BACKGROUND: Little is known about the content or quality of non-communicable disease (NCD) care in humanitarian settings. Since 2014, Médecins Sans Frontières (MSF) has provided primary-level NCD services in Irbid, Jordan, targeting Syrian refugees and vulnerable Jordanians who struggle to access NCD care through the overburdened national health system. This retrospective cohort study explored programme and patient-level patterns in achievement of blood pressure and glycaemic control, patterns in treatment interruption, and the factors associated with these patterns. METHODS AND FINDINGS: The MSF multidisciplinary, primary-level NCD programme provided facility-based care for cardiovascular disease, diabetes, and chronic respiratory disease using context-adapted guidelines and generic medications. Generalist physicians managed patients with the support of family medicine specialists, nurses, health educators, pharmacists, and psychosocial and home care teams. Among the 5,045 patients enrolled between December 2014 and December 2017, 4,044 eligible adult patients were included in our analysis, of whom 72% (2,913) had hypertension and 63% (2,546) had type II diabetes. Using visits as the unit of analysis, we plotted the following on a monthly basis: mean blood pressure among hypertensive patients, mean fasting blood glucose and HbA1c among type II diabetic patients, the proportion of each group achieving control, mean days of delayed appointment attendance, and the proportion of patients experiencing a treatment interruption. Results are presented from programmatic and patient perspectives (using months since programme initiation and months since cohort entry/diagnosis, respectively). General linear mixed models explored factors associated with clinical control and with treatment interruption. Mean age was 58.5 years, and 60.1% (2,432) were women. Within the programme's first 6 months, mean systolic blood pressure decreased by 12.4 mm Hg from 143.9 mm Hg (95% CI 140.9 to 146.9) to 131.5 mm Hg (95% CI 130.2 to 132.9) among hypertensive patients, while fasting glucose improved by 1.12 mmol/l, from 10.75 mmol/l (95% CI 10.04 to 11.47) to 9.63 mmol/l (95% CI 9.22 to 10.04), among type II diabetic patients. The probability of achieving treatment target in a visit was 63%-75% by end of 2017, improving with programme maturation but with notable seasonable variation. The probability of experiencing a treatment interruption declined as the programme matured and with patients' length of time in the programme. Routine operational data proved useful in evaluating a humanitarian programme in a real-world setting, but were somewhat limited in terms of data quality and completeness. We used intermediate clinical outcomes proven to be strongly associated with hard clinical outcomes (such as death), since we had neither the data nor statistical power to measure hard outcomes. CONCLUSIONS: Good treatment outcomes and reasonable rates of treatment interruption were achieved in a multidisciplinary, primary-level NCD programme in Jordan. Our approach to using continuous programmatic data may be a feasible way for humanitarian organisations to account for the complex and dynamic nature of interventions in unstable humanitarian settings when undertaking routine monitoring and evaluation. We suggest that frequency of patient contact could be reduced without negatively impacting patient outcomes and that season should be taken into account in analysing programme performance.