Prognostic role of pneumonia in supracricoid and supraglottic laryngectomies

Introduction 

Partial laryngeal surgery in patients with squamous cell carcinoma aims to preserve the natural protective, respiratory and phonatory functions of the larynx in contrast to total laryngectomy where all these functions are lost.1 Partial excision of the larynx with alteration of its physiological mechanisms affects not only the phonatory but also the swallowing function with possible aspiration that can lead to life-threatening complications, mainly pneumonia.2, 3 Despite any attempt to preserve and reconstruct a normal physiology of the larynx, conservative approaches, such as supraglottic and supracricoid partial laryngectomies cause a consistent alteration of the laryngeal anatomy and physiology affecting laryngeal elevation, laryngeal closure and bolus progression. These impaired function need to be managed by postoperative swallowing rehabilitation.

The occurrence of PP represents a primary source of increased cost and morbidity not only after partial laryngectomy, but also after other major head and neck surgery, particularly with a tracheotomy, occurring in 20% or more of cases.4

The aetiology of PP following partial laryngeal surgery, however, is most likely multifactorial considering that several host and tumour factors such as age, obesity, poor preoperative pulmonary function tests, smoking history, type of surgery, extension of disease and blood gas analysis values might play a crucial role in predisposing for PP.5, 6

In the attempt to minimise PP onset patients eligible for partial laryngeal surgery are carefully pre-selected upon pulmonary and performance status, nevertheless PP remains the most important and feared complication even in this selected population, despite antibiotic prophylaxis.7

The aim of this study is to review our experience with supracricoid and supraglottic laryngectomies at our Institution over a 20 years period in a series of 416 consecutive patients, to investigate possible PP risk factors. Since several reports from other districts8, 9 suggest a possible PP prognostic significance, we also investigated the prognostic role of PP in our series.

Methods 

The charts of 416 consecutive laryngeal cancer patients that underwent a partial laryngectomy at the Otolaryngology Head and Neck Surgery Department of the University of Florence from 1980 to 2000 were retrospectively reviewed. We considered eligible for partial surgery laryngeal cancer patients with: Karnofsky physical index superior to 80%, no dispnea whilst climbing two flights of stairs, satisfactory pulmonary functional tests according to Morris and Koski,10 and basal pO2 values above 85mmHg. Furthermore, patients with severe cardio-vascular diseases, diabetes mellitus, gastro-oesophageal reflux, chronic obstructive pulmonary disease and rheumatoid arthritis were considered not eligible for partial laryngectomy.

Surgery included 309 supraglottic horizontal laryngectomies (SHL), 85 supracricoid laryngectomies with crico-hyoidopexy (CHP) and 22 with crico-hyoid-epiglottopexy (CHEP).

Clinical characteristics of the series with several clinical conditions potentially involved in the onset of PP are summarised in Table 1. All tumours were squamous cell carcinomas (SCC) classified using UICC TNM staging system.11 Of the 416 cases, 392 (94.2%) were previously untreated whilst 13 patients were treated as salvage procedure following a radiotherapy failure and 11 a laser CO2 failure. The mean follow-up time was 67.9 months (range 60–300 months).

Table 1. Series overview

	Series overview
	No. of patients	416
	Sex (M/F)	382/31
	Age at intervention
	Range	20–79
	Mean	57
	Median
	Body mass index (BMI)
	<18.5	10 (2.40%)
	18.5–25	190 (45.67%)
	25–30	167 (40.14%)
	30–40	49 (11.77%)
	>40	0
	Smoking history
	No smokers	10 (2.40%)
	Light smokers (<20 p-y)	25 (6.00%)
	Moderate smokers (20–40 p-y)	158 (37.98%)
	Heavy smokers (>40 p-y)	223 (53.60%)
	History of previous pneumonia
	No	287 (69%)
	Yes	129 (31%)
	Preoperative haemoglobin values
	⩽14g/dL	35 (8.5%)
	>14g/dL	381 (91.5%)
	Preoperative pO2 values
	⩽90mmHg	273 (65.62%)
	>90mmHg	143 (34.37%)
	Stage of disease
	I	132 (31.7%)
	II	154 (37%)
	III	66 (15.86%)
	IV	64 (15.38%)
	Type of surgery
	Supraglottic horizontal laryngectomies	309 (74.27%)
	Supracricoid laryngectomies with crico-hyoidopexy	85 (20.43%)
	Supracricoid laryngectomies with crico-hyoid-epiglottopexy	22 (5.28%)
	Neck dissection
	No	163 (31.18%)
	Elective	162 (38.94%) (114 MRND+48 SND)
	Therapeutic	91 (21.87%) (all MRND)
	Postoperative pneumonia	73
	Deaths	3 (4.10%)
	Early pneumonia	26 (35.61%)
	Ab ingestis pneumonia	44 (60.27%)

We defined a patient affected by PP if at least three of the following conditions were found: temperature above 37.5°C, purulent sputum with an increase in volume (>20%), positive blood and/or sputum culture, white cell blood count more than 10,000/mm3 or less than 3000/mm3, positive findings at chest X-rays, persistent localised signs on chest examination (crackles, bronchial breathing and wheeze).

For what concerns PP in our series we defined three groups of patients: Group 0 (G0), patients without PP; Group I (GI), patients developing PP within the first postoperative 7–9 days, before any attempt to achieve oral food intake; Group II (GII), those who developed PP because of aspiration at the time of achievement of oral intake. The postoperative period was considered as the period of time from the end of the surgical procedure until hospital discharge. To understand our grouping it is relevant to mention that our Institutional policy previews that all patients after surgery must be admitted to a swallowing rehabilitation program by a speech therapist, starting with tongue exercises from the third postoperative day; oral intake, always under assistance of the speech therapist, usually starts around the ninth postoperative day, in general never before postoperative day seven.

In our series antibiotic prophylaxis with 1g cefazolin intravenously was standard administrated. According to international guidelines this prophylaxis is reported as able to control infective complications like pulmonary infections, wound infections, urinary tract infections and septic phlebitis.7

Results 

We documented a PP in 73 cases (17.54%): early PP arose in 26 patients (GI: 6.25%), whereas ab ingestis PP developed in 44 cases (GII: 10.65%); three PP patients died in the postoperative period because of sepsis with respiratory failure (0.07%) and were excluded from our analysis. The remaining 343 patients (G0: 83.2%) were free from PP, amongst them 48 patients (13.99%) did experience postoperative major swallowing problems without PP. Overall the incidence of major postoperative swallowing problems is 95 out of 416 (22.83%).

In 42 of the 70 PP patients (60%) microbiological data of the sputum culture are available. The most common micro-organism detected was Staphylococcus aureus. Amongst them, multiple bacteria were isolated in 22 patients. According cultural analysis no differences between GI and GII were recorded for what concerns the type of micro-organism responsible for PP.

Overall, the postoperative complication rate, including pulmonary and non pulmonary complications (urinary tract infections, septic phlebitis, wound infections and postoperative bleeding), was 25.90% (107 out of 413). A higher incidence of non pulmonary complication was documented in PP patients (13 out of 70 vs. 37 out of 343) (Fisher’s exact test p=0.057).

Definitive functional unsuccessful results were recorded in only 13 cases out of 413 (3.14%). Amongst them, 10 were PP patients: seven underwent a total laryngectomy, three patients a permanent gastrostomy because of untreatable swallowing problems; the remaining three were free from PP but experienced laryngeal stenosis that prevented decannulation (Fisher’s exact test p<0.001).

The mean time of nasogastric tube placement was 15.3 days (range 36, median 14) for G0, 21.2 days (range 40, median 15) for GI, and 29.3 days for GII (range 52, median 20) (Fisher’s exact, p=0.11). The mean length of time before tracheotomy tube removal was 28.1 (range 485, median 22) for G0, 17.6 (range 205, median 31) for GI and 21.40 (range 84, median 33) for GII (Fisher’s exact; p=0.10). The mean discharge time of G0 (29.3 days; range 85, median 32) was significantly lower from those documented in GI and GII patients (38.39 days, range 61, median 37; 46.17 days, range 101, median 34, respectively) (Fisher’s exact; p=0.005).

Univariate and multivariate logistic regression analyses 

In Table 2, Table 3 logistic regression analysis according to the main prognostic factors for each type of PP are summarised.

Table 2. Univariate and stepwise multivariate analysis for two groups: no pneumonia and yes early pneumonia GI

	Early pneumonia 26 cases
	Variable	Univariate analysis			Multivariate stepwise analysis
		Odd ratio	Z	p Value	Odd ratio	Z	p Value
	Tobacco exposure
	No
	Yes	0.97 (0.55;1.70)	−0.10	0.91
	BMI
	<30
	⩾30	2.80 (1.22;6.44)	2.44	0.02	3.14 (1.29;7.59)	2.54	0.01
	Previous pneumonia
	No
	Yes	1.20 (0.46;3.13)	0.39	0.69
	Age (years)
	<60
	⩾61	2.18 (0.97;4.87)	1.90	0.05	2.49 (1.07;5.67)	2.13	0.03
	Sex
	Female
	Male	1.04 (0.23;4.62)	0.05	0.95
	Hb
	⩽14g/dL
	>14g/dL	2.01 (0.46;8.75)	0.93	0.35
	pO2
	⩽90mmHg
	>90mmHg	1.09 (0.49;2.42)	0.23	0.82
	Type of surgery
	Supracricoid surgery
	Supraglottic surgery	1.30 (0.55;3.09)	0.61	0.55
	Stage
	I–II
	III–IV
		1.22 (0.84;1.79)	1.08	0.28
	Neck dissection
	No
	Yes	0.87 (0.38;1.94)	−0.34	0.73

Table 3. Univariate analysis and stepwise multivariate analysis for two groups: no pneumonia and yes ab ingestis GII pneumonia

	Ab ingestis pneumonia 44 cases
	Variable	Univariate analysis			Multivariate stepwise analysis
		Odd ratio	Z	p Value	Odd ratio	Z	p Value
	Tobacco exposure
	No
	Yes	0.80 (0.53;1.21)	−1.02	0.32
	BMI
	<30
	⩾30	1.28 (0.68;2.38)	0.72	0.44
	Previous pneumonia
	No
	Yes	0.54 (0.28;1.04)	−1.83	0.07	0.56 (0.29;1.09)	−1.70	0.09
	Age (years)
	<60
	⩾60	1.51 (0.81;2.84)	0.31	0.19	1.37 (0.72;2.61)	0.98	0.32
	Sex
	Female
	Male	3.20 (1.34;7.65)	2.62	0.01	1.61 (0.92;2.82)	1.68	0.09
	Hb
	⩽14g/dL
	>14g/dL	0.69 (0.30;1.58)	−0.85	0.39
	pO2
	⩽90mmHg
	>90mmHg	0.73 (0.38;1.37)	−0.97	0.32
	Type of surgery
	Supracricoid surgery
	Supraglottic surgery	0.83 (0.39;1.75)	−0.48	0.962
	Stage
	I–II
	III–IV	1.11 (0.83;1.49)	0.73	0.47
	Neck dissection
	No
	Yes	2.37 (1.13;4.95)	2.31	0.01	2.52 (1.13;5.63)	2.26	0.02

In GI, BMI>30 and age>60 were significant prognostic factors in both univariate and multivariate analyses (p=0.02 and 0.05, p=0.01 and 0.03, respectively). In GII, patients receiving a neck dissection were at higher risk of ab ingestis PP in both analyses (p=0.01 and p=0.02, respectively). Smoke habits, history of previous pneumonia, haemoglobin values and pO2 blood levels, TNM Stage and type of reconstructive surgery were not significantly associated with the onset of GI or GII pneumonia.

Prognostic impact of PP on disease-free and actuarial survival 

We evaluated the prognostic impact of PP (GI and GII) in terms of disease-free and actuarial survival. These results are summarised in Table 4, Table 5. According to disease-free survival, tumour stage at diagnosis, haemoglobin blood level under 14g/dL and early pneumonia were statistically significant, whilst type of surgery, history of previous pneumonia and pO2 blood gas level under 90mmHg lost their statistical significance in multivariate model. According to 5 years actuarial survival, univariate analysis indicated analysis tumour stage (p<0.001), history of previous pneumonia (p=0.010) and early-PP (p=0.001) as prognostic factors; however, only advanced tumour stage and early-PP maintain their prognostic value in multivariate analysis.

Table 4. Statistical analysis for disease-free survival

	Disease-free survival
	Variables	Univariate analysis				Stepwise multivariate analysis
		Score	HR (95% CI)	Z	p	HR	Z	p
	Surgery
	SHL	0
	CHP	1	0.50 (0.30;0.84)	−2.61	0.009	0.79 (0.47;1.35)	−0.83	0.408
	CHEP	2	0.75 (0.33;1.72)	−0.67	0.504	0.98 (0.39;2.44)	−0.04	0.970
	Sex
	Male	0
	Female	1	0.91 (0.46;1.79)	−0.27	0.788
	Age (yr)
	<60	0
	⩾60	1	1.03 (0.72;1.47)	0.17	0.815
	Stage
	Stage I–II	0
	Stage III–IV	1	2.67 (1.88;3.79)	5.53	0.000	4.41 (2.33;8.35)	4.56	0.000
	Previous pneumonia
	No	0
	Yes	1	0.62 (0.43;0.88)	2.61	0.010	0.70 (0.48;1.02)	1.85	0.065
	Other complications
	No	0
	Yes	1	1.55 (0.85;2.82)	1.46	0.120	1.17 (0.36;3.72)	0.27	0.790
	Pneumonia
	No	0
	Early-GI	1	2.02 (1.08;3.78)	2.23	0.026	1.87 (1.00;3.49)	1.97	0.049
	Ab ingestis-GII	2	1.39 (0.83;2.32)	1.26	0.207	1.40 (0.83;2.35)	1.30	0.195
	Smoke (py)
	No smokers	0
	<20	1	3.92 (0.84;18.3)	1.74	0.076
	20<py<40	2	1.71 (0.39;7.38)	0.72	0.469
	>40	3	1.84 (0.42;7.92)	0.82	0.411
	Hb
	⩽14g/dL	0
	>14g/dL	1	3.29 (1.04;10.3)	2.04	0.012	1.99 (1.01;17.4)	1.99	0.047
	pO2
	⩽90mmHg	0
	>90mmHg	1	0.58 (0.39;0.87)	−2.60	0.006	0.62 (0.41;0.95)	−2.17	0.030
	BMI
	<30	0
	⩾30	1	0.88 (0.56;1.39)	−0.52	0.060

Table 5. Statistical analysis for actuarial survival

	Actuarial survival
	Variables	Univariate analysis				Stepwise multivariate analysis
			HR	Z	p	HR	Z	p
	Surgery
	SHL	0
	CHP	1	0.58 (0.30;1.11)	−1.64	0.101	0.75 (0.20;2.71)	−0.44	0.663
	CHEP	2	1.34 (0.58;3.10)	0.69	0.488	0.96 (0.19;4.88)	−0.04	0.970
	Sex
	Male	0
	Female	1	1.21 (0.55;2.64)	0.49	0.621
	Age (yr)
	<60	0
	⩾60	1	1.05 (0.67;1.66)	0.25	0.803
	Stage
	Stage I–II	0
	Stage III–IV	1	2.70 (1.93;4.30)	5.24	0.000	3.57 (2.30;5.55)	5.67	0.000
	Previous pneumonia
	No	0
	Yes	1	0.53 (0.34;0.83)	2.76	0.006	0.82 (0.51;1.33)	0.78	0.437
	Other complications
	No	0
	Yes	1	1.47 (0.68;3.19)	0.99	0.321
	Pneumonia
	No	0
	Early-GI	1	3.29 (1.68;6.44)	3.48	0.001	14.0 (2.49;78.6)	3.00	0.003
	Ab ingestis-GII	2	1.87 (1.02;3.41)	2.04	0.041	1.53 (0.24;9.65)	0.46	0.648
	Smoke (py)
	No smokers	0
	<20	1	1.22 (0.29;5.07)	0.28	0.776
	20<py<40	2	0.67 (0.19;2.37)	0.61	0.544
	>40	3	0.74 (0.21;2.58)	0.46	0.642
	Haemoglobin
	⩽14g/dL	0
	>14g/dL	1	1.30 (0.48;3.49)	0.52	0.602
	pO2
	⩽90mmHg	0
	>90mmHg	1	0.68 (0.36;1.25)	−1.23	0.219
	BMI
	<30	0
	⩾30	1	1.02 (0.59;1.57)	0.10	0.923

The Kaplan–Meier estimated disease-free and actuarial survival is shown in Figure 1A and C. Accordingly, early-PP was an adverse prognostic factor in terms of both disease-free and actuarial survival (Fig. 1B and D) (log-rank test p=0.030 and p<0.001, respectively).

View Large Image Download to PowerPoint

Figure 1. General and early-PP related Kaplan–Meier survival curves for disease-free survival (A and B) and actuarial survival (C and D).

Discussion 

During deglutition, the larynx elevates under the tongue base and moves forward placing the epiglottis to protect the laryngeal inlet, this movement stretches the crico-pharyngeal region contributing to upper oesophageal sphincter opening. Airway protection is guaranteed by vocal cords and arytenoids adduction. Any surgical procedure that significantly affects these mechanisms, like partial laryngectomy, usually results in some degree of aspiration during the swallow with patients always requiring postoperative rehabilitation.

In supraglottic laryngectomy, the removal of the larynx above the true vocal cords and the hyoidopexy reducing the distance between base of tongue and glottic region is responsible for early swallowing impairment despite the maintenance of laryngeal closure valve.

In supracricoid laryngectomy both vocal cords are removed with only preservation of one or both arytenoids, thus, directly affecting laryngeal closure. In partial laryngectomies, oedema of the arytenoids may also contribute to laryngeal closure failure. Furthermore, food might collect around the arytenoids thus, increasing the risk of aspiration.12, 13

To minimise PP, elderly patients and those with poor pulmonary function tests are considered not eligible for conservative laryngeal surgery because of the high risk of aspiration, due to a decreased ability to cough and clear secretions. Despite this selection, PP still represents the most frequent potentially life-threatening complication in partial laryngectomized patients, requiring more intensive care, delay rehabilitation and discharge from the hospital and inevitably leading to greater health care costs.3

In our series, PP had an incidence of 16.8%. Being postoperative deaths confined amongst PP patients with peri-operative mortality in this group of 4.2%, thus highlight the clinical relevance of such a complication. Moreover, we documented a statistically significant higher rate of non pulmonary complications in the PP group (18.84%). Furthermore, most of all definitive functional unsuccessful results were PP patients.

Together with the impairment of laryngeal functions, other factors related to tumour stage, type of surgical resection and host clinical conditions have been reported as additional risk factors for PP, thus suggesting a multifactorial aetiology.5, 6

It is well known that patients who require tracheotomy after major head and neck surgery (always performed during reconstructive laryngectomies) experience a higher incidence of pulmonary complications. Data from the literature indicate that almost 50% of these patients suffer of pulmonary complications including PP, early after surgery despite a course of antibiotic prophylaxis,4 with elderly subjects and patients with co-morbid respiratory disease being at highest risk.

Interestingly, the analysis of our data evidences that PP not only develops as a result of food aspiration following the attempts to restore oral food intake, but can arise within the first 7–9 postoperative days, before any swallowing rehabilitation. According to this finding we divided PP in two separate groups: GI enclosing early PP patients, GII those with ab ingestis PP.

Being patients undergoing partial laryngectomy pre-selected, we postulated that the occurrence of an early PP might have a different pathogenesis and prognostic impact when compared to the ab ingestis one. Theoretically, early PP could be linked to aspiration of blood, saliva and secretion during and immediately after surgery, events occurring in every patient. The low rate of early PP documented in our series, conversely, raises the question of its pathogenesis and potential prognostic significance. Accordingly, we evidence that early PP is sustained by different factors than ab ingestis PP. Age greater than 60 years and BMI greater than 30 were the prognostic factors linked to early PP (p=0.03 and p=0.01, respectively). On the other hand, partial laryngectomy with neck dissection statistically predicted ab ingestis PP (p=0.02).

It is noteworthy that factors influencing the onset of early PP are strictly host related, whilst ab ingestis PP seems to be related to type of surgical procedure, possibly affecting postoperative abilities that are crucial for swallowing rehabilitation.

Then, we analysed the prognostic impact of these complications by Cox univariate and multivariate tests. According to disease-free survival, univariate analysis indicates as significant the following factors: type of laryngectomy (supraglottic laryngectomy vs. supracricoid, p=0.010), stage III–IV of disease (p<0.001), history of previous pulmonary disease (p=0.010), occurrence of early PP (p=0.026), preoperative haemoglobin values less than 14g/dL (p=0.042), preoperative pO2 values less than 90mmHg at blood gas analysis (p=0.006). In the stepwise multivariate analysis, stage of disease (p<0.001), occurrence of early-PP (p=0.049), preoperative haemoglobin values less than 14g/dL (p=0.047) and preoperative pO2 values less than 90mmHg at blood gas analysis (p=0.030) maintained their prognostic relevance.

At variance, actuarial survival analysis suggest that clinical evidence of nodal disease (p<0.001), stage III–IV of disease (p<0.001), previous pulmonary disease (p=0.006), occurrence of early (p=0.001) and ab ingestis PP (p=0.041) were negative prognostic predictors in univariate analysis, whilst in multivariate tests only advanced stage (p<0.001) and occurrence of early PP (p=0.003) maintained its significance. The key prognostic role of early PP is confirmed by the Kaplan–Meier curves analysis (Fig. 1), overall suggesting that those patients who experienced early PP have a higher risk of cancer failure (log-rank test p=0.030) with poor prognosis (log-rank test p<0.001).

The fact that PP is predictive of poor outcome also in surgery of other districts,14, 15 supports the hypothesis that host immune system might be potentially involved in host cancer response. Data from the literature16, 17 indicate that head and neck cancer patients, also because of poor general condition due to tobacco/alcohol exposure, show host immune system impairment able to affect cancer prognosis. Accordingly, here, to the best of our knowledge, in patients who underwent partial reconstructive laryngectomies, we first describe early PP and report its negative impact in terms of disease-free and actuarial survival. Of course, only additional analyses evaluating host immune response in patient with this new clinical entity might address this issue and is currently under investigation at our Institution.

In conclusion, our study indicates a prognostic significance of early-onset pneumonia when compared with a classical late-onset pulmonary infection; if this evidence will be confirmed also by further investigations it should be wise to pay particular attention to the postoperative course of such patients at risk with a stricter follow-up schedule, and maybe these patients might be eligible for adjuvant treatment.