO2, and pCO2 were used for further analyses.
Patients and their relatives were asked about the date of disease onset. Patients with dysphagia were offered percutaneous endoscopic gastrostomy (PEG) and NIV at the onset of respiratory insufficiency. Treatment with riluzole (Rilutek®) was recommended for all patients, some of whom refused the drug. Age of patients, usage of PEG, therapy with riluzole, and the El Escorial criteria were used in survival analyses, as they might have impact on the survival besides respiratory symptoms, functional tests, and AGA 8-10.
The indications for NIV were orthopnea, dyspnea, and morning headache, and one of the following: PaCO2 > 6.0 kPa, nocturnal oxygen desaturation <90% for 1 cumulative minute or MIP < −60 cmH2O, SNP < 40 or FVC < 50% predicted 3, 11. NIV treatment was carried out using a bi-level positive airway pressure device. IPAP was increased until the patient's AGA normalized and the symptoms of hypoventilation disappeared, or until the increase was no longer tolerated by the patient. The efficient use of NIV was defined as usage at least 4 h per night. The usage was checked by smart card. None of the patients opted for IMV.
Statistical analyses
To avoid collinearity in multivariable models, bivariate associations were first screened using the Pearson's correlation, the independent samples t-test, and Fisher's exact test for numeric–numeric, numeric–binary, and binary–binary pairs of variables, respectively. Survival was analyzed using proportional hazards regression (i.e. the Cox model; see the Results section for details), and the Kaplan–Meier method and the exact log-rank test (for two-group comparison regarding the use of NIV).
All the cases were complete for the analyses excluding the measures of respiratory function. For analysis of the measures of respiratory function, data were missing in 36 cases: only Norris scale data were missing in one case; data from both the Norris scale and ALSFRS in 23 cases; the former two and oxygen saturation in one case; oxygen saturation, pO2, and pCO2 in 25 cases; and from all five measures in 11 cases. The missing data were imputed by means of logistic regression, whereby a separate model was built for each variable from all the available information (excluding the other measures of respiratory function) and the cutoff value of predicted probability for classification was determined from ROC analysis as the one maximizing the sum of sensitivity and specificity.
For assessing the effect of NIV, the median of piecewise (i.e. two-adjacent-point) slopes (with time expressed in days) was calculated for each patient as the outcome measure, whereby it was assumed that at death, MIP, MEP, and SNP decrease to zero, whilst FVC reaches 100 (and the trend from the last two observations was extrapolated for the patients who were still alive at the time of the study). The median slope was then compared between groups with regard to NIV using the independent samples t-test. This is a very crude and simplistic approach, but it was chosen because of the scarcity of available data (as detailed in the Results section) and because intrasubject comparisons would be biased, whilst sophisticated modern models (generalized mixed-effect models, that is, random-effect nonlinear models) would require making unfounded assumptions about the growth-curve form and the correlational structure of the parameters. Statistical analyses were performed using SPSS for Windows 15.0 software (SPSS Inc., Chicago, IL, USA).
Results
We analyzed the data of 189 patients (91 men and 98 women) with ALS. The mean age of disease onset was 62 years (median, 63 years; range, 35–82 years). At the onset of the disease, 60 patients had predominantly the bulbar form, whilst the remainder had the spinal form. Sixty-two patients required PEG. Ninety-nine patients were treated with riluzole. One hundred and nine patients were diagnosed at last examination as definite ALS according to the revised El Escorial criteria, 55 as probable, 11 possible and 14 as suspected ALS. One hundred and thirty-three patients died during the study.
A hundred and nine (58% of all included) patients reported orthopnea, dyspnea or morning headaches during the disease. However, when they were asked or examined using scales respiratory problems were found more frequently. The presence of respiratory symptoms was not always in accordance with abnormalities in AGA and FVC (Table 1). Amongst those patients who died, 40 (30%) never complained of respiratory symptoms. Thirteen (10%) had no respiratory problems even during the last 3 months of their life (Table 1). Similarly, the daytime AGA, MIP, SNP and FVC revealed normal parameters in some patients even during the last 3 months of life: pCO2, pO2, and oxygen saturation were normal in 24 (20%), 19 (16%) and 25 (21%) of 118 patients, respectively, FVC > 50% in six (7%) patients, and MIP and SNP in one (4%) patient. Only 4% of patients did not notice respiratory symptoms nor was daytime AGA abnormal during the last 3 months of life (Table 2).
| Symptoms (no., %) | Norris-r < 3 (no., %) | ALSFRS-r < 4 (no., %) | |
|---|---|---|---|
| 109 (58%) | 95 (62%) | 122 (84%) | |
| No. of evaluated patients | 188 | 153 | 145 |
| With examined AGA and FVC | 59 | 44 | 48 |
| Normal pO2, pCO2, FVC | 11 | 1 | 0 |
| Abnormal pO2 or pCO2 or FVC | 14 | 13 | 17 |
| Abnormal 2 of pO2, pCO2 or FVC | 12 | 11 | 11 |
| Abnormal pO2, pCO2 and FVC | 22 | 19 | 20 |
| No symptoms (no., %) | Norris-r = 3 (no., %) | ALSFRS-r = 4 (no., %) | |
| 40 (30%) | 35 (32%) | 19 (17%) | |
| No. of evaluated patients who died | 133 | 110 | 109 |
| Examination during the last 3 months of their life | 13 (10%) | 12 (11%) | 7 (6%) |
- Respiratory symptoms do not always coincide with abnormalities in AGA or FVC. Even at the examination during the last 3 months of life, some patients do not notice symptoms.
| No. of examined patients | 118 | 118 | 118 | 88 | 24 | 24 |
|---|---|---|---|---|---|---|
| Normal pCO2 (%) | Normal pO2 (%) | Normal oxygen saturation (%) | FVC > 50% (%) | MIP > −60 cmH2O (%) | SNP > 40 cmH2O (%) | |
| No. of patients at last examination | 51 (43) | 49 (42) | 58 (49) | 39 (44) | 4 (17) | 5 (21) |
| Examined during the last 3 months of their life | 24 (20) | 19 (16) | 25 (21) | 6 (7) | 1 (4) | 1 (4) |
| And without symptoms | 11 (9) | 5 (4) | 7 (6) | 3 (3) | 1 (4) | 0 |
- At the time of the disease (including patients who died during the study), not all patients have abnormalities in daytime arterial gas analyses (AGA), forced vital capacity (FVC), maximal inspiratory pressure (MIP), or sniff nasal pressure (SNP). Some of them already notice symptoms, but not all.
One hundred and eight patients who reported symptoms of respiratory insufficiency, which started, on average, 17 months after disease onset (range, 0–108 months), died, on average, in 9 months (range, 0–33 months).
Symptoms of respiratory insufficiency preceded abnormalities in pCO2, pO2 and oxygen saturation, and FVC < 50% in 78% (78 patients had symptoms first, whilst in 22 patients abnormal pCO2 preceded the symptoms), 71% (82 patients had symptoms first, whilst in 34 patients abnormal pO2 preceded the symptoms), 76% (81 patients had symptoms first, whilst abnormal oxygen saturation preceded the symptoms in 26 patients) and 60% (125 patients had symptoms first, whilst FVC < 50% preceded the symptoms in 83 patients), respectively.
Initially, the Cox model without measures of respiratory function was fitted, with respiratory symptoms coded as either absent, absent after NIV or present despite NIV. Older age at onset (P = 0.002), the presence of respiratory symptoms (P = 0.025) and absence of riluzole (P = 0.048) were found to be significantly associated with shorter survival, whilst the diagnostic criteria according to the revised El Escorial were marginally associated (P = 0.064) (Table 3). Amongst patients with respiratory symptoms, regular use of NIV decreased the difference between them and patients without respiratory symptoms (i.e., improved survival). This is evidenced by the higher P-value (0.074) and lower estimated hazard ratio (HR = 1.54) for the comparison of patients with respiratory symptoms receiving NIV to patients without respiratory symptoms; with regard to the comparison of patients with respiratory symptoms not receiving NIV to patients without respiratory symptoms P = 0.007, HR = 1.84.
| Predictor | b | P | HR (95% CI) |
|---|---|---|---|
| Age at ALS onset (years) | 0.031 | 0.002 | 1.031 (1.011, 1.051) |
| Gender (female versus male) | 0.026 | 0.887 | 1.030 (0.722, 1.468) |
| ALS diagnosis according to the revised El Escorial criteria (possible, probable or certain versus suspect) | 0.644 | 0.064 | 1.891 (0.956, 3.740) |
| ALS type at onset (bulbar versus spinal) | 0.045 | 0.835 | 1.070 (0.704, 1.627) |
| Respiratory symptoms | 0.025 | ||
| Yes with NIV versus no | 0.429 | 0.074 | 1.536 (0.960, 2.459) |
| Yes without NIV versus no | 0.610 | 0.007 | 1.840 (1.180, 2.869) |
| Therapy with Rilutek (yes versus no) | −0.365 | 0.048 | 0.694 (0.483, 0.996) |
| PEG (yes versus no) | −0.078 | 0.716 | 0.919 (0.601, 1.405) |
- b, regression coefficient; P, statistical significance; HR, estimated hazard ratio; CI, confidence interval.
Next, a Cox model without the information on observed respiratory symptoms (retaining the remaining six predictors) was fitted for each of the five measures of respiratory function in turn. To avoid confounding, the 46 patients who received NIV were excluded from those analyses (i.e. sample size for these analyses was 143). Any abnormality in each measure (Norris-r, ALSFRS-r, pO2, pCO2, and oxygen saturation from AGA) tended to increase the hazard ratio, but no effect was statistically significant. Oxygen saturation came closest to statistical significance (P = 0.065) and had the highest estimated hazard ratio, so it might be considered as the most reliable predictor amongst these measures (Table 4).
| Measure of respiratory function in the model | b | P | HR (95% CI) |
|---|---|---|---|
| Norris-r | 0.330 | 0.147 | 1.390 (0.891, 2.170) |
| ALSFRS-r | 0.272 | 0.292 | 1.313 (0.791, 2.178) |
| pO2 | 0.335 | 0.131 | 1.398 (0.905, 2.160) |
| pCO2 | 0.240 | 0.260 | 1.271 (0.838, 1.930) |
| Oxygen saturation from AGA | 0.404 | 0.065 | 1.497 (0.974, 2.301) |
- b, regression coefficient; P, statistical significance; HR, estimated hazard ratio; CI, confidence interval.
- To avoid confounding, the 46 patients who received NIV were excluded from these analyses; in all models, P-values for age at ALS onset and treatment with Rilutek were < 0.05, and P-value for the dichotomized ALS diagnosis according to the revised El Escorial criteria was between 0.05 and 0.15.
Univariate analysis confirmed that NIV significantly prolonged survival in patients with symptoms of respiratory insufficiency. Of the 46 patients with respiratory insufficiency who regularly used NIV for more than 4 h per day, eight (17%) were still alive at the end of the study compared to only nine (13%) of the 67 patients with respiratory insufficiency who did not use NIV. The patients with NIV had a longer estimated survival (mean, 36 months; 95% CI, 29–44 months after the onset of symptoms; median, 30 months; 95% CI, 22–38 months) compared to those who refused NIV (mean, 35 months; 95% CI, 27–42 months after the onset of symptoms; median, 24 months; 95% CI, 19–29 months: P = 0.004).
Non-invasive positive-pressure ventilation not only improved survival, but also influenced respiratory function. FVC was continuously followed in 111 patients. In the patients with NIV, the rate of FVC reduction (as estimated by the median of pairwise slopes) was statistically significantly lower than in those without NIV (P = 0.007; Table 5). MIP, MEP, and SNP were followed in 40 patients. The decline in these indicators of respiratory function also decreased with the introduction of NIV (Table 5), although the change in the median slope was not statistically significant.
| Indicator of respiratory function | NIV | Median two-point slope | P | |
|---|---|---|---|---|
| Mean | SD | |||
| FVC | No | 0.279 | 0.392 | 0.007 |
| Yes | 0.149 | 0.094 | ||
| MIP | No | −0.247 | 0.327 | 0.391 |
| Yes | −0.159 | 0.112 | ||
| MEP | No | −0.255 | 0.479 | 0.158 |
| Yes | −0.112 | 0.144 | ||
| SNP | No | −0.125 | 0.398 | 0.620 |
| Yes | −0.087 | 0.099 | ||
- NIV, non-invasive positive-pressure ventilation; FVC, forced vital capacity; MIP, maximal inspiratory pressure; MEP, maximal expiratory pressure; SNP, sniff nasal pressure.
Discussion
Our data indicate that the patients who reported symptoms or had signs of respiratory insufficiency on routine clinical examination had a statistically significant shorter survival than those without such symptoms/signs. Similarly, abnormalities in Norris-r and ALSFRS-r tended to be associated with a shorter survival, although these were not statistically significant. Patients with respiratory insufficiency at disease onset have a poor prognosis compared to those with the bulbar or spinal forms 12. These findings were consistent with reports that dyspnea is generally associated with diaphragmatic dysfunction 13 and with the neuropathologic finding of predominantly phrenic nerve abnormalities 14.
In our study, about one-third of the patients never mentioned respiratory symptoms. Oliver et al. 15 found that 15% of ALS patients did not experience dyspnea. The reason for this discrepancy may be that patients in the terminal phase are not able to attend the clinic, although some of our patients were examined during the last 3 months of their life, or even died in hospital, yet they reported no symptoms.
Respiratory symptoms often (in more than 60% of patients) precede the abnormalities in daily AGA and FVC. Tsara et al. 16 reported dyspnea in only 39.3% of patients, although other patients without dyspnea had abnormal results of FVC and MIP. Similowski et al. 17 described patients without dyspnea but with FVC as low as 38% of predicted values. In addition, hypercapnia can be found in patients without breathlessness at rest or orthopnea 18. We found that abnormal daytime pO2, pCO2, and oxygen saturation from AGA appeared to be associated with a shorter survival, but no association was statistically significant. Oxygen saturation is considered to be the most reliable predictor amongst AGA tests. Blood exchange abnormalities (high pCO2) are generally late findings 2, especially on daytime examination. We used daytime AGA because they can be performed at regular examinations, and their values used to check the NIV settings. Nocturnal oximetry or even polysomnography with morning AGA can detect nocturnal hypoventilation earlier, but these procedures are inconvenient for regular follow-up of disease progression in the outpatient clinic. Daytime AGA do not always coincide with dyspnea or orthopnea 18, 19, which is contrary to the findings of Polkey et al. 20, who found no patients to be hypercapnic or hypoxic without respiratory symptoms. About one-fifth of the patients in our study never had any abnormalities in daytime AGA; even with asking about respiratory symptoms some percentages of patients still remain with no symptoms nor abnormalities in respiratory functional tests.
Patients with respiratory failure have a shorter survival, but with regular use of NIV during sleep this can be somewhat improved. The decline in FVC, MIP, MEP, and SNP with disease progression decelerates after the introduction of NIV. Similar findings have been published previously 21-23. With bi-level positive airway pressure, survival can be statistically significantly prolonged for up to 12 months 22, 24 or even longer 21, 25. Nocturnal non-invasive ventilation therefore favorably influences the quality of life, prolongs survival 25, 26, and improves respiratory function.
The AAN recommends the introduction of NIV when orthopnea is present, SNP falls below 40 cm, MIP < −−60 cm, FVC < 50%, or nocturnal oximetry is abnormal 3. We have added to these criteria diurnal pCO2, which should be more than 6.0 kPa, as recommended for patients with chronic respiratory failure 11. Recommended limits for morning blood-gas pCO2 are slightly higher (more than 6.5 kPa) 18.
We have also confirmed the findings of several previous studies that longer survival is associated with younger age at disease onset 8 and the use of riluzole 9.
In conclusion, the symptoms and signs of respiratory insufficiency are important parameters associated with a shorter survival. An abnormal result of daytime AGA also indicates a bad prognosis, but this is generally a late sign. No single test is superior for the detection of impending respiratory failure. Symptoms, abnormal respiratory functional tests, and abnormal daytime AGA do not always coincide; they can appear in a different sequence or even be normal despite obvious respiratory problems. Any respiratory abnormality should therefore alert the clinician to the possibility of respiratory insufficiency and prompt him to discuss NIV with the patient, because NIV prolongs survival and improves respiratory function.
Acknowledgements
The authors thank Prof. Janez Zidar for review of the manuscript, and Dr Dianne Jones for language review. This work was supported by a research grant from the Slovenian Research Agency, No. P3-0338.
Disclosure of conflicts of interests
The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.
Statement on validity of statistical methods
The last author is a professional statistician, and he performed the statistical analyses.
