Volatile Organic Compounds of Malignant Breast Cancer Wounds: Identification and Odors

Introduction. During the metabolic processes of malignant wounds, bacteria produce a large amount of volatile organic compounds (VOCs) that are responsible for malodors and may have a major impact on the patient’s quality of life with a risk of isolation. Objective. A translational study was conducted on 32 malignant breast wounds by combining the identification of bacterial strains present on wounds, the identification of VOCs produced by these bacterial strains, and sensory evaluation to assess odor intensity and quality of odorous bacteria. Materials and Methods. Thirty-two patients with malignant breast cancer wounds > 10 cm² at various stages of the disease (curative or palliative) were included in the protocol. Volatile organic compounds were collected from primary dressings by headspace solid-phase microextraction and then analyzed by gas chromatography separation coupled with a mass spectrometer detector analysis. Microbiological samplings were taken and analyzed on agar plates. The odors of selected bacteria were assessed by a panel of staff members. Results. Proteus mirabilis and Fusobacterium necrophorum seem to produce the strongest and most typical malignant wound odor. The VOCs were analyzed and dimethyl disulfide, dimethyl trisulfide, phenol, indole, and 3-methylbutanal were found to be produced by bacteria generating the most typical wound odor. Conclusions. This study suggests the bacteria present in wounds may be responsible for odors. In addition, these findings could pave the way to engineer new types of dressings and to develop an evaluation method to assess their efficiency both quantitatively and qualitatively as well as improve quality of palliative care and comfort for women with malignant wounds.

Malignant wounds occur when tumor cells infiltrate cutaneous tissue, blood, or lymph vesselsand are often associated with malodors.^1-3 Indeed, this progression can lead to tissue hypoxia and the formation of necrosis, providing an ideal place for anaerobic bacteria to grow, to which the presence of odors is attributed.⁴ Malodors associated with necrotic wounds strongly affect patients, contributing to their social life and their social isolation.⁵ Thus, a better understanding of the etiology of this symptom and its effects is of major interest to further clinical practices.

Odors are the result of volatile organic compounds (VOCs), which are presumably produced by microorganisms during their metabolic processes.⁶ Shirasu et al⁷ described VOCs originating from malodorous, malignant wounds and concluded the sulfurous odor, which seems the most disturbing, is likely linked to the dimethyl trisulfide (DMTS) produced by aerobic bacteria, such as Pseudomonas aeruginosa, without in any way underestimating the impact of anaerobic bacteria. However, this study⁷ was carried out on a restricted number of patients (5) with wounds on numerous sites (breast, ear, nose, and throat).

The aim of the present study is to characterize malignant wound odors associated with breast cancer and determine their origin by extending this study to a substantial number of patients (32). The study was conducted by systematically performing a combination of analyses: sensory analysis, identification of bacterial strains, and chemical analysis by gas chromatography. Until now, the direct assessment of wound odors^8,9 has been limited to the assessment of odor intensity performed by 1 person (usually a member of the nursing staff). Therefore, sensorial perceptions were subject to each assessor’s personal sensitivity. In this study, the authors use sensory analysis to reliably describe odors associated with malignant wounds and the identified bacteria.

The identification of VOCs in odorous and odorless wounds and VOCs produced by various bacterial strains is performed using an analytical protocol that combines an extraction of VOCs by headspace solid-phase microextraction (HS-SPME) followed by gas chromatography separation coupled with a mass spectrometer detector (GC-MS).¹⁰

Microbiological analyses were carried out to identify bacterial strains in malignant wounds and their relationships with the presence or absence of odors. The 10 most frequent strains are presented herein.

This study started during medical consultations with the collection of used dressings and patients’ assessment of odors. The dressing samples were then analyzed to identify strains and the present VOCs. The 8 most malodorous were placed into culture and described by a panel of the nursing staff.

Patient consent and ethical aspects
Patients were informed of the study conditions and written informed consent was obtained. Each patient then was assigned a study number. All data and results were compiled into a unique database. The study was carried out with full respect of the ethics that apply in this area and the current regulations. The People’s Protection Committee approved the study protocol.

Collection of dressing samples and clinical data
Regarding the sample group, 32 patients with malignant breast cancer wounds > 10 cm² at various stages of the disease (curative or palliative) were included in the protocol, providing 3 visits on days 0, 21, and 42 during consultation. In light of the particular context regarding the management of oncology patients, mostly under a palliative care situation, the initially planned follow-up could not be fully respected. Therefore, 32 patients were involved in the protocol on day 0, 28 patients on day 21, and 18 patients on day 42. Among the patients not in all 3 consultations, 3 died, 1 was transferred to a palliative care unit, and 4 did not attend 1 of the appointments for psychosocial reasons. Moreover, some wounds improved with partial epithelialization, which did not allow for sampling as that could have been traumatic and deleterious for the progression of the wound.

During each consultation, a used primary dressing sample (that had direct contact with the wound) was collected. A piece of used primary wound dressing, about 1 cm² in size, was introduced in the 20-mL SPME vial and kept at -80°C for further analysis. Dressings moistened with water to enable their removal or dressings associated with heavy bleeding were excluded, leading to a total of 62 samples.

Following removal of the primary dressing and wound cleanse with physiological saline solution, microbiological samplings were taken with a single-use curette (Volkman curette, 50 mm), calibrated on the deepest and/or most necrotic zone. Biological samples were brought to the laboratory without delay for bacteriological analysis. Overall, 60 samples were collected.

Finally, the odor emanating from the wound was assessed at 2 stages. Before removing the dressing, a randomly selected member of the nursing staff rated the wound odor and reported if, in his/her view, the odor was controlled by the dressings (no perceptible odor), partially controlled (weak odor), or uncontrolled (strong odor). After removing the dressing, the patient and the nurse rated the intensity of the wound odor using a 4-point scale: absent, weak, moderate, or strong odor.

Microbiological analysis
For each biological sample, the detection of aerobic, facultative anaerobic, or strict anaerobic bacteria was carried out in a quantitative way. On arrival at the laboratory, samples were immediately incubated in brain heart infusion (BHI) broth and then diluted 1:10 and/or 1:100 in BHI depending on the clinical context. Then, 1-µL seeding of each broth was realized on different agar media:

• 1 chocolate agar (incubated under CO₂)
• 2 Columbia CNA agars (1 incubated under CO₂ and 1 incubated anaerobically)
• 1 Drigalski agar (incubated aerobically)
• 1 CPS agar (incubated aerobically)
• 1 Chapman agar (incubated aerobically)
• 1 Pseudomonas agar (incubated aerobically)
• 1 Staphylococcus aureus Elite agar (incubated aerobically)
• 1 Schaedler agar (incubated anaerobically)
• 1 Schaedler NeoVanco agar (incubated anaerobically)
• 1 Sabouraud Genta Chloramphenicol test tube

The agar plates were incubated at 37°C for 1 to 5 days. All plates were analyzed and all bacteria were identified and numerated as aerobic, facultative anaerobic, and strict aerobic flora.

Identification of VOCs from primary dressings by GC-MS
Volatile organic compounds were collected from the primary dressings by HS-SPME and then analyzed by GC-MS analysis.

Prior to extraction, SPME fibers were conditioned before each HS-SPME analysis. The used primary dressing or the bacterial cultures were prepared in the 20-mL SPME vial. The fully automated HS-SPME procedure was as follows.

First, the vial was equilibrated at 50°C for 10 minutes before the SPME fiber (Carboxen-PolyDiMethylSiloxane 75 µm [Supelco-Sigma-Aldrich, Saint Quentin Fallavier, France]) was placed into the HS of the sample for extraction and maintained at 50°C for 30 minutes. Analyses were performed on an Agilent GC 7890A and MS 5975C (Agilent Technologies, Les Ulis, France) equipped with a MPS autosampler Gerstel (RIC, Saint-Priest, France). The analytical column DB5-MS (30 m x 0.25 mm, 1-µm film thickness [Agilent Technologies]) was used with helium as carrier gas in constant flow mode at 1 mL/minute.

Then, the SPME fiber was desorbed directly in the GC injector for 10 minutes at 260°C in splitless mode. The temperature program was 40°C held for 3 minutes then raised to 280°C at 10°C/minute and held for 3 minutes. The mass spectrometer was scanned from m/z 20 to 300 and samples were analyzed using electronic ionization using 70 eV. The identity of each volatile compound was confirmed by comparing their retention time and mass spectra with those of pure standard compounds (Sigma-Aldrich) after GC-MS analysis.

Sensory evaluation
The odors of the 8 selected bacteria were assessed by a panel of 32 staff members (randomly selected nurses and doctors; 28 women, 4 men; age range, 26–50 years) at the Institut Curie (Paris, France) accustomed to taking care of malignant wounds. None reported major olfaction impairment (acute or chronic rhinitis or airway obstruction).

The bacterial cultures were grown in 3 culture mediums: tryptic soy broth (TSB), BHI, and combined TSB + BHI. Then, bacteria were dish-cultured on agar gel over a period of 3 days at 37°C. On the day of assessment, the petri dishes were placed in inhalers (COOPER, Melun, France) and sealed using a plastic film. The closed inhalers were maintained for 30 minutes before the start of the test so that the volatile compounds can disperse in the gaseous phase, enabling a good assessment of the odor.

Participants tested the 8 samples in random order. For each sample, they rated the odor intensity on an 8-point scale from 0 (weak) to 7 (strong). Then, they described the odor by choosing up to 2 attributes among the following: putrid, excrement-like, pestilential, rotten, urine-like, and vegetal. Lastly, participants assessed the odor typicality by answering the yes/no question: “Is this odor typical of a malignant wound?”

Analysis of VOCs produced by bacterial strains by GC-MS
Among the 54 bacteria identified from biological samples of tumoral wounds, 8 were selected as the most odorous: Escherichia coli, P aeruginosa, Alcaligenes faecalis, Proteus mirabilis, S lugdunensis, Acinetobacter baumanii, Porhyromonas asaccharolytica, and Fusobacterium necrophorum. Based on the bacterial flora present in the samples of the most odorous wounds, these bacteria were selected from the literature^11,12 and informal odor assessments performed by the laboratory technicians in charge of analysis.

The analysis of VOCs produced by 4 of these 8 bacteria was conducted using GC-MS (P mirabilis, F necrophorum, P aeruginosa, and E coli). Proteus mirabilis and F necrophorum were selected based on the results of the sensory analysis because they produced an intense odor typical of malignant wounds. Pseudomonas aeruginosa and E coli were selected because they are responsible for inducing VOCs (DMTS) commonly present in wounds.⁷ The bacterial cultures were grown in TSB or BHI culture medium, according to the bacterial strain, over a period of 3 days at 37°C. The petri dish was not sealed with a parafilm because the work was done under a sterile hood. The strains studied were P mirabilis (TSB), F necrophorum (BHI), E coli (TSB), and P aeruginosa (TSB). The analysis was carried out within 24 hours under the same conditions as those outlined in the previous microbiological analysis subsection.

Wound odors
The frequency of odor intensities perceived by patients and staff members at each of the 3 sample collection dates is shown in Figure 1. At day 0, 75% of wounds were found to present weak to moderate odors and no wound was assessed as having a strong odor. Between days 0 and 21, wound odor intensities increased; the percentage of weak intensity decreased, whereas the percentage of moderate and strong intensity significantly increased (Khi² = 12.2; P = .007). At this point, wounds with a strong odor were treated using systemic antibiotics (metronidazole). As expected, it induced a reduction in the overall intensity of wound odors, with still 75% of wounds having weak to moderate odors between days 21 and 42.

Figure 2 indicates the percentage of wounds for which the odor was absent (controlled), weak (partially controlled), or moderate to strong (uncontrolled); assessment was executed by the nurse during care provision. Although no strong odor was reported at days 0 and 42 (Figure 1), the odors were not fully controlled by local treatments (dressing or activated charcoal), ie, some odors were perceived by the nursing staff before removing the dressing. This highlights the fact that even weak or moderate odors can be difficult to treat and control and can potentially be an issue for patients.

Characterization of bacterial flora and link with wound odor
Bacterial flora analysis following wound sampling proved to be predominantly polymicrobial. Fifty-four different bacteria were identified on the wounds, with 37 aerobic and facultative anaerobic bacteria and 17 strict anaerobic bacteria. The 10 most frequently identified bacteria are listed in Table 1. They are mainly commensal, pathogenic, or opportunistic pathogenic germs.

Table 2 shows that if strict anaerobic bacteria are present, the percentage of odorous wounds (weak, moderate, or strong intensity) is substantially higher (Khi² = 6.78; P = .009) than when they are absent. Therefore, wound odor could constitute an indicator of wound microbiota imbalance.

Identification of VOCs on used dressings
From the 62 used primary wound dressings analyzed, 20 corresponded to odorless wounds, 26 to wounds with weak odor, and 16 to wounds with moderate odor (Table 3). A total of 88 different VOCs were identified from wound dressings by GC-MS. An example of VOC identification is given in Figure 3. Samples from odorless wounds had an average of 7 volatile compounds per sample, those with a weak odor had an average of 9 volatile compounds per sample, and those with a moderate odor had an average of 11 volatile compounds per sample. The increase in odor intensity seems to be associated with the presence of a greater number of volatile compounds (Table 3). Five main VOCs were found on more than 40% of samples (Table 4^7,13,14): dimethyl disulfide (DMDS), DMTS, phenol, indole, and 3-methylbutanal.

Intensity and description of odors by sensory evaluation
Thirty-two staff members participated in sensory evaluation on the 8 most odorous bacteria (E coli, P aeruginosa, A faecalis, P mirabilis, S lugdunensis, A baumanii, P asaccharolytica, and F necrophorum). The assessment of odor intensity shows that all bacteria develop odors but with a wide range of different intensities (Table 5).

The odor descriptions were summarized in an 8 (bacteria) x 6 (attributes) contingency matrix, where the number of occurrences of each descriptive term was reported for each bacteria. The matrix was analyzed with correspondence analysis (CA).

Figure 4 represents the subspace 1-2 of the CA map. The first axis, representing about 66% of the total variance, opposes the bacteria E coli, A faecalis, and P aeruginosa, which were described as vegetal/urine-like, to the other bacteria described as rotten/putrid/pestilential. The second axis, representing about 21% of the total variance, opposes P asaccharolytica, which was described as excrement-like, to the other bacteria. The first axis seems to represent a hedonic axis with very unpleasant odors on the right side and moderately unpleasant to neutral odors on the left side. Odor intensity is related to its unpleasantness; the more intense, the more unpleasant. The coordinates of the bacteria on the first axis are significantly correlated with their normal intensity (r = 0.97; P < .001).

Table 5 indicates that bacteria such as A baumanii, P asaccharolytica, F necrophorum, and P mirabilis produced odors typical of malignant wounds contrary to the others (E coli, P aeruginosa). Overall, the most typical odors also are the most intense. Both F necrophorum and P mirabilis produced strong and typical odors. The majority of participants described those odors as rotten/putrid/pestilential (Figure 4).

Although its median intensity (5) is not the highest, P mirabilis showed the highest typicality rating (67%). Conversely, F necrophorum and P asaccharolytica had the highest odor intensity but a slightly lower typicality, although not significant (Khi² = 0.67; P > .05). Furthermore, P mirabilis and F necrophorum produced odors described in a similar way (Figure 4). Thus, one hypothesis is that they, when concurrently present, may produce an odor of both high intensity and high typicality.

Identification of VOCs specific from strain bacteria
Four strains (P mirabilis, F necrophorum, P aeruginosa, and E coli) were selected from the results of the sensory analysis due to their production of an intense typical odor of malignant wounds. Table 6 outlines the most frequently identified VOCs in these 4 strains under examination. Indole was detected for the 4 bacteria. Both F necrophorum and P mirabilis produce a greater amount of VOCs (from which DMDS and DMTS are detected) than P aeruginosa and E coli. The compounds 3-methylbutanal and phenol also were identified.

In this study of 32 patients with malignant breast cancer wounds, the investigators highlighted a variety of microbiota present and the link between these bacteria and VOC-related odors. In simultaneously combining the analysis of microbiota and the identification of VOCs from evolving malignant wounds, these results suggest the bacteria present on the wounds were probably responsible for odors. Proteus mirabilis was the most significant bacteria associated with the presence of odor, as shown by the sensory assessments of typicality.

Methyl sulfides (DMDS and DMTS) appear to play a large role in the unpleasantness of malignant wounds. Shirasu et al⁷ associated this extremely disturbing feature with the presence of volatile organosulfur compounds, particularly DMTS. In the present study, the odors appear to be due to 4 of the 10 mostly identified volatile compounds (DMDS, DMTS, indole, and phenol) (Table 4^7,13,14). However, the production of DMDS and DMTS is frequently cited in the scientific literature as the main VOCs emitted by the decomposing body.¹⁵ As a result, there seems to be a strong correlation between the presence of methyl sulfides (alone or in combination) and a strong smell of putrefaction. Moreover, a combination study of GC-MS and sensory perception analysis concludes that DMTS and DMDS are the compounds that induce the unpleasantness of stool odor.¹⁶ All participants herein did not agree on the typical fecaloid odor of these components, but all agreed on their foul and putrid characteristics.

The smell of corpses, as those of stool, is strong in symbolism and has an important cultural anchorage that may explain the repulsive effect expressed towards patients with malodorous tumoral wounds (eg, social isolation and rejection) and inappropriate measures of protection occasionally undertaken by the nursing staff caring for patients. For instance, this feeling of repulsion was expressed during the current sensory evaluations conducted on bacterial odors. For some samples, the current participants expressed their repulsion by their attitude or comments (oral and written). The investigators also question the possible neuropsychic effects of volatile compounds such as carbon disulfide. Exposure to toxic compounds at very low doses, but continuously and during a period of fragility, may insidiously cause changes in neurophysiological and neuropsychological functions. Tumoral wounds of the breast are located on the upper part of the thorax, thus, the patient inhales the scent continuously.

In light of these results, it is relevant to develop tools to assess wound odors in clinical practice that take into account both quantitative and qualitative aspects and to study the psychosensory or neurophysical impact of these odors on patients, their relatives, and their friends. These tools also would be useful to assess the efficiency of dressings.

Up until now, dressings containing charcoal with very low selective absorbency are advised in the first-line treatment of malodorous wounds.¹⁷ They act as an odor filter, even though they are not effective in terms of disease etiology.¹⁸ The results herein confirm these dressings have limited effectiveness to treat wound odors. The prescription of metronidazole, an antibiotic active against anaerobic bacteria and also the best-documented treatment, has convincing results in the treatment of odor symptoms.^14,19 However, resorting to antibiotics outside an episode of infection is a questionable alternative due to the risk of developing bacterial resistance to antibiotics, their impact on the bacterial colonization on the wound,²⁰ and the potential side effects.

For some years, alternatives have emerged that show an efficiency to this olfactory phenomenon. A wide range of anti-odor treatments (both conventional and alternative) has been tested in response to this problem, such as the application of essential oils,²¹ yogurt,²² turmeric,²³ or chlorophyll.²⁴ However, no investigation on the efficacy of these treatments has been conducted yet. Those studies could provide innovative solutions to improve quality of palliative care and comfort for patients with malignant wounds.

This transdisciplinary study has limitations as it is quite a challenge to mix disciplines such as microbiology, wound care, sensory perception, and analytical chemistry.

Given the small number of patients for a large number information (clinical, chemical, and microbiological), the results of this study should be interpreted with caution. Eight months after the end of the study, 27 of the 32 included patients died, highlighting the advanced stage of their disease and the effort to participate to a clinical research protocol during this period of life.

The identification of bacteria by polymerase chain reaction may have allowed to have more complete microbiological results such as the detection of bacteria present in minute amounts in samples, bacterial virulence genes, or detection of resistance by genetic methods.

In this study, the investigators found that anti-odor dressings may use 2 parameters to increase their efficiency. The first is their ability to adsorb VOCs (eg, with charcoal), thus lowering the malodorous VOCs emitted; yet, even at low intakes, the smell can disturb the patient and caregiver and be repulsive. The second parameter is to add a fragrance to the dressing that would cover the malodorous VOCs and hence produce a more pleasant odor; the idea is to lower its intensity while changing its odor (ie, a bouquet of VOCs) with a smell likely to suit everyone regardless of culture. The combination of these 2 parameters would drastically increase patients’ well-being. The efficiency of this combination could be assessed with tools that would take into account quantitative and qualitative aspects of the dressings. The establishment of a study around the volatile compounds of tumoral wounds and their possible consequences on behavior and psyche could be conducted with the collaboration of psycho-oncologists and neuropsychiatrists. To achieve this, new sensory perception and characterization studies of VOCs from different groups of patients with tumoral and nontumoral wounds would be necessary, in addition to clinical evaluations (neurophysiological, neuropsychological, and wound).

The authors wish to thank Aurélie Baffie, RN, Marie-Christine Falcou, Clinical Research Associate, Alexandra Rivat, RN, and Souhir Neffati, MS, of Institute Curie (Paris, France) for their support, as well as their partner, Nursing Research Hospital Program (Directorate of Health Care Supply, Paris, France).

Authors: Aurelie Thuleau, Ing¹; Jose Dugay, PhD²; Catherine Dacremont, PhD³; Zaineb Jemmali, MD²; Jacqueline Elard¹; Yann De Ricke, PhD⁴; Nathalie Cassoux, MD, PhD¹; Sarah Watson, MD, PhD¹; Marie-Christine Escande, PhD¹; and Isabelle Fromantin, RN, PhD^1,5

Affiliations: Institut Curie, Paris, France; ESPCI Paris Tech, Paris, France; French National Center for Scientific Research (CNRS), Paris, France; Ensemble hospitalier de l’Institut Curie, Paris, France; and Equipe d’accueil Clinical Epidemiology and Ageing (CEpiA), Université Paris Est Créteil, Créteil, France

Correspondence: Isabelle Fromantin, RN, PhD, Institut Curie, Anesthesia and Reanimation, 26 rue d’Ulm, 75005 Paris, France; isabelle.fromantin@curie.fr

Disclosure: This research was supported by the Hospitalization and Care Organization Division of the Minister of Health (France).