Elevation of Prostaglandin E₂ in Lung Cancer Patients with Digital Clubbing

Abstract

To test the hypothesis that digital clubbing in patients with lung cancer reflects elevated systemic levels of prostaglandin E₂ (PGE₂), we quantified levels of the major urinary PGE2 metabolite in 29 patients with lung cancer with or without clubbing. We found that digital clubbing was associated with markedly elevated urinary levels of PGE2 metabolite. These findings highlight a potential role for PGE2 in the pathogenesis of digital clubbing. Furthermore, this result identifies a lung cancer subpopulation that may uniquely benefit from cyclooxygenase inhibition.

The cause of digital clubbing has been a medical mystery since it was first recognized in the fifth century BC. Here we present evidence from a prospective study for the molecular pathogenesis of clubbing and discuss its potential therapeutic implications. Several lines of evidence suggest a role for the cyclooxygenase (COX) metabolite, prostaglandin E₂ (PGE₂), in the etiology of digital clubbing and hypertrophic osteoarthropathy (HOA). First, observations that HOA is commonly found in patients with pathologic shunting around the pulmonary circulation led to the hypothesis that a circulating factor, normally inactivated in the lung, is central to HOA pathogenesis. Prostaglandins fit this profile as they are subject to first-pass catabolism in lung by 15-hydroxyprostaglandin dehydrogenase.^1,2 Second, PGE₁ treatment leads to signs of HOA in infants with cyanotic heart disease and in adults with liver disease.^2,3 Third, primary HOA has recently been linked to inactivating mutations in the gene encoding 15-hydroxyprostaglandin dehydrogenase.² Finally, marked elevations in the major urinary PGE₂ metabolite, 11α-hydroxy-9,15-dioxo-2,3,4,5-tetranor-prostane-1,20-dioic acid (PGE-M), were recently reported in a lung cancer patient with HOA.¹ In this patient, symptoms of HOA, including debilitating bilateral leg pain, were completely relieved by the COX-2 inhibitor, rofecoxib, and promptly recurred with rofecoxib discontinuation.

PATIENTS AND METHODS

A prospective study of urine PGE-M levels in newly diagnosed, adult patients with stage III/IV lung cancer with and without digital clubbing was initiated at the Massachusetts General Hospital Thoracic Oncology Clinic. Twenty-nine patients were accrued. To limit pharmacologic confounding, patients who had taken aspirin, nonsteroidal antiinflammatories, or corticosteroids within 14 days, 3 days, and 30 days, respectively, were excluded from the study. Time restrictions were selected to reflect agent half lives. Similarly, patients with dialysis-dependent renal failure or systemic inflammatory illness were excluded. Clubbing was identified by clinical examination by a single physician with extensive experience managing patients with thoracic malignancies (K.R.K.). Examination included simple visual inspection and Schamroth’s test (Figure 1).⁴ Importantly, clubbing was coded before obtaining urine. The study was approved by the Partners Human Subjects Research Committee. Patients provided written informed consent. Urine was obtained before initiation of therapy. Urine was desedimented by centrifugation and stored at −78°C until analysis. Urine PGE-M levels were measured in duplicate as described previously.⁵

FIGURE 1.

Clubbing in a patient with lung cancer. Schamroth’s sign, or the obliteration of the normal window at the base of opposed nail beds, is evident in the lower panel.

RESULTS AND CONCLUSIONS

Patient characteristics, urine PGE-M levels, and distribution of clubbing among patient subgroups are presented in Table 1. Significantly elevated levels of urinary PGE-M were detected in patients with clubbing. The median urinary PGE-M level was 2.3-fold higher in patients with clubbing compared with patients without clubbing (data not shown). Statistically significant differences in urinary PGE-M levels were not detected when patients were segregated by smoking status, gender, cancer stage, or cancer histology. Fisher’s exact test testing revealed no statistically significant imbalances in smoking status, gender, cancer stage, or cancer histology between patients with clubbing and patients without clubbing. Age did not vary significantly between patients with or without clubbing (clubbing: 60 ± 4 years and no clubbing: 64 ± 3 years; p = 0.47, Wilcoxon rank sum test). Multivariate linear regression analysis with the logarithm of urinary PGE-M as the dependent variable yielded clubbing as the only significant covariate (p = 0.017); gender was of borderline significance (p = 0.07).

TABLE 1.

Patient Characteristics, Urine PGE-M (Average ± SE), and Clubbing Distribution among Patient Subgroups

	Number (%)	Urinary PGE-M (ng/mg Cr)	pa	Number with Clubbing	pb
Clubbing
Yes	10 (34)	62 ± 20	0.03	—	—
No	19 (66)	24 ± 5		—	—
Smoking status
Current	11 (38)	45 ± 18	0.41	6/11	0.11
Not current	18 (62)	33 ± 8		4/18
Gender
Male	8 (28)	60 ± 24	0.09	3/8	1.0
Female	21 (72)	29 ± 6		7/21
Cancer stage
III	15 (52)	23 ± 4	0.16	3/15	0.13
IV	14 (48)	53 ±15		7/14
Histology
Adenocarcinoma	17 (59)	48 ± 13	0.24	7/17	0.45
Other	12 (41)	22 ± 5		3/12

^aWilcoxon rank sum test.

^bFisher’s exact test.

PGE-M, prostaglandin E₂ metabolite.

This study affirms our hypothesis that clubbing in patients with lung cancer is significantly associated with elevated levels of PGE₂. In the context of the evidence described earlier in the text, these data suggest that PGE₂ may be causally linked to clubbing and HOA. Promising preclinical data prompted interest in COX-2 inhibitors in lung cancer therapy. Nevertheless, clinical trials demonstrated poor efficacy in unselected patients.⁶ Some evidence indicates that COX-2 inhibitors may specifically benefit those patients with lung cancers that express high levels of COX-2 (and thus produce large quantities of PGE₂). Our data suggest that simple visual inspection of the digits may identify patients with lung cancer more likely to benefit from COX-2 inhibitors.