The Device and the Drug Are a Single Product
A drug delivery device is not a device that happens to contain a drug. To FDA it is a combination product, and the moment you put a formulation inside your hardware, three things change at once: the review path, the manufacturing quality system, and the definition of a failure. A pump that runs perfectly but delivers 12 percent under the programmed dose has failed. So has an autoinjector whose needle shield works flawlessly but whose glass barrel leaches silicone into a protein formulation over an 18-month shelf life.
That coupling is what makes these programs longer and costlier than a comparable standalone device. Teams that budget for a mechanical product and discover a pharmaceutical program halfway through are the ones that run out of money.
Three Families, Three Engineering Problems
Injectors. Pens, autoinjectors, and on-body injectors are spring, gas, or motor driven systems that push a defined volume through a fine needle in a controlled time. The hard parts are needle insertion depth repeatability, injection time with a viscous biologic (a 20 cP formulation through a 27G needle behaves nothing like water), and confirming to the patient that the full dose went in. Prefilled syringe compatibility dictates the entire mechanical architecture, so the primary container decision comes first, not last.
Infusion and ambulatory pumps. These are electromechanical systems with a control loop, an alarm architecture, and often connectivity. Occlusion detection, air-in-line sensing, free-flow protection, and battery behavior during a fault dominate the risk file. Because they are electrically powered and patient-contacting, they carry the full weight of IEC 60601 electrical safety testing, plus the collateral standards for programmable systems and alarms. Power architecture is not a detail here; the runtime and shutdown behavior of the pack is a safety requirement, which is why battery pack design belongs in the early architecture review.
Inhalers. Metered-dose, dry powder, and soft mist devices are aerosol physics problems dressed as plastic parts. The metric that matters is not volume but delivered dose uniformity and the fine particle fraction that reaches the lung. Powder flow, deagglomeration geometry, moisture ingress, and the patient's own inspiratory flow rate all move that number, and none can be simulated to the accuracy you need. Inhaler programs live or die on cascade impactor testing.
Dose Accuracy Is the Budget Everything Else Spends
Write the dose accuracy specification before anything else, as a tolerance with a statistical basis, not a target. Something like plus or minus 5 percent of label claim at 95 percent confidence over the full temperature and shelf-life range. Then build a stack-up that allocates that budget across every contributor: primary container fill volume, plunger friction and its drift after storage, spring force tolerance, needle bore variation, dead volume, and temperature effects on formulation viscosity.
Most teams underestimate friction drift. A siliconized glass barrel stored for a year can require noticeably more break-loose force than a fresh one, and a spring sized to the fresh value will deliver a slow or incomplete injection at end of shelf life. That is a tolerance stack-up problem with a time axis, and it needs accelerated aging data behind it, not a safety factor pulled from intuition.
The User Is Inside the Control Loop
Delivery devices are handed to patients with tremor, low vision, arthritic grip, or a fresh diagnosis and no training. Use error is a leading cause of delivery device recalls, and FDA expects a human factors validation study with at least 15 participants per distinct user group, run with production-equivalent devices and the real instructions. Formative studies come earlier and cheaper, and they are where you find out that users pull the device off the skin at the click instead of holding for the full ten seconds. This work follows the process described in usability engineering under IEC 62366, and its findings feed straight back into ISO 14971 risk management as design inputs, not as documentation written after the fact.
Materials, Sterility, and Container Closure
Anything touching the formulation is evaluated for extractables and leachables, which is a different and often harsher test program than the patient-contact assessment covered by ISO 10993 biocompatibility testing. Cyclic olefin polymers, specific grades of polypropylene, and fluoropolymer-coated stoppers exist because ordinary commodity resins fail these studies. Expect to lock material grades and even supplier lot sources early, because a resin change late in the program restarts stability testing.
Sterility strategy also constrains design. Ethylene oxide needs gas paths and a defined aeration cycle; gamma degrades many polymers and can discolor or embrittle them; steam is off the table for most assembled electromechanical devices. The tradeoffs are laid out in medical device sterilization methods, and the choice interacts with the primary container, which is often supplied already sterile and cannot be reprocessed.
The Regulatory Path Is Decided by the Primary Mode of Action
For a combination product, FDA assigns a lead review center based on the primary mode of action. A device delivering an already-approved drug in an approved indication may reach market through a 510(k) or a supplement to an existing application. A novel formulation, a new route of administration, or a new indication pulls the program into a drug application with the device as a constituent part, and the timeline roughly doubles. A Request for Designation resolves the question formally when it is unclear. Either way manufacturing falls under 21 CFR Part 4, requiring a quality system that satisfies both device design controls and drug cGMP. Read the FDA approval process for medical devices for the device half, and assume the drug half adds review cycles rather than replacing them.
Practical sequencing advice: fix the primary container and the formulation viscosity range first, run formative human factors on a crude functional model within the first few months, and get an accelerated aging plan started before you have a final design, because stability data is almost always the schedule's critical path.
Start With the Right Architecture
Projects House develops drug delivery hardware end to end, from primary container selection and mechanism design through human factors, verification testing, and transfer to a validated contract manufacturer. Describe your formulation, dose, and target user through our contact form and we will map the pathway and the realistic schedule.