Identifikasi Pola Volatile Organic Compound (VOC) Menggunakan Electronic Nose (E-Nose) pada Pasien Kanker Servik Sebelum dan Sesudah Terapi

Selma Mutiara Hani, dr. Muhammad Ary Zucha, Sp.OG, Ph.D;dr. Dian Kesuma Pramudya Nurputra, Ph.D., M.Sc., Sp.A.

2023 | Tesis | S2 Ilmu Kedokteran Dasar dan Biomedis

Latar Belakang: Metode pemeriksaan biomarker volatile sudah banyak dikembangkan karena lebih nyaman, murah dan tidak invasif namun tetap reliable. Dalam perjalanan kanker serviks, monitor respon terapi dapat dilihat melalui biomarker pada darah atau urin dan dapat bermanfaat sebagai modalitas biomarker prognosis baru untuk menurunkan mortalitas kanker servik di Indonesia

Tujuan: Mengidentifikasi VOC (komponen aromatik, alkana, aldehida, asam, keton, alkohol dan ester) pada sampel urin pasien kanker serviks sebelum terapi dan sesudah terapi.

Metode: Penelitian observasional dengan pendekatan case-control menggunakan sampel urin pasien kelompok sebelum terapi dan sesudah terapi. Sampel urin diambil 2 kali (first void dan mid stream urine) di waktu yang berbeda, kemudian dianalisis pola VOC menggunakan electronic nose. Terdapat 7 sensor gas pada e-nose kemudian dilakukan analisis distribusi data dan metode klasifikasi menggunakan XGBOOST. Analisis statistik menggunakan Principal Component Analysis (PCA) dan Linear Discriminant Analysis (LDA) untuk melihat pola persebaran antara 2 kelompok. Evaluasi dengan kurva ROC dan analisis SHAP untuk melihat perbedaan prediksi antara kedua kelompok

Hasil: Sepuluh pasien sebelum dan sesudah terapi masing-masing memiliki rerata usia yang tidak berbeda bermakna yaitu 53,6 dan 53,6 tahun. Variabel klinis yang berbeda bermakna antara 2 kelompok adalah stadium FIGO (p=0.0024). Pola VOC pasien sebelum dan sesudah terapi menunjukkan tipe dispersing, berbeda dengan pasien sehat yang mayoritas bertipe clustering. Sebanyak 79 sampel pasien sebelum kemoterapi dan 84 sesudah terapi didapatkan, analisis VOC menunjukkan sensor yang paling berpengaruh adalah S5 menyumbang porsi terbesar dalam memprediksi data sesudah terapi sedangkan S3 menyumbang porsi terbesar dalam meprediksi data sebelum terapi. Akurasi menggunakan model random forest untuk membedakan kedua kelompok mencapai 81,9%.

Kesimpulan: Secara umum, pola urin kanker servik sesudah terapi menyerupai urin sebelum terapi. Namun, secara spesifik sensor S5 (Carbon monoxide, ethanol, hydrogen, isobutane, methane, and propane) mampu memprediksi pasien kanker sesudah terapi sedangkan S3 (ethanol, hydrogen, isobutane, dan methane) memprediksi urin sebelum terapi. Validasi interna menilai akurasi e-nose dalam membedakan kelompok sebelum dan sesudah terapi mencapai 81,9%. Hal ini didukung oleh studi pendahulu yang mampu membedakan pola dan senyawa VOC antara pasien sehat dan kanker serviks dengan baik. Perlu analisis validasi klinis lanjutan untuk menilai sensitivitas dan spesifitas e-nose untuk studi lanjut.

 Kata Kunci: Kanker Servik, Terapi, VOC, electronic nose

Background: Methods of examining volatile biomarkers have been developed because they are more convenient, inexpensive and non-invasive but still reliable. In the case of cervical cancer, monitoring the response to therapy can be seen through biomarkers in blood or urine and can be useful as a new prognostic biomarker modality to reduce cervical cancer mortality in Indonesia Aim: To identify pattern of VOCs (aromatic components, alkanes, aldehydes, acids, ketones, alcohols and esters) in urine samples of cervical cancer patients before and after therapy.

Methods: An observational study with a case-control approach using urine samples, the subjects were divided into 2 groups, namely the pre-therapy and post-therapy groups. Urine samples were taken 2 times (first void and midstream urine) at different times, then the VOC pattern was analyzed using an electronic nose. There are 7 gas sensors on the e-nose then an analysis of data distribution and classification methods is carried out using XGBOOST. Statistical analysis used Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) to see distribution patterns between the 2 groups. Evaluation with ROC curve and SHAP analysis to see the difference in predictions between the two groups.

Results: Ten patients each group for pre therapy and post therapy groups were enrolled and had significant difference for FIGO stage (p=0.024). The VOC pattern of patients before and after therapy showed the dispersing type, in contrast to the majority of healthy patients were clustering type. A total of 79 patient samples before chemotherapy and 84 after therapy were obtained, VOC analysis showed that the most influential sensor was S5 contributing the largest portion in predicting data after therapy while S3 contributed the largest portion in predicting data before therapy. The accuracy using the random forest model as internal validation to distinguish between the two groups reached 81.9%.

Conclusion: Volatile Organic Compound (VOC) in urine samples of cervical cancer patients before and after therapy had similar patterns generally, but e-nose detection showed S5 (Carbon monoxide, ethanol, hydrogen, isobutane, methane, and propane) could predict samples after therapy while S3 (ethanol, hydrogen, isobutane, and methane) predict before therapy. The accuracy of e-nose in differentiating two groups reached 81.9%. This is supported by preliminary studies which were able to distinguish VOC patterns and compounds between healthy patients and cervical cancer. A clinical validation assessment should be obtained for further studies. Keywords: Cervical cancer, Therapy, VOC, electronic nose

Kata Kunci : Cervical cancer, Therapy, VOC, electronic nose