{"id":11047,"date":"2025-10-30T08:50:51","date_gmt":"2025-10-30T07:50:51","guid":{"rendered":"https:\/\/oytec.es\/la-fda-i-la-ia-farmaceutica-claus-de-lesborrany-2025-sobre-credibilitat-de-models\/"},"modified":"2026-10-17T17:24:40","modified_gmt":"2026-10-17T15:24:40","slug":"la-fda-i-la-ia-farmaceutica-claus-de-lesborrany-2025-sobre-credibilitat-de-models","status":"publish","type":"post","link":"https:\/\/www.oytec.es\/ca\/la-fda-i-la-ia-farmaceutica-claus-de-lesborrany-2025-sobre-credibilitat-de-models\/","title":{"rendered":"La FDA i la IA farmac\u00e8utica: claus de l’Esborrany 2025 sobre credibilitat de models"},"content":{"rendered":"

[et_pb_section fb_built=\u00bb1″ _builder_version=\u00bb4.14.7″ width=\u00bb100%\u00bb max_width=\u00bb76%\u00bb module_alignment=\u00bbleft\u00bb custom_margin=\u00bb|-155px||||\u00bb custom_padding=\u00bb0px|||||\u00bb global_colors_info=\u00bb{}\u00bb][et_pb_row _builder_version=\u00bb4.14.7″ _module_preset=\u00bbdefault\u00bb custom_padding=\u00bb||2px|||\u00bb global_colors_info=\u00bb{}\u00bb][et_pb_column type=\u00bb4_4″ _builder_version=\u00bb4.14.7″ _module_preset=\u00bbdefault\u00bb global_colors_info=\u00bb{}\u00bb][et_pb_text _builder_version=\u00bb4.14.7″ text_font=\u00bbMontserrat||||||||\u00bb text_text_color=\u00bb#000000″ header_font=\u00bbMontserrat||||||||\u00bb header_font_size=\u00bb14px\u00bb header_line_height=\u00bb1.5em\u00bb text_orientation=\u00bbjustified\u00bb custom_padding=\u00bb|75px|18px|1px||\u00bb z_index_tablet=\u00bb500″ global_colors_info=\u00bb{}\u00bb]<\/p>\n

La FDA situa la credibilitat com a condici\u00f3 indispensable per utilitzar IA<\/strong> en decisions reguladores farmac\u00e8utiques, en un marc proporcional al risc que exigeix tra\u00e7abilitat, transpar\u00e8ncia i validaci\u00f3 al llarg de tot el cicle de vida del model. Aquest enfocament ha estat sintetitzat i divulgat per Guillem L\u00f3pez, CSV & QA Manager d’OYTEC<\/strong>, autor de l’an\u00e0lisi \u00abCredibilitat de la IA farmac\u00e8utica: Claus de l’esborrany de la FDA\u00bb<\/strong> publicat al n\u00famero 18 de la revista FARMABIOTEC. <\/h1>\n

[\/et_pb_text][\/et_pb_column][\/et_pb_row][et_pb_row _builder_version=\u00bb4.14.7″ _module_preset=\u00bbdefault\u00bb min_height=\u00bb481.5px\u00bb custom_margin=\u00bb-20px||-50px||false|false\u00bb global_colors_info=\u00bb{}\u00bb][et_pb_column type=\u00bb4_4″ _builder_version=\u00bb4.14.7″ _module_preset=\u00bbdefault\u00bb global_colors_info=\u00bb{}\u00bb][et_pb_image src=\u00bbhttps:\/\/oytec.es\/wp-content\/uploads\/Captura-revista-Guillem-800×561.png\u00bb title_text=\u00bbOYTEC-FDA-Guillem\u00bb align=\u00bbcenter\u00bb align_tablet=\u00bbcenter\u00bb align_phone=\u00bb\u00bb align_last_edited=\u00bbon|desktop\u00bb _builder_version=\u00bb4.14.7″ max_width=\u00bb40%\u00bb module_alignment=\u00bbcenter\u00bb min_height_tablet=\u00bb500px\u00bb min_height_phone=\u00bb500px\u00bb min_height_last_edited=\u00bboff|desktop\u00bb custom_margin=\u00bb|44px|37px|||\u00bb z_index_tablet=\u00bb500″ global_colors_info=\u00bb{}\u00bb][\/et_pb_image][\/et_pb_column][\/et_pb_row][et_pb_row column_structure=\u00bb3_4,1_4″ _builder_version=\u00bb4.14.7″ _module_preset=\u00bbdefault\u00bb custom_padding=\u00bb2px||2px|||\u00bb global_colors_info=\u00bb{}\u00bb][et_pb_column type=\u00bb3_4″ _builder_version=\u00bb4.14.7″ _module_preset=\u00bbdefault\u00bb global_colors_info=\u00bb{}\u00bb][et_pb_text _builder_version=\u00bb4.14.7″ text_font=\u00bbMontserrat||||||||\u00bb text_text_color=\u00bb#000000″ header_2_font_size=\u00bb22px\u00bb background_color=\u00bbRGBA(255,255,255,0)\u00bb custom_padding=\u00bb|0px||65px||\u00bb z_index_tablet=\u00bb500″ global_colors_info=\u00bb{}\u00bb]<\/p>\n

Exig\u00e8ncies reguladores<\/span><\/h2>\n

L’esborrany \u201cConsiderations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products\u201d<\/strong>, em\u00e8s el gener de 2025, defineix que s’ha d’establir un marc de credibilitat en els models ajustat al risc d’aquest proc\u00e9s, tenint en compte tant la influ\u00e8ncia que el model tindr\u00e0 en la presa de decisions, com les conseq\u00fc\u00e8ncies que tindria la presa d’una decisi\u00f3 incorrecta (Model influence vs. Decision consequence).
\n[\/et_pb_text][et_pb_text _builder_version=\u00bb4.14.7″ text_font=\u00bbMontserrat||||||||\u00bb text_text_color=\u00bb#000000″ header_2_font_size=\u00bb22px\u00bb background_color=\u00bbRGBA(255,255,255,0)\u00bb width=\u00bb100%\u00bb custom_padding=\u00bb|0px||65px||\u00bb z_index_tablet=\u00bb500″ global_colors_info=\u00bb{}\u00bb]Es proposa la constituci\u00f3 d’un pla per establir la credibilitat<\/strong> i un informe corresponent en el qual es demostri la idone\u00eftat del model per al context d’\u00fas.<\/p>\n

Entre els elements clau del pla haur\u00e0 d’estar inclosa la definici\u00f3 de la pregunta d’inter\u00e8s, el context d’\u00fas, la garantia de risc i la descripci\u00f3 t\u00e8cnica de com s’ha desenvolupat el model<\/strong> (incloent-hi dades, l’entrenament, les m\u00e8triques d’avaluaci\u00f3, entre d’altres).<\/strong> Una vegada establert, s’haur\u00e0 d’executar i registrar apropiadament l’adequaci\u00f3 dels resultats al context d’\u00fas.
\n[\/et_pb_text][\/et_pb_column][et_pb_column type=\u00bb1_4″ _builder_version=\u00bb4.14.7″ _module_preset=\u00bbdefault\u00bb global_colors_info=\u00bb{}\u00bb][et_pb_image src=\u00bbhttps:\/\/oytec.es\/wp-content\/uploads\/Grafica-riesgo-Guillem.png\u00bb title_text=\u00bbGr\u00e0fica-risc-Guillem\u00bb _builder_version=\u00bb4.14.7″ _module_preset=\u00bbdefault\u00bb transform_scale=\u00bb95%|95%\u00bb transform_translate=\u00bb-24px|28px\u00bb global_colors_info=\u00bb{}\u00bb][\/et_pb_image][\/et_pb_column][\/et_pb_row][et_pb_row module_class=\u00bb et_pb_row_fullwidth\u00bb _builder_version=\u00bb3.25″ width=\u00bb89%\u00bb width_tablet=\u00bb80%\u00bb width_phone=\u00bb\u00bb width_last_edited=\u00bbon|desktop\u00bb max_width=\u00bb89%\u00bb max_width_tablet=\u00bb80%\u00bb max_width_phone=\u00bb\u00bb max_width_last_edited=\u00bbon|desktop\u00bb min_height=\u00bb877.5px\u00bb custom_padding=\u00bb8px|||||\u00bb z_index_tablet=\u00bb500″ make_fullwidth=\u00bbon\u00bb global_colors_info=\u00bb{}\u00bb][et_pb_column type=\u00bb4_4″ _builder_version=\u00bb3.25″ custom_padding=\u00bb|||\u00bb global_colors_info=\u00bb{}\u00bb custom_padding__hover=\u00bb|||\u00bb][et_pb_image src=\u00bbhttps:\/\/oytec.es\/wp-content\/uploads\/Diagrama-ejemplo-de-entrenamiento-de-una-red-neuronal-800×435.png\u00bb title_text=\u00bbDiagrama exemple d'entrenament d'una xarxa neuronal\u00bb align=\u00bbcenter\u00bb _builder_version=\u00bb4.14.7″ _module_preset=\u00bbdefault\u00bb custom_margin=\u00bb27px|||||\u00bb global_colors_info=\u00bb{}\u00bb][\/et_pb_image][et_pb_text _builder_version=\u00bb4.14.7″ text_font=\u00bbMontserrat||||||||\u00bb text_text_color=\u00bb#000000″ header_2_font_size=\u00bb22px\u00bb background_color=\u00bbRGBA(255,255,255,0)\u00bb min_height=\u00bb72px\u00bb custom_margin=\u00bb-18px|||||\u00bb custom_padding=\u00bb19px|0px||0px||\u00bb z_index_tablet=\u00bb500″ global_colors_info=\u00bb{}\u00bb]<\/p>\n

Com tot sistema GMP cr\u00edtic, una vegada es troba en fase productiva, no ha de faltar la planificaci\u00f3 de manteniment i revisi\u00f3 peri\u00f2dica del rendiment<\/strong>. Per\u00f2, en el cas dels models d’IA \u00e9s especialment important pel fet de ser sistemes que aprenen del nostre proc\u00e9s, i aquest est\u00e0 evolucionant constantment (fenomen del Datadrift). \u00c9s crucial assegurar-nos que es mant\u00e9 aquesta adequaci\u00f3 al context d’\u00fas mitjan\u00e7ant la monitoritzaci\u00f3 i el reentrenament. <\/p>\n

[\/et_pb_text][et_pb_text _builder_version=\u00bb4.14.7″ text_font=\u00bbMontserrat||||||||\u00bb text_text_color=\u00bb#000000″ header_2_text_color=\u00bb#000000″ header_2_font_size=\u00bb22px\u00bb background_color=\u00bbRGBA(255,255,255,0)\u00bb custom_padding=\u00bb|2px||0px||\u00bb z_index_tablet=\u00bb500″ global_colors_info=\u00bb{}\u00bb]<\/p>\n

Aportaci\u00f3 d’OYTEC i el seu autor<\/h2>\n

Des d’OYTEC, Guillem L\u00f3pez destaca la necessitat d’identificar amb precisi\u00f3 la pregunta d’inter\u00e8s (QoI) i el context d’\u00fas<\/strong> per poder establir un Credibility Assessment Plan<\/em> i un Credibility Assessment Report <\/em>que connectin evid\u00e8ncia t\u00e8cnica amb l’ajust a les necessitats de planta, integrant les normatives GxP i marcs de refer\u00e8ncia com GAMP5 Ed.2 i ISPE per assegurar reproductibilitat i control del canvi.<\/p>\n

 <\/p>\n

Com OYTEC implementa la IA<\/h2>\n

OYTEC implementa solucions d’IA en entorns GxP juntament amb partners tecnol\u00f2gics com Aizon<\/strong>, alineant plans i informes de credibilitat amb el marc de la FDA.<\/p>\n

Per aprofundir, podeu consultar l’article de Guillem L\u00f3pez a FARMABIOTEC (n\u00fam. 18, p\u00e0gs. 66\u201371)<\/a> i contactar amb els nostres experts.<\/a><\/p>\n

[\/et_pb_text][\/et_pb_column][\/et_pb_row][\/et_pb_section][et_pb_section fb_built=\u00bb1″ _builder_version=\u00bb3.22″ custom_padding=\u00bb9px|||||\u00bb z_index_tablet=\u00bb500″ global_colors_info=\u00bb{}\u00bb][et_pb_row module_class=\u00bb et_pb_row_fullwidth\u00bb _builder_version=\u00bb3.25″ width=\u00bb89%\u00bb width_tablet=\u00bb80%\u00bb width_phone=\u00bb\u00bb width_last_edited=\u00bbon|desktop\u00bb max_width=\u00bb89%\u00bb max_width_tablet=\u00bb80%\u00bb max_width_phone=\u00bb\u00bb max_width_last_edited=\u00bbon|desktop\u00bb custom_padding=\u00bb11px|0px|38.0469px|0px|false|false\u00bb z_index_tablet=\u00bb500″ make_fullwidth=\u00bbon\u00bb global_colors_info=\u00bb{}\u00bb][et_pb_column type=\u00bb4_4″ _builder_version=\u00bb3.25″ custom_padding=\u00bb|||\u00bb global_colors_info=\u00bb{}\u00bb custom_padding__hover=\u00bb|||\u00bb][et_pb_divider _builder_version=\u00bb3.21.4″ z_index_tablet=\u00bb500″ global_colors_info=\u00bb{}\u00bb][\/et_pb_divider][et_pb_text _builder_version=\u00bb3.27.4″ text_font=\u00bbRoboto||||||||\u00bb text_text_color=\u00bb#000000″ z_index_tablet=\u00bb500″ global_colors_info=\u00bb{}\u00bb]\u00daltimes Not\u00edcies<\/strong>
\n[\/et_pb_text][et_pb_blog fullwidth=\u00bboff\u00bb posts_number=\u00bb3″ include_categories=\u00bb3,5,17,7″ show_more=\u00bbon\u00bb show_date=\u00bboff\u00bb _builder_version=\u00bb3.21.4″ z_index_tablet=\u00bb500″ box_shadow_style=\u00bbpreset3″ global_colors_info=\u00bb{}\u00bb][\/et_pb_blog][et_pb_button button_url=\u00bbhttps:\/\/oytec.es\/blog-oytec\/\u00bb button_text=\u00bbM\u00e9s not\u00edcies\u00bb button_alignment=\u00bbcenter\u00bb _builder_version=\u00bb4.14.7″ z_index_tablet=\u00bb500″ global_colors_info=\u00bb{}\u00bb][\/et_pb_button][\/et_pb_column][\/et_pb_row][\/et_pb_section][et_pb_section fb_built=\u00bb1″ _builder_version=\u00bb4.9.3″ _module_preset=\u00bbdefault\u00bb background_color=\u00bb#eef1f7″ background_image=\u00bbhttps:\/\/www.oytec.es\/wp-content\/uploads\/2021\/03\/Footer_Oytec_400.jpg\u00bb background_enable_image=\u00bboff\u00bb custom_margin=\u00bb||-55px|||\u00bb custom_padding=\u00bb3px||3px|||\u00bb saved_tabs=\u00bball\u00bb global_colors_info=\u00bb{}\u00bb][et_pb_row column_structure=\u00bb1_5,1_5,1_5,1_5,1_5″ _builder_version=\u00bb4.14.7″ _module_preset=\u00bbdefault\u00bb module_alignment=\u00bbcenter\u00bb global_colors_info=\u00bb{}\u00bb][et_pb_column type=\u00bb1_5″ _builder_version=\u00bb4.9.2″ _module_preset=\u00bbdefault\u00bb global_colors_info=\u00bb{}\u00bb][et_pb_text _builder_version=\u00bb4.14.7″ _module_preset=\u00bbdefault\u00bb custom_margin=\u00bb-9px||20px||false|false\u00bb global_colors_info=\u00bb{}\u00bb]BARCELONA<\/strong>
\n[\/et_pb_text][et_pb_text admin_label=\u00bbTexto\u00bb _builder_version=\u00bb4.14.7″ text_font=\u00bbRoboto||||||||\u00bb text_text_color=\u00bb#636363″ header_3_font=\u00bb|||on|||||\u00bb header_6_text_color=\u00bb#7c7c7c\u00bb background_layout=\u00bbdark\u00bb custom_padding=\u00bb0px|||||\u00bb z_index_tablet=\u00bb500″ global_colors_info=\u00bb{}\u00bb]<\/p>\n