FACT-G

FACT-G in Cancer Trials and Economic Evaluation

The Functional Assessment of Cancer Therapy General, usually shortened to FACT-G, is the general core of the FACIT family of questionnaires, and oncology trials use it to record health-related quality of life as patients report it. Its answers add up to a total score and four well-being subscale scores, none of which carries preference weights, so a health economist who wants a QALY from a FACT-G trial needs a second step. That step is either a mapping function to EQ-5D utilities or the FACT-8D, a preference-based classification built from nine FACT-G items and valued with general population surveys. Cella and colleagues reported the development of FACT-G version 2 in 1993, after a validation programme that involved 854 patients with cancer and 15 oncology specialists.

Development of the general measure and the current version

Cella and colleagues started with a pool of 370 overlapping items for breast, lung and colorectal cancer. Preselected criteria cut the pool to 38 items, and factor and scaling analyses of those items in 545 patients with mixed cancer diagnoses produced the 28-item FACT-G, version 2.

That 1993 version gave a total score and subscale scores for physical, functional, social and emotional well-being, plus a subscale for satisfaction with the treatment relationship. The authors reported that the scale discriminated between patients by stage of disease, performance status rating and hospitalisation status, and that it was sensitive to change over time.

The questionnaire in use today is version 4. FACIT.org lists it as 27 items covering four domains of health-related quality of life in cancer patients, which are physical, social or family, emotional and functional well-being. The treatment relationship subscale of the 1993 version is not among them. FACIT.org gives the recall period as the past 7 days, the response format as a 5-point Likert-type scale and the completion time as 5 to 10 minutes. A 7-item short form, the FACT-G7, is also listed.

Scoring the subscales and the total score

FACT-G scoring is a summation, so a higher score always means better quality of life once the negatively worded items have been turned round. The FACIT scoring guidance reverses each negatively stated item by subtracting the response from 4 and then sums the raw item scores, each 0 to 4, within each subscale.

Missing answers are handled by proration. FACIT.org allows a subscale score to be prorated from the average of the answered items, provided more than 50% of the subscale's items were answered, so a 7-item subscale needs a minimum of 4 of 7 items. The total is the sum of the subscale scores.

$$S_k = \frac{n_k}{a_k} \sum_{i=1}^{a_k} x_{ik}$$

where $S_k$ is the score for subscale $k$, $n_k$ is the number of items in that subscale, $a_k$ is the number of items the patient answered, and $x_{ik}$ is the score for answered item $i$ after any reversal, on the 0 to 4 scale. With 27 items scored 0 to 4, the total can run between 0 and 108.

The total therefore treats every item as equally important and every step between response categories as equal. Those are conventions of a summated rating scale, and they say nothing about how much the public or patients would give up to avoid a particular problem.

The FACIT system and its disease-specific modules

The FACT-G sits inside a larger measurement system. Webster, Cella and Yost described the Functional Assessment of Chronic Illness Therapy (FACIT) system in 2003 as a collection of quality of life questionnaires targeted to the management of chronic illness, with the FACT-G as its general core and suited to patients with any form of cancer. They also noted that extensions of the questionnaire had been validated in other chronic illnesses.

Disease and symptom modules are added to the 27 core items. The FACT-Hep, for example, consists of the FACT-G plus an 18-item hepatobiliary subscale. The FACIT.org library now describes the system as more than 100 distinct self-report questionnaires with more than 700 unique items for adults. A trial that reports a module such as the FACT-Hep therefore also holds the FACT-G core, the part that the FACT-8D and the mapping studies described below draw on.

Interpreting change and minimally important differences

A difference in FACT-G total score has to be judged against what patients regard as a meaningful change, the idea behind a minimal clinically important difference. Estimates are specific to a population and a method, so the source of any threshold matters as much as its size.

Eton and colleagues estimated minimally important differences for the breast cancer version in 2004, using a phase III trial in metastatic breast cancer (739 patients) and a pain study (129 patients). They combined distribution-based criteria, one third and one half of a standard deviation and one standard error of measurement, with anchor-based comparisons against performance status, current pain and response to treatment. Scores tracked performance status and pain but not treatment response, and the combined estimate for the FACT-G total was 5 to 6 points.

Direction of change matters too. Cella, Hahn and Dineen compared FACT score changes with retrospective global ratings of change after 2 months in 308 cancer patients, and reported in 2002 that patients who said they had worsened showed larger score changes than those who reported a comparable improvement. Correlations between actual score changes and patients' ratings were modest, about 0.07 to 0.35. A threshold estimated for improvement may therefore not apply to deterioration, and none of these estimates says what a change is worth in QALY terms.

Routes that turn FACT-G answers into utilities

A cost-utility analysis needs a health state utility value for each health state in a model, on a scale where full health is 1 and dead is 0. A FACT-G total of 54 out of 108 is not a utility of 0.5, because neither end of the summated scale is anchored on full health or dead and no preferences were used to weight its items.

The FACT-8D preference-based classification

King, Norman and colleagues published the FACT-8D in 2021 as a multi-attribute utility instrument derived from the FACT-G. Item selection drew on psychometric analysis of existing FACT-G data and on patient input. Shiroiwa and colleagues describe the classification as nine of the 27 FACT-G items forming eight dimensions, which are nausea, pain, fatigue, sleep, work, worry, sadness and support from family or friends.

The first value set came from an online discrete choice experiment with members of an Australian general population panel, and 1,644 of the 1,737 panel members who accessed the survey completed at least one choice pair. The largest utility decrements were for the highest levels of pain (minus 0.40) and nausea (minus 0.28), and the worst health state was valued at minus 0.54, considerably worse than dead. Further value sets have since been published, and Shiroiwa and colleagues reported in 2025 that health states were scored consistently higher in the United States and Australia than in their Japanese set, estimated from 2,320 participants.

The FACT-8D needs no new questionnaire, so it can be scored from FACT-G data already collected in a trial. It plays the same role for the FACT-G that the EORTC QLU-C10D plays for the EORTC QLQ-C30, and the two classifications have different dimensions because their parent questionnaires differ. FACT-8D utilities, whichever country's value set scores them, are not UK EQ-5D-5L values, so under PMG51 they also fall outside the NICE reference case and would sit in supporting or sensitivity analyses.

Mapping FACT-G scores to the EQ-5D

The second route is a mapping study in patients who completed both the FACT-G and a preference-based measure. Teckle and colleagues used 367 patients with breast, colorectal or lung cancer, split into development and validation samples of 184 and 183, and found that the physical, functional and emotional well-being scores best explained EQ-5D and SF-6D utilities. A censored least absolute deviations model performed best for the EQ-5D, with a root mean squared error of 0.095.

Hagiwara and colleagues mapped both the FACT-G and the QLQ-C30 to the EQ-5D-5L index with the Japanese value set, in 908 patients for the FACT-G analysis. They recommended two-part beta regression for direct mapping and ordinal logistic regression for indirect mapping, with cross-validated root mean squared errors of 0.121 and 0.120 for the FACT-G, against 0.101 for the QLQ-C30. Predicted and observed values agreed except in very good health for both questionnaires and in poor health for the FACT-G alone.

The value set inside a mapping algorithm decides where the algorithm can be used. NICE's interim methods statement on implementing the EQ-5D-5L value set (PMG51, 27 August 2026) says the UK EQ-5D-5L value set should be used to derive all utility values for adults in reference-case analyses, so an algorithm that predicts EQ-5D-3L values or Japanese EQ-5D-5L values does not meet that requirement as it stands.

Illustrative worked example of scoring and utility choice

All figures below are illustrative. A patient leaves two items blank in a 7-item subscale, and the five answered items, after reversal of negatively worded items, score 3, 3, 4, 2 and 3.

1. Check that proration is allowed. Five of seven items is about 71% of the subscale, which is more than 50%, so the subscale can be prorated.

2. Prorate the subscale. The answered items sum to 3 + 3 + 4 + 2 + 3 = 15. Applying the formula, 15 × 7 = 105, and 105 ÷ 5 = 21, so the subscale score is 21.

3. Calculate the total. If the patient's other three subscale scores are 20, 16 and 18, the FACT-G total is 21 + 20 + 16 + 18 = 75 out of a possible 108.

4. Turn a utility difference into QALYs. Suppose a trial reports that a new treatment improves FACT-G scores, and the FACT-8D gives mean utilities of 0.72 and 0.66 in the two arms over 1.5 years. The gain is 0.72 − 0.66 = 0.06, and 0.06 × 1.5 = 0.09 QALYs. At an incremental cost of GBP 4,500, the cost per QALY is 4,500 ÷ 0.09 = 50,000. For a NICE submission this figure would be a supporting analysis, because FACT-8D utilities are not UK EQ-5D-5L values.

5. Repeat with a mapped utility. If a mapping algorithm that predicts UK EQ-5D-5L values instead gives a utility gap of 0.04, the gain is 0.04 × 1.5 = 0.06 QALYs, and 4,500 ÷ 0.06 = 75,000, so the cost per QALY is GBP 75,000.

The total of 75 is a descriptive score and says nothing about the patient's utility. The same trial difference gives a ratio of GBP 50,000 or GBP 75,000 per QALY depending on whether the FACT-8D or a mapping algorithm supplies the utilities, so the choice of route can move a technology across a decision threshold. Under PMG51 only the mapped UK EQ-5D-5L figure could be the reference case, and the FACT-8D result would be shown as a sensitivity analysis.

How the FACT-G differs from nearby measures

The FACT-G is a patient-reported outcome measure and a disease-specific instrument written for people with cancer, but general across cancer types, with site-specific concerns left to the modules. It is a non-preference-based measure, since its scores come from summed ratings and not from a valuation exercise.

The closest rival is the EORTC QLQ-C30, the core of the EORTC modular system. Both questionnaires have a cancer core and add-on modules, but the QLQ-C30 reports 15 separate scores on 0 to 100 scales and its standard scoring gives no single total, while the FACT-G gives a single summed total with four subscales, including a social or family well-being domain. In the same Japanese patients, Hagiwara and colleagues found mapping errors for the FACT-G slightly larger than for the QLQ-C30.

The EQ-5D is a generic instrument with value sets that produce utilities directly and can be compared across disease areas, while the FACT-G can pick up concerns such as nausea, worry or family support that a generic measure does not name. The FACT-8D is a scoring system applied to FACT-G answers and has no questionnaire of its own.

Limitations and common misreadings

Most errors with FACT-G data in an economic evaluation come from treating a summated profile as if it were a preference-weighted index. Others come from lifting thresholds or algorithms out of the population and version in which they were estimated.

• Reading the total as a utility. A total of 81 out of 108 is 75% of the maximum score, which has no link to a utility of 0.75.
• Borrowing a threshold from another population. The 5 to 6 point estimate of Eton and colleagues comes from breast cancer data, and improvement and worsening may need different thresholds.
• Ignoring missing items. A total built from prorated subscales assumes the skipped items behave like the answered ones, and proration is allowed only when more than half of a subscale's items were answered.
• Treating mapped values as observed. Mapped utilities carry prediction error, Hagiwara and colleagues found poorer agreement in poor health for the FACT-G, and an algorithm built on a United States or Japanese value set does not give UK reference-case utilities.
• Pooling across versions. The 1993 version 2 had 28 items and a treatment relationship subscale, so totals from older trials are not on the same footing as version 4 totals.

Sources

• Cella DF, et al. The Functional Assessment of Cancer Therapy scale: development and validation of the general measure. Journal of Clinical Oncology. 1993;11(3):570-579. doi:10.1200/JCO.1993.11.3.570. https://pubmed.ncbi.nlm.nih.gov/8445433/
• Cella D, Hahn EA, Dineen K. Meaningful change in cancer-specific quality of life scores: differences between improvement and worsening. Quality of Life Research. 2002;11(3):207-221. doi:10.1023/A:1015276414526. https://pubmed.ncbi.nlm.nih.gov/12074259/
• Eton DT, et al. A combination of distribution- and anchor-based approaches determined minimally important differences (MIDs) for four endpoints in a breast cancer scale. Journal of Clinical Epidemiology. 2004;57(9):898-910. https://pubmed.ncbi.nlm.nih.gov/15504633/
• FACIT.org. FACT-G, Functional Assessment of Cancer Therapy General (version 4). Accessed 17 October 2026. https://www.facit.org/measures/fact-g
• FACIT.org. Scoring. Accessed 17 October 2026. https://www.facit.org/scoring
• FACIT.org. FACIT Searchable Library. Accessed 17 October 2026. https://www.facit.org/facit-searchable-library
• FACIT.org. FACT-Hep. Accessed 17 October 2026. https://www.facit.org/measures/fact-hep
• Hagiwara Y, Shiroiwa T, Taira N, et al. Mapping EORTC QLQ-C30 and FACT-G onto EQ-5D-5L index for patients with cancer. Health and Quality of Life Outcomes. 2020;18:354. doi:10.1186/s12955-020-01611-w
• King MT, Norman R, et al. The Functional Assessment of Cancer Therapy Eight Dimension (FACT-8D), a multi-attribute utility instrument derived from the cancer-specific FACT-General (FACT-G) quality of life questionnaire: development and Australian value set. Value in Health. 2021;24(6):862-873. doi:10.1016/j.jval.2021.01.007
• National Institute for Health and Care Excellence. Interim methods statement: implementing the EQ-5D-5L value set (PMG51). Published 27 August 2026. https://www.nice.org.uk/process/pmg51/chapter/statement-on-using-the-eq-5d-5l-value-set
• Shiroiwa T, King MT, Campbell R, Murata T, Shimozuma K, Fukuda T, Norman R. Japanese value set for the Functional Assessment of Cancer Therapy Eight Dimension (FACT-8D) cancer-specific preference-based quality of life instrument. Health and Quality of Life Outcomes. 2025;23:109. doi:10.1186/s12955-025-02442-3
• Teckle P, et al. Mapping the FACT-G cancer-specific quality of life instrument to the EQ-5D and SF-6D. Health and Quality of Life Outcomes. 2013;11:203. doi:10.1186/1477-7525-11-203
• Webster K, Cella D, Yost K. The Functional Assessment of Chronic Illness Therapy (FACIT) Measurement System: properties, applications, and interpretation. Health and Quality of Life Outcomes. 2003;1:79. doi:10.1186/1477-7525-1-79