為什麼一切從「好問題」開始？
論文寫不好，多數情況不是寫作技巧的問題，而是研究問題本身就含糊。一個含糊的問題會讓緒論說不出 gap、方法選不對工具、結果不知道該回答什麼、討論變成感想文。
好的研究問題具備三個特徵：可被檢驗 (testable)、有意義 (significant)、可行 (feasible)。三者缺一，論文都會出問題。
Most weak papers don't fail because of writing — they fail because the research question itself was vague. A vague question means the introduction has no clear gap, the methods don't fit, the results have nothing specific to answer, and the discussion becomes commentary.
A good research question is testable, significant, and feasible. Lacking any one of these breaks the paper.
一、PICO：把問題結構化
PICO 是循證醫學裡用來把問題結構化的工具，但它對任何「比較」型的生醫研究都適用。也常擴充為 PICOS（加 Study design）或 PICOTS（加 Time、Setting）。
PICO originated in evidence-based medicine for structuring questions, but it works for any comparative biomedical study. It often extends to PICOS (+ Study design) or PICOTS (+ Time, Setting).
P — Population / Problem
研究對象是誰？年齡、性別、診斷、細胞型別、物種…越具體越好。
❌「癌症病人」
✅「III–IV 期非小細胞肺癌、未接受過免疫治療的成人病人」
Who are your subjects? Age, sex, diagnosis, cell type, species — the more specific, the better.
❌ "cancer patients"
✅ "treatment-naïve adults with stage III–IV non-small-cell lung cancer"
I — Intervention / Exposure
你「介入」或「觀察」的因子是什麼？藥物、治療、暴露、基因敲除、scRNA-seq 處理流程…
❌「免疫治療」
✅「Pembrolizumab 200 mg 每 3 週一次，共 4 週期」
The factor you intervene with or observe — drug, treatment, exposure, gene knockout, scRNA-seq pipeline.
❌ "immunotherapy"
✅ "Pembrolizumab 200 mg every 3 weeks for 4 cycles"
C — Comparator
你拿來比較的對照組是什麼？安慰劑、標準療法、wild-type、健康對照…
❌「沒有比較」
✅「化療 (carboplatin + paclitaxel) 標準療法」
What you compare against — placebo, standard care, wild-type, healthy controls.
❌ "no comparison"
✅ "standard chemotherapy (carboplatin + paclitaxel)"
O — Outcome
你想測量什麼結果？必須事先定義並可量化。
❌「病人狀況改善」
✅「24 個月 progression-free survival (PFS) 比例」
What outcome you measure — must be pre-specified and quantifiable.
❌ "patient improvement"
✅ "24-month progression-free survival (PFS) rate"
二、FINER：你的問題值得做嗎？
PICO 幫你把問題寫清楚；FINER 幫你判斷該不該做這個問題。Hulley 等人在《Designing Clinical Research》中提出此 5 項判準。
PICO clarifies the question; FINER tells you whether it's worth doing. Hulley et al. (Designing Clinical Research) proposed these 5 criteria.
| FINER | 自問 | 紅燈 |
|---|---|---|
| Feasible | 樣本量、預算、時間、技術都到位嗎？Sample size, budget, time, expertise — all available? | 需要 1000 例稀有疾病樣本，但你只有 50 例Needs 1000 rare-disease samples; you have 50 |
| Interesting | 研究者、領域、社會關心嗎？Do you, the field, and society care? | 「我有空所以做這個」"I'm doing this because I have free time" |
| Novel | 能在既有文獻上加些什麼？確認、推翻、延伸？Does it add to existing literature — confirm, refute, extend? | 主題已有 3 篇 Cell 論文、結論一致Topic has 3 Cell papers with the same conclusion already |
| Ethical | 能通過 IRB / IACUC？符合 Helsinki / Belmont？Will it pass IRB / IACUC? Follow Helsinki / Belmont? | 無法獲得受試者知情同意Cannot obtain informed consent |
| Relevant | 能影響臨床實踐、政策、後續研究？Will it affect clinical practice, policy, or follow-up research? | 「結果不會改變任何決策」"The result wouldn't change any decision" |
三、Aim / Objective / Hypothesis
三者經常被混用，導致緒論最後一段交代不清。記住層級：Aim（最廣）→ Objective（具體可達成步驟）→ Hypothesis（可被否證的預測）。
These three are often muddled, leaving the last paragraph of the introduction unclear. Hierarchy: Aim (broadest) → Objective (concrete steps) → Hypothesis (falsifiable prediction).
| Aim (目標) | Objective (細目) | Hypothesis (假設) | |
|---|---|---|---|
| 性質 | 寬泛意圖Broad intent | 具體可量化步驟Concrete measurable steps | 可被否證的預測Falsifiable prediction |
| 動詞 | investigate / explore / understand | quantify / compare / identify | will / increases / decreases |
| 例 | 了解 IFN-γ 對 PBMC 的免疫調控Understand IFN-γ immune regulation in PBMCs | (1) 量化 ISG 表達；(2) 鑑定上調超過 2× 的轉錄因子(1) Quantify ISG expression; (2) identify TFs upregulated >2× | IFN-γ 處理會使 ISG 表達在 24 小時內顯著上升 (≥4×, p<0.05)IFN-γ will significantly upregulate ISG expression within 24 h (≥4×, p<0.05) |
四、好壞研究問題對照
❌ 模糊版
「我想研究腸道菌與肥胖的關係。」
"I want to study the relationship between gut microbiota and obesity."
✅ 可檢驗版
「在 18–45 歲的台灣成人中 (P)，Akkermansia muciniphila 相對豐度 (I) 與身體質量指數 (BMI) (O) 是否呈負相關？並比較體重正常 (BMI 18.5–24) 與肥胖 (BMI ≥27) 兩組 (C)。」
"In Taiwanese adults aged 18–45 (P), is relative Akkermansia muciniphila abundance (I) negatively correlated with body mass index (O), comparing normal-weight (BMI 18.5–24) and obese (BMI ≥27) groups (C)?"
❌ 無法檢驗
「Single-cell 技術在癌症研究中很重要。」
(這是觀點，不是問題。沒有變數、沒有對照、無法用數據回答。)
"Single-cell technology is important in cancer research."
(This is an opinion, not a question. No variables, no comparison, no data can answer it.)
✅ 可量化
「相較於 bulk RNA-seq，scRNA-seq (10x v3) 是否能在三陰性乳癌樣本中多偵測出至少 30% 的稀有 (<1%) 細胞亞群？」
"Does scRNA-seq (10x v3) detect at least 30% more rare (<1%) cell subpopulations than bulk RNA-seq in triple-negative breast cancer samples?"
五、研究問題定型決策樹
🌳 一週內鎖定研究問題的流程
六、可直接套用的問題與假設範本
# 比較型 / Comparative In [POPULATION], does [INTERVENTION] compared with [COMPARATOR] result in [OUTCOME] over [TIMEFRAME]? # 關聯型 / Associative Among [POPULATION], is [EXPOSURE] associated with [OUTCOME], after adjusting for [CONFOUNDERS]? # 預測型 / Prognostic In [POPULATION], can [BIOMARKER / MODEL] predict [OUTCOME] with AUC > 0.75 in an external validation cohort? # 方法學 / Methodological Compared with [CURRENT METHOD], does [NEW METHOD] improve [METRIC] by at least [THRESHOLD] on [BENCHMARK]?
# 方向性假設 / Directional H1: [Variable A] will increase / decrease [Variable B] by at least [effect size] in [population]. # 非方向性假設 / Non-directional H1: There is a significant difference in [outcome] between [group 1] and [group 2]. # 虛無假設 / Null H0: There is no difference in [outcome] between [group 1] and [group 2].
# Aim (one sentence, broad) The aim of this study is to [verb: investigate / characterize / compare] [topic] in [population/system]. # Objectives (numbered, measurable) Objective 1: To [verb: quantify / identify / compare] [X] using [method]. Objective 2: To [verb] [Y] using [method]. Objective 3: To [verb] [Z] using [method].
📝 自我檢測
1. 以下哪個研究問題最符合 PICO 結構，且能寫成可檢驗的論文？
1. Which research question best matches the PICO structure and is testable?
2. FINER 五個準則中，「N (Novel)」的合理判斷是？
2. Among the FINER criteria, the correct interpretation of "N (Novel)" is?
3. 關於 Aim、Objective、Hypothesis 的區辨，下列何者最正確？
3. Which best describes the distinction between Aim, Objective, and Hypothesis?